Joanne got the email on August 20th. Subject line: “Sponsored Discovery Is Being Phased Out.”

She runs a mid-sized household goods brand out of Kuala Lumpur — $130,000 GMV/month across Shopee, Lazada, and TikTok Shop. Lazada was 22% of that mix two years ago. Today it’s 11%. Her team was already drafting the Q4 budget when the email landed, and the migration notice — move your automated campaigns to Sponsored Max by September 18th or we’ll move them for you — turned a routine budget meeting into an existential one.

“Do we even keep the Lazada store? The commission went up again this month. Traffic keeps sliding. And now they’re changing the ads again.”

Her performance marketer pushed back with one number: Lazada was still her highest-ROAS marketplace. Nine times, against six on Shopee. Highest average order value too.

She is not alone. I’ve had a version of this conversation with four different brand owners since June. Everyone can see Lazada shrinking. Almost nobody has updated how they run it.

So I spent the week pulling the data — platform announcements, market trackers, and our own client accounts — to answer her properly. Here’s what I found.

Reality Bites: From 16% to 9% in One Year

Let’s start with the numbers. I’ll admit I’d been quoting the 2024 figures in meetings without realising how stale they had gone.

Across the six major SEA markets, Lazada’s share of combined platform GMV fell from 16% in 2024 to 9% in 2025 — GMV down roughly 28% year-on-year, in a market that grew +22.8% to US$157.6 billion (Source: Cube Tradewinds Q4 2025 dataset, Jun 2026; Momentum Works, Apr 2026). One caveat worth naming: trackers cut this data differently, and Momentum Works doesn’t publish Lazada’s exact share — but every series I could find points the same direction. The decline ran straight through Q4, the peak quarter, which makes it hard to blame seasonality. Shopee took 59% of the region. TikTok Shop took 31%, growing +66%.

On this dataset, TikTok Shop is now roughly three times Lazada’s size in Southeast Asia. If you’ve been suspecting Lazada slipped to third place in most markets, the numbers back you up.

The market-by-market picture is uneven, and the unevenness is the useful part:

• Singapore is Lazada’s strongest relative market — 22% GMV share (US$1.7B), still #3 behind Shopee’s 48% and TikTok Shop’s 31%, but with the highest AOV of the three platforms and real defensibility through RedMart and LazMall. Even here, GMV fell 23% year-on-year (Source: Cube, 2026).

• Malaysia — #2 by traffic, roughly 6.4M monthly web visits against Shopee’s much larger base (Source: Similarweb, Jul 2026). Still a real channel. [confirm: no reliable Lazada Malaysia GMV-share figure exists publicly — the draft deliberately uses traffic rank here]

• Thailand — #2 by traffic, but visits fell ~6% month-on-month in the latest read (Source: Similarweb, Jul 2026).

• Philippines — GMV down 34% in a market that grew: US$3.0B in 2025 against ~US$4.5B in 2024, share 28% → 16%. The only major platform that shrank while the market expanded 20%+ (Source: Cube, Apr 2026).

• Vietnam — hard to justify for most brands now. Roughly 3% share against Shopee’s 53% and TikTok Shop’s 44% (Source: Q&Me via China Digital Retail Report, 2026).

• Indonesia — a maintenance channel. Around 7% of GMV in the region’s largest market; in tracked beauty categories, Lazada’s monthly GMV fell 63% year-on-year to under 2% share (Source: Databoks/Katadata, Jun 2025; Magpie IQ, 2026).

Traffic tells the same story from a different angle. In Indonesia in May 2025, Shopee, Tokopedia, and Blibli all grew month-on-month while Lazada fell 16% — the only major marketplace losing visitors while every rival gained (Source: Semrush via Databoks, Jun 2025). App-level download and MAU data is thinner: Sensor Tower’s State of Mobile 2026 names Lazada among category leaders without breaking out numbers, and the iPrice quarterly tracker that operators used to cite went dormant years ago.

The honest summary: every independent traffic series that exists points the same direction, and none of them points up.

Then there’s what the parent company is telling you without saying it. In June 2026, Lazada cut ~5% of its workforce across all six markets — its second regional round since January 2024 (Source: Malay Mail, 24 Jun 2026). In August 2026, Alibaba folded its international commerce arm into one consolidated “Alibaba E-commerce Group” — and in the June-quarter earnings release, the segment where Lazada lives shrank 1% year-on-year while management’s commentary praised AliExpress by name and didn’t mention Lazada once (Source: Alibaba SEC 6-K, Aug 2026).

When your parent stops mentioning you to investors, you’re probably not the growth story anymore.

What Lazada Is Actually Doing About It

Lazada’s response is the part worth studying, because as far as I can tell it is not trying to win back what it lost. It is deliberately shrinking into something specific.

The strategy has a name: “Confidence Commerce.” Announced as the 2026 positioning, it is an explicit pivot away from GMV-and-subsidy competition toward brand authenticity, curation, and logistics reliability (Source: Digital in Asia, Mar 2026).

In practice, five moves define it:

1. The LazMall bet. 170,000 verified brand stores. LazMall brands grew GMV +39% during 11.11, million-dollar brands +53% since 9.9 2025, LazMall AOV +30% year-on-year, buyer penetration at 65% (Source: Lazada press releases, Sep–Dec 2025 — the platform’s own numbers, so read them as directionally true and promotionally framed). Lazada commissioned a survey of 6,000 SEA consumers that found 31% would pay a 10–30% premium for guaranteed authenticity, with beauty, electronics, and lifestyle as the top categories. That survey exists to justify the strategy — but the strategy itself is coherent: if you can’t win on price or entertainment, win on trust.

2. Following Shopee’s fee moves — off a lower base. Lazada Malaysia raised commissions to as high as 22.5% in November 2025, revised again on 21 May 2026, and on 14 August 2026 rolled out a new “all-inclusive” commission table that folds campaign vouchers, basic LazCoins, and free shipping into a single rate — the same day Shopee Malaysia updated its own table (Source: The Star, Nov 2025; Zetpy, Aug 2026). The same-day timing doesn’t look like coincidence. But note the base: Lazada’s total take remains the lowest of the three platforms in most categories. On a $100 skincare sale, a seller typically keeps more on Lazada than on Shopee or TikTok Shop (Source: Digital in Asia fee comparison, Jun 2026). We covered the repricing math two issues ago in The Fee Table Is Your New P&L.

3. Plugging into Taobao. In October 2025, Lazada achieved system-level integration with Tmall for the first time — a dedicated project team routing China’s Tmall merchants directly into SEA through Lazada storefronts (Source: EqualOcean, Oct 2025). For local sellers this cuts both ways: more cross-border price pressure in open categories, and a clearer reason for Lazada to protect LazMall as the space where authenticated local distribution still commands a premium.

4. Quick commerce in its fortress market. RedMart Now — 30–45 minute grocery delivery in Singapore, built in about 100 days, launched February 2026 (Source: Momentum Works, Feb 2026). Lazada is reinforcing the one market where it still has structural advantages rather than spending to defend the ones it’s losing.

5. A US$100 million affiliate push and an AI layer. The LazAffiliate fund (May 2025), AI Lazzie shopping assistant, virtual try-on, review summarisation — with AI-influenced orders up 36% during 9.9 2025, per Lazada’s own reporting.

Notice what’s missing from that list: any serious play for content commerce. Livestream and short-video GMV hit US$49.7 billion in 2025 — 32% of the entire SEA platform market, up from 20% a year earlier (Source: Momentum Works, Apr 2026). That is the fastest-growing chunk of the region’s ecommerce, TikTok Shop and Shopee Live own it, and Lazada has effectively conceded it.

So is the premium-brand focus working? Two things are true at once. LazMall’s own metrics are growing while the platform around it shrinks — the mall is holding, the bazaar is emptying. Whether you call that a successful repositioning or a controlled retreat depends on whether you’re Alibaba’s CFO or a seller deciding where Q4 budget goes.

Who Still Shops on Lazada

The demographic data here is thinner than any of us would like — no independent 2026 survey publishes a clean age-income cross-tab by platform. But the behavioural evidence is consistent:

• Highest average order value of the three major platforms, in every market where it’s measured (Source: Cube, 2026).

• Category skew toward considered purchases: electronics and appliances are ~25–26% of Lazada GMV in Singapore and the Philippines — categories where authenticity risk makes shoppers pay for LazMall’s guarantee.

• Lowest purchase frequency: Singapore Lazada shoppers average 6 orders/year — they come less often and spend more per visit.

• An older skew, directionally: every analyst note frames TikTok Shop’s growth as “younger demographics,” which makes Lazada’s remaining base the mirror image — urban, older, more affluent, brand-loyal. Directional, not measured. [confirm: happy to soften or cut this line if you want only measured claims]

Broadly: the deal-seekers went to TikTok Shop, the habitual browsers live on Shopee, and what’s left on Lazada is closer to a purchase channel than a discovery channel — shoppers who know what they want, care whether it’s genuine, and spend more when they arrive.

Hold that thought, because it’s the foundation of the playbook below.

What Our Own Accounts Show

We manage Lazada stores and ad accounts across pet care, personal care, electronics accessories, tyres, fashion, and grocery. Anonymised and averaged, three patterns repeat across nearly every account — and they’ll look familiar to anyone who reads their own seller centre honestly.

Pattern one: CPC is up 40–60% while the pond shrinks. For one FMCG brand’s Lazada store, average CPC went from RM0.25 in 2025 to RM0.38 in 2026 — a 52% increase — while storewide monthly GMV fell from ~RM48,000 to ~RM15,000 in the same comparable month. For a sister brand in an adjacent category, July 2026 impressions were down 89% year-on-year — from 306,000 to 34,000 — and CPC spiked to RM0.96 against a RM0.16–0.24 historical range. Sales plateaued; clicks got dearer. That’s not an optimisation problem. That’s an auction with fewer shoppers and the same number of advertisers: less inventory to bid on, so every click costs more.

Pattern two: ROAS holds — but mostly because we keep cutting spend. The same accounts still report 8–12x ROAS. It looks healthy in a monthly deck. The mechanism is less comforting: as traffic thins, we pull budget back to the small core of demand that still converts, and the ratio survives while absolute revenue shrinks quarter after quarter. In our accounts at least, reported ROAS has been the last metric to show the decline. Watch absolute ad-attributed GMV and impression volume instead — they showed us the truth about six months earlier.

Pattern three: ads have become the store. Across our Lazada accounts, ads-attributed sales now run at 60–80% of storewide GMV — for one electronics-accessories brand, ads contributed above 60% of store revenue every single month for a year. On our Shopee accounts the same brands show similar or higher ads dependence, but on a traffic base that isn’t shrinking. On Lazada, organic discovery has thinned to the point where the ad platform is the storefront. Turn off ads on Shopee and you lose growth. Turn off ads on Lazada and, for many stores, you lose the store.

One more pattern worth naming because it shaped the playbook: the stable accounts are the boring ones. The brand that held ROI between 7x and 10x for twelve straight months wasn’t doing anything clever — consistent always-on Sponsored Max, budgets matched to the demand that actually exists, no heroic scaling attempts. On 2026 Lazada, boring is a strategy.

The Ads Stack Just Changed Underneath You

Now the practical layer — because Lazada has quietly rebuilt its entire advertising product line over eighteen months, and most sellers’ campaign structures date from two product generations ago.

The timeline every Lazada advertiser needs on one page:

• April 2025 — Sponsored Store discontinued. Campaign creation blocked 4 April, all active campaigns force-stopped 19 April (Source: Lazada Sponsored Solutions, Apr 2025).

• May 2025 onward — Sponsored Max becomes the default. The GMV-Max-style automated product: Target ROAS, Max CPC, or auto bidding; search + recommendation + an exclusive “all platform traffic” placement; Store and Product variants.

• January–February 2026 — ROAS Protection. Miss 90% of your target ROAS on an eligible Sponsored Max campaign and Lazada rebates vouchers up to 80% of daily spend. Conditions apply: out of learning phase, ≥5 daily orders, target ≤1.3x the recommendation, no same-day target edits (Source: Lazada Sponsored Solutions, Jan 2026).

• March 2026 — product grouping in Sponsored Max: several SKUs in one campaign with a group-level target — Lazada’s answer to Shopee’s ad groups.

• 11 August 2026 — Cost Per Sale beta. No upfront wallet: set a commission rate (say 10%), pay only on confirmed orders, deducted from next-day store earnings, 30-day last-click attribution. Selected sellers, all six markets. It’s ads priced like affiliate — the clearest signal yet of where Lazada thinks seller trust stands.

• 20 August 2026 — Sponsored Discovery phase-out. The announcement in Joanne’s inbox: SD-Automated campaigns auto-upgrade to Sponsored Max Store, SD-Promoted Specific Products to Sponsored Max Product, migration window 19–30 September 2026, opt-out deadline 18 September (Source: Lazada Sponsored Solutions, 20 Aug 2026). Bids, budgets, product selections, and campaign status carry over automatically.

Two details in that last announcement matter more than the headline. First, Sponsored Discovery Manual — the keyword campaigns — survives. Lazada is the last of the three platforms where you can still bid on a keyword by hand; TikTok Shop stopped accepting manual bids in July 2025. Second, the direction of travel seems clear: everything automated is converging into Sponsored Max, keyword intent is being resold through the brand suite (Sponsored Brand Search added generic keywords for 9.9 2026), and the platform’s own claim — “over 80% of sellers prefer scaling with Sponsored Max” — tells you where the roadmap money went. Control is being traded for guarantees. Take the guarantees, but read the eligibility rules before you rely on them.

What we’ve learned running money through this stack — the operational fine print no announcement mentions:

• The learning phase eats short campaigns. Pause a campaign, resume it for a Mega day, and it re-enters learning right when you need stable delivery. Keep always-on campaigns running through the quiet weeks; scale budgets into peaks rather than resurrecting dead campaigns.

• Budgets can’t be reduced before 9:00 AM — a real constraint on Mega Sales Accelerator days. Either accept the morning’s high budget or pause the campaign entirely; there is no early-morning throttle.

• Sponsored Max regularly can’t hit low targets on thin traffic. We’ve set 5x targets that the system simply couldn’t achieve — not because the target was ambitious but because the demand wasn’t there. When the recommended target keeps drifting up, that’s the auction telling you the pond shrank.

• Generic keywords in manual campaigns are mostly a tax now. Across our accounts, manual generic-keyword campaigns consistently underperform Sponsored Max on both traffic and conversion. Keep manual for brand terms and a shortlist of proven category keywords; let automation fish for the rest.

• Expect a daily overspend buffer — a RM10 budget spending RM13 is normal platform behaviour, not a bug. Plan monthly, not daily.

• Sponsored Brand Search doesn’t show GMV in its dashboard — you’re flying on clicks and traffic quality. Fine for brand defence; don’t judge it like a performance product.

Lazada’s own recommended structure is worth keeping as the baseline:
70–80% of budget in CPC/performance products,
20–30% in display;
roughly 40% of the year’s spend in Mega campaigns,
35% always-on, 25% on payday and mid-month spikes (
Source: Lazada On-Platform Solutions Playbook, Oct 2025).
Adjust in 20–40% increments and wait a full week — target-ROAS systems punish fidgeting.

The Closer’s Playbook

Everything above points the same way, at least to me. Lazada in 2026 is the closer, not the opener. Demand gets created elsewhere now — TikTok, Meta, creators, and increasingly Shopee’s content surfaces. Lazada is where demand you’ve already built comes to convert: shoppers who know what they want, care whether it’s genuine, and spend more when they arrive.

Put plainly: Lazada is not a very useful platform if you’re not an established brand. It rewards four situations — a strong brand presence, a base of repeat customers, strong account-manager support inside the platform, or a brand entering a new market with latent demand waiting to be unlocked. Build the demand and create the awareness somewhere else, then make sure you’re on Lazada to capture the conversion — usually at a higher order value, sometimes with a higher-value virtual bundle doing the lifting. We covered virtual bundles back in Boosting Conversion with Virtual Bundles (Apr 2025). And if resources are tight, Lazada is probably not your first launch. It’s the second store you open once the first one has proven people want you.

One more shift that doesn’t show up in any tracker, because we only see it from inside the accounts: exposure on Lazada is increasingly negotiated, not bought. Since last year, more of the platform’s exposure decisions have been consolidating into Singapore — regional control over campaign slots and visibility mechanics that local teams used to hand out. That makes working your key account manager, incubation manager, and category manager more important than ever; they’re the route into the entry points that still move real volume — LazCoins, Surprise Box, digital vouchers, and flash deals. Those channels look old-fashioned next to a TikTok live, but on a purchase-intent platform they’re where the remaining traffic concentrates. If nobody on your team has a monthly conversation going with your category manager, you’re invisible in exactly the places Lazada still shows people things.

Here’s what all of that means for each seat at the table.

The brand owner: keep the store, cap the ambition, reset the target.

Lazada still earns its place in the mix on three conditions: it’s profitable at fully loaded take rate, your category matches its remaining strengths (beauty, electronics, home, mother-and-baby, considered FMCG), and it costs your team less than 20% of their marketplace hours. Set targets that respect the channel: flat-to-single-digit growth at protected contribution margin — not the +30% you’re chasing on TikTok Shop. And decide by market, not by logo: Malaysia and Singapore, keep and run properly; Thailand and Philippines, keep if profitable, watch quarterly; Vietnam, exit unless you have a specific reason; Indonesia, maintenance mode — presence, brand protection, zero growth spend. If you’re launching a new brand or entering a new market on tight resources, sequence it: prove demand on the discovery platforms first, open Lazada when there’s something to close. What you should NOT do: exit a profitable Lazada store to fund TikTok losses. Nine-times ROAS money is still money, and the shoppers there — remember the AOV data — are your highest-value customers on any marketplace.

The performance marketer: restructure for the September migration, then stop chasing scale.

Before September 18th, review every Sponsored Discovery campaign rather than letting the auto-upgrade decide for you — the settings carry over, but a campaign structure built for 2024 traffic shouldn’t be grandfathered into 2026. Consolidate to one Sponsored Max Store campaign as the base layer, Sponsored Max Product for heroes and launches, manual keywords for brand terms only. Set target ROAS from contribution margin, not from last year’s benchmark. Qualify for ROAS Protection and treat it as downside insurance. Test Cost Per Sale the moment you’re offered the beta — on a conversion-capture channel, paying only on confirmed orders is exactly the right risk allocation. What you should NOT do: judge the channel on ROAS alone (watch impressions and absolute ad-attributed GMV — they lead by six months), scale budget into a shrinking auction because Q4 targets demand it, or resurrect paused campaigns two days before 11.11.

The store ops team: LazMall status and the human channel matter more than anything else you control.

On a platform whose entire strategy is brand trust, marketplace-tier stores are being structurally deprioritised. Get into LazMall if your category and paperwork allow it; if you’re already in, treat its service metrics — response time, fulfilment speed, rating — as hard constraints, because the trust positioning only pays sellers who look the part. Then work the people: keep a standing monthly rhythm with your key account manager, incubation manager, or category manager, and pitch for the entry points they control — LazCoins placements, Surprise Box, digital voucher slots, flash deals, Super Brand Days, category campaigns. With exposure decisions consolidating regionally, the sellers who show up prepared with a pitch get the slots; the ones waiting for the self-serve console get what’s left. Keep PDPs current even with thin traffic — the shopper who does arrive converts at high AOV and reads the reviews. What you should NOT do: staff Lazada like it’s 2022. One capable person running store ops across Lazada MY and SG part-time is the right shape for most USD 1–5M brands — as long as that person owns the manager relationship, not just the listings.

The finance lead / planner: reprice on the new all-inclusive table and re-forecast flat.

The 14 August all-inclusive commission table changed per-SKU economics — vouchers and free shipping you used to fund separately now live inside the rate, so your old net-margin math is stale. Rebuild it per category. Budget Lazada as a profit line, not a growth line: if the P&L needs Lazada to grow 25% next year, the problem is the P&L. And model the cross-border wildcard — the Tmall pipe means open-category price competition gets worse, which is one more argument for LazMall and brand-protected assortment. The companion workbook to this issue has the math scaffold: fee comparison across the three platforms, and a closer calculator that flags when your Lazada line stops earning its keep.

That’s the whole framework — four seats, one job and one trap each.

Is Lazada Still an Effective Channel?

Yes — for a narrower job than the one you originally hired it for.

On our accounts it is usually the best-margin marketplace click in the region — the highest-AOV shoppers, generally the lowest take rates, and a parent company that has chosen curation over combat. That’s a real position, and for the right categories a bankable one. It is also a platform whose GMV halved as a share of the region in twelve months, whose organic traffic is evaporating, and whose future is being decided in Hangzhou by people who no longer mention it to investors.

Both things are true. The operators who get hurt are the ones acting on only one of them — either abandoning a profitable closing channel in a panic, or budgeting for growth that the platform itself has stopped pursuing.

This week: pull your last 12 months of Lazada data and look at three numbers — impression trend, absolute ad-attributed GMV trend, and fully loaded contribution margin per order on the new commission table. Those three numbers will tell you which seat’s playbook above you’re not running yet.

Then run it.

Related reading from the archive:

• The Fee Table Is Your New P&L — repricing a SEA catalogue when commissions move (Aug 2026)

• The Tightening Squeeze — Shopee Malaysia’s 2026 fee changes (Jan 2026)

• Thinking Full Funnel — orchestrating Shopee, TikTok Shop, Facebook & affiliates (Jan 2026)

• The Affiliate Arbitrage — 30–40% of GMV through creator networks (Jan 2026)

Affiliate marketing stopped being the free channel. Here’s what eighteen months of MCN fees, paid creators, dead outreach trackers and one very expensive December taught us about running it now.

All figures are shown in $ for regional comparability. Our own client numbers originate in local currency across Malaysia, Singapore, the Philippines and Vietnam.

Hafiz messaged a creator in June.

He runs a mid-sized consumer brand across Malaysia and Singapore — Shopee and TikTok Shop, a bit under $1M GMV a month, decent margins, a category where demos actually work. The creator was someone his team had seeded in late 2024. She’d made three videos, one of them did $11,000 in a week, and then she went quiet.

She replied in two days. Politely. She’s with an agency now. The agency handles her brand deals. If he wants her on a campaign, he should speak to them — and they run a six-month partner package, $1,200, plus 12% commission, plus a sample commitment.

Hafiz forwarded me the screenshot with one line: “Since when do I pay to talk to my own affiliate?”

Then he told me the rest. His team had a Q1 target of 300 Shopee outreaches. They did zero. On TikTok they’d targeted 260 and managed 64, all in January, then it stopped. His affiliate pool on TikTok reads 16,760 registered creators. In March, 180 of them made a sale.

He is not alone. I hear a version of this every week now, and the numbers behind it are worse than most operators realise.

Here’s the shift, in one sentence: the affiliate is no longer applying to you. You are applying to the affiliate. And a channel that three years ago cost you a commission percentage and a bit of admin now has a media budget, a vendor list, a contract with a one-year minimum, and a cost-per-video you probably can’t quote off the top of your head.

We’ve written about affiliate twice before — the 90-day build in January, and the starter playbook back in August 2025. Both still hold as sequencing. Both are now incomplete, because both assumed a market where recruiting a creator was mostly a function of effort.

That market is gone. This is the update.

The reckoning: content ate a third of the region

Start with the macro, because it explains why creators got scarce.

Content commerce — live and short-form video — hit US$49.7 billion in 2025, 32% of Southeast Asia’s total platform GMV, up from 20% in 2024 (Momentum Works, April 2026). Total platform ecommerce was US$157.6 billion. So a third of everything sold in this region now moves through content, and that share grew by 12 percentage points in twelve months.

The platforms confirm it from their own side of the ledger:

• Shopee: orders from live streaming and short-form video grew more than 50% YoY and now account for more than 25% of physical goods orders in Southeast Asia (Sea Limited Q2 2026 earnings call, 11 August 2026). That’s orders, not GMV — a more conservative measure than most people quote.

• Shopee, off-platform: affiliate orders from creators with linked Facebook accounts grew more than 85% quarter-on-quarter. Over five million creators globally had connected Facebook to the Shopee Affiliate Program as of March 2026 — roughly half of them new to the programme. Instagram went live across all eight core markets on 1 July 2026.

• TikTok Shop: SEA GMV of US$45.6 billion in 2025, roughly double the prior year (Momentum Works, February 2026). Malaysia posted over 130% YoY sales growth during Ramadan/Aidilfitri 2026 (Malay Mail, April 2026).

• Lazada joined Meta’s affiliate partnerships programme on 8 July 2026, a year after committing US$100 million to LazAffiliate with commissions up to 36% during 6.6.

• Supply side: affiliates produced over 100 million pieces of Shopee Live and Shopee Video content in 2024 — three times the prior year.

Read those together. Every major platform in this region spent 2025 and 2026 pushing affiliate off-platform onto Meta surfaces and pouring subsidy into creator supply. That is not a campaign. That’s infrastructure.

And here’s the part sellers feel: when the platforms recruit five million creators for you, the creators stop being scarce — but good creators become the scarcest thing in the market. Supply exploded at the bottom. Nothing changed at the top.

Meanwhile the mechanics quietly got more expensive on the seller’s side of the invoice. On Shopee Malaysia, the platform-funded affiliate commission is now capped at a flat ~$1 per completed order (as of 14 February 2026), and from 24 May 2026 the seller-funded portion on different-shop social media links rose from 30% to 50% of your set rate. Attribution stayed at seven days, last-click. Translation: Shopee capped what it pays and raised what you pay.

TikTok moved in the same direction from a different angle. GMV Max now optimises with affiliate costs, coupons and fees as inputs (TikTok World ‘26, May 2026) — the bidding engine has internalised your commission line.

The toll booth nobody publishes a price for

Now the part Hafiz ran into.

Agencies — MCNs, TAPs, creator agency partners, whatever your platform rep is calling them this quarter — have become the gatekeepers to the creators who actually convert. And a growing number of them charge brands to get in the door.

We work with six of them for one client. Here’s what the door costs, from our own signed terms:

That’s an 86× spread between the cheapest and most expensive route to a piece of affiliate content — $17 a video at one end, $1,500 at the other, for a deliverable most brand owners describe with the same three words: “a creator video.”

Two things about that table should bother you.

First, there is no published market price. I went looking. The trade’s own analysts concede the MCN and affiliate-network segment “has no published market size” (Digital in Asia, July 2026). The two Malaysian agencies most often named in this conversation — call them L MCN and A MCN — publish no rate card of any kind. L MCN’s own site claims 1,000+ creators and $100M+ monthly GMV as of January 2026; it does not name a price. A MCN markets 0% commission to its creators, which means it monetises somewhere else. When we asked, the answer was a campaign fee.

Second, the platforms regulate agency conduct but not agency pricing. TikTok Shop Malaysia’s MCN Agency Violation Handling rules took effect 23 February 2026 — a 120-point annual credit system with removal at zero, covering everything from impersonating the platform to bad-mouthing rival agencies. There is not one line in it about what an agency may charge a brand. Shopee went the other way and made scarcity official: its MCN programme in Malaysia is invitation-only, and accredited MCNs get “exclusive MCN-tier commission rates” their unaccredited competitors can’t access (AnyMind, March 2026).

So: an unpriced market, a conduct-regulated but not price-regulated intermediary layer, and a platform-granted rate advantage you can only reach through an agency. If you have ever wondered why the invoices started appearing, that’s the structure.

I’ll be honest about our own results, because they’re the useful part — and because the fault is shared. Across six MCNs on one brand, TikTok GMV from MCN-sourced creators ran $25,036 in January 2026 and $1,558 in July. Two of the six never activated at all; one is logged in our tracker as “MIA.” But we also signed all six to one-year minimums, pointed every one of them at the same eight SKUs at the same rate, and then did almost nothing to keep the creators they sent us engaged after month one. Some of that decay is theirs. A good part of it is ours.

The lesson isn’t “don’t use MCNs” — we still run six, and the good ones do work our own team cannot do at that speed. It’s that we bought them as one interchangeable category when they are six different specialities, signed them all to the same brief, and then treated the creators they delivered as somebody else’s job. That’s not diversification. It’s a bidding war we’re funding on both sides.

The Five Rungs: what a creator video actually costs you

Here’s the framework, and it’s deliberately not a tier chart or a 90-day plan. Both of those exist in the archive already. What’s missing from most affiliate operations I see is a cost ladder — the honest ranking of every way to get a creator to post, priced per activated video.

Five rungs. Cheapest first.

Rung 1 — Open plan (works, and quietly getting more expensive)

You set a commission, publish it, and creators self-select. Costs nothing to manage.

It works. Let’s be clear about that, because the fashionable position is to dismiss it. Across our brands the open plan still delivers a real share of affiliate GMV with no headcount attached to it, and it is the only rung that scales while you sleep. Nobody should switch it off.

What’s changed is the trend and the price. Open-plan contribution is drifting down across every account we run, and the reason isn’t platform mechanics — it’s that creators now shop the rate. The open plan used to be a doorway. It has become a price list, and you are being compared on it, item by item, against every other brand in the category. A creator with a working audience has forty offers in their inbox and a rough sense of what each category should pay. If your published rate sits below the market, you don’t get a complaint. You get silence, and a slowly falling number.

The rate itself has moved, and it moves differently by market. In Singapore, affiliate commissions now typically start at 15% and climb from there — that’s the floor of the conversation, not the ambitious ask. Malaysia, the Philippines and Indonesia still clear at 8–12% in most categories, with 15–20% needed for launches and peak windows. Same regional creator pool, tiering differently by market. If you run one commission grid across Southeast Asia, Singapore is where you are losing creators and you probably can’t see it in the blended number.

Peak season is where the gap shows up hardest. In the peak week of December 2025 for one of our brands, the open plan produced $0.00 while the targeted plan produced $80,864. Not “underperformed” — zero, for that week. In peak season every creator worth having is already committed to somebody who asked them personally and paid more to do it.

So: keep the open plan on, budget nothing to it, and start reading it as a demand signal rather than a channel. When open-plan contribution falls two months in a row, your published rate has fallen behind the market. That’s the useful information in it.

One thing to stop doing: reporting the pool count as if it were an asset. One of our TikTok pools reads 16,760 registered affiliates and delivered 180 with sales in a month — a 1.1% activation rate. The 16,760 has been quietly inflating client decks for two years.

Rung 2 — Direct outreach (cheap in cash, expensive in discipline)

DM-ing creators yourself, through Seller Centre chat or the creator marketplaces. Our own benchmark: 75 outreaches a week converts to 10–15 recruited affiliates, a 15–20% acceptance rate. That’s a good ratio, and it’s the cheapest real recruitment there is.

It also fails more often than any other rung, because it fails silently. Our outreach tracker for one fashion client shows 612 first contacts across a year, and zero second or third touches — the 14-day and 30-day follow-up columns are built into the sheet and have never once been filled. There’s no response field at all, which means we could not compute a reply rate if we wanted to. Against a Q1 target of 300 Shopee outreaches, the actual was zero.

Tools help at the margin. Xiami (虾秘) is a Chrome extension that batch-sends invitation messages and product cards to Shopee creators across MY, SG, TH, VN, PH and TW. It’s real, it works, and we run it on every client where we hold Shopee seller access and chat permissions — which is the actual constraint, not the licence fee. TikPai is a Chinese-built equivalent for TikTok, though its documented focus is TikTok Shop US rather than this region. Use both with your eyes open: they automate messaging inside a platform interface, which is exactly the behaviour policy teams tend to revisit.

Worth saying plainly — the tool is not the tactic. Seller Centre chat outreach works because somebody is doing 75 a week and following up twice. Xiami makes that person faster. It does not make them exist.

Rung 3 — Targeted plan and the top-30 (the highest-yield rung, by a distance)

Personal invitation, upgraded commission, a named account manager, first access to new SKUs. Our activation rate on this rung is above 50% — three to five times anything else.

The concentration data says this is where the money is, and it also says where the risk is. On one brand’s TikTok in a single month, the top ten affiliates produced 70.4% of all affiliate GMV, and one creator alone did 29.3%. The following month that creator did $868 — a 94% collapse. On the same brand’s Shopee, the top ten produced 18.3%. Two completely different risk profiles, run for two years on one playbook.

If you take one operational thing from this issue: your TikTok affiliate programme is a concentrated portfolio and your Shopee affiliate programme is a diversified one, and they need different management. TikTok needs succession planning for your top three creators. Shopee needs volume and reactivation.

Rung 4 — MCN and agency supply ($17–$240 a video, and specialising fast)

MCNs work. I want to be careful here, because the easy line is that agencies are a tax and you should build everything in-house. That’s wrong, and it’s expensive advice. A good MCN puts twenty relevant creators in front of you in a fortnight. Your own team, doing 75 outreaches a week at a 15–20% acceptance rate, takes two months to get to the same place. Speed is the product, and it’s a real product.

What almost nobody does properly is match the agency to the job. MCNs have specialities, exactly the way affiliates do. Some are live-first and can fill a livestream calendar; some are short-form volume machines that will produce a thousand videos and not much thought; some hold a small roster of performance creators with genuine conversion history; some are excellent in one market and useless across the border. Buying “an MCN” is like buying “an affiliate” — the category tells you nothing, the speciality tells you everything. Every disappointing MCN engagement I can point to was a speciality mismatch we didn’t diagnose until month three.

Two market shifts to plan around. Demand for MCNs is rising — more brands, more enablers, more platform programmes pointing at the same agencies. And average quality of the creators they can spare is dipping, which is the arithmetic consequence of the first thing. The better agencies know this and are changing how they work with brand owners and enabler partners: named creator lists up front instead of a black-box roster, fee-plus-performance structures, stated output floors, market-specific rosters, and co-managed communities where the agency briefs and we hold the relationship. Those are the ones worth signing. The ones still selling an unnamed pool at a flat fee are the ones whose numbers decay.

The reframe that fixed our own thinking: an MCN buys you scale, not engagement. They will get a pool built faster than you can. They will not keep it warm. Every creator they source still needs the weekly huddle, the sample turned around in 48 hours, the leaderboard, the check-in call — and none of that transfers to the vendor, no matter what the deck says. Managing your own affiliates is the operator’s job, and it becomes more important the faster an MCN scales you, not less. Scale without engagement is how you end up with 16,760 registered affiliates and 180 selling.

Guardrails we now apply:

• Match the speciality to the need, in writing. Live hours, short-form volume, or performance creators — name which one you’re buying and what “good” looks like for it.

• No one-year minimums on a first engagement. Three months with a stated output floor. Renew on evidence.

• Price the deliverable, not the relationship. Cost per activated video is the only number that compares across vendors.

• Set the entry bar before they send anyone. Ours: minimum trailing 30-day GMV in the category, a live posting history, and a content style that matches the brand. Without a written bar you get whoever has spare capacity this month.

• Keep sample approval on your side. Several agencies ask for self-approval rights. Give that up and your sample budget becomes their creator-acquisition budget.

• Segment the roster, don’t stack it. Six MCNs chasing the same few hundred creators in one market, for the same eight SKUs at the same rate, is a fee stack rather than reach. Split them by category, format or market.

• Make platform tier access explicit. If the pitch is Shopee MCN-tier commission rates, get the tier in writing. It’s the one thing you genuinely cannot buy any other way.

And treat the creators they send you as the top of your funnel, not the bottom. The ones who post twice and convert graduate into your own targeted plan on Rung 3, where activation runs above 50% and you own the relationship. The MCN found them. It shouldn’t keep renting them to you.

Rung 5 — Paid affiliates (new, and we are mid-experiment)

This is the rung that didn’t exist in our 2025 playbook, and it’s where a lot of our budget is being tested right now.

Three shapes, all live:

Paid top affiliates — $4,500 to $8,000 for two posts. These creators have KOL-level followings, but we’re not buying them as KOLs. We’re buying them as affiliates who happen to have reach — chosen off their conversion history, not their follower count, and paid a fee on top of commission. The bet is that a creator who already sells at scale is worth a guaranteed fee to lock in before a competitor makes the same offer.

Bundled affiliate packages — $8,000 to $10,000 for a cohort. Running on two brands now, a consumer goods brand and an F&B franchise. You buy a group of creators and a content volume commitment rather than individual deals. Cheaper per video, less control per video.

Livestream hours at scale — 1,000 hours of affiliate livestream across Malaysia and Vietnam, packaged by an MCN, running for a Korean beauty brand. This is the most interesting bet in the portfolio, because it tests whether affiliate live can be bought as inventory the way media is bought — a fixed number of hours, two markets, one contract.

None of these have a verdict yet, and I’d rather say that than pretend. What I will say is that the framing has changed: we’re no longer asking “how do we recruit affiliates for free.” We’re asking what an activated creator video is worth, and then buying it at the cheapest rung that delivers. That’s a media-buying question. Affiliate has become a media channel with a commission tail.

What we’ve tried on engagement, and what it actually returned

Recruitment gets the attention. Engagement is where the GMV is, and it’s the part almost nobody costs properly.

Here’s our honest ledger.

Flat tiered cash rewards. Three months, all affiliates, no competition: hit $500 in a month, get $25; $1,500, get $75; $3,000, get $150. Total cash cost across both platforms for the full quarter: $7,075. Against roughly $270,000 of Shopee affiliate GMV in the same window — 2.6%. It worked, in the narrow sense that winners grew month over month (27 → 26 → 36 on Shopee, 4 → 13 → 26 on TikTok). It also paid the same $150 to a creator who did $6,700 and one who did $19,300, and several of the top “affiliates” collecting it were cashback and ad networks, not creators at all.

Commission upsize windows. Temporarily lifting rates for 3–5 days around D-Day, Mall Day, Pay Day. This is the most reliable lever we have and the least glamorous. A 14% upsize window produced $15,018 in five days against $1,910 of commission. The dormant pool responds to price, immediately, and then goes quiet again.

Video contests. A back-to-school video contest cost $2,900. A Raya contest cost $1,100 with a target of 200+ videos. Cheap, decent output, no durable behaviour change after the prize is awarded.

Offline gatherings. A buka puasa dinner for the top 20 affiliates, roughly $1,500–2,000. A warehouse visit, $1,500. An MCN-run offline training day, $2,000. These don’t show up in any ROI column and I’d defend every dollar — the top-30 relationship is the thing competitors can’t buy, and it is built in rooms, not DMs.

Targeted campaign invitations without engagement behind them. The cautionary tale. Across sixteen TikTok targeted campaigns over three months: 148 creators accepted, 148 added the products to their showcase, and 2 posted content. Fourteen of the sixteen campaigns produced exactly $0.00. The five campaigns set at 20% commission produced $1,690; the eleven at 12% produced nothing.

Read that again. A 1.4% content fulfilment rate on creators who had already said yes. Acceptance is not activation. Adding a product to a showcase costs a creator nothing and tells you nothing.

What we haven’t properly run yet: a genuine shared cash pool, a live leaderboard with stakes, and a formal graduation path from top affiliate to retained brand ambassador on a monthly fee. That last one is the most obvious gap in our own operation — we already pay $4,000–8,000 for individual creator deals, so the retainer version is a smaller step than it looks.

The measurement problem: your affiliate ROAS is wrong

Most affiliate reporting in this market — ours included, until recently — computes one headline number:

Affiliate ROAS = Affiliate GMV ÷ Commission paid

On that basis our numbers look excellent. One brand’s full-year 2025: 16.97x on Shopee, 8.02x on TikTok. Affiliate GMV of $3.66M against $11.0M storewide — 26.1% contribution on Shopee, 53.7% on TikTok. In Q3 2026 to date those contributions are 43.8% and 67.2%.

Now look at what’s not in the denominator: samples. Video boosting. Ad spend behind affiliate content. MCN fees. Incentive payouts. Contest prizes. Agency retainers. The salary of the person doing the outreach.

In December 2025 that same brand shipped 621 samples on TikTok in one month. At even $25 landed cost, that’s $15,525 of product that appears nowhere in an ROAS of 8x. The Shopee AMS budget ladder for 2026 runs from $5,000 in April to $13,000 in December — $65,000 of ad money pushing affiliate content, sitting outside the metric entirely. The six MCN fees sit outside it. The $7,075 of incentive sits outside it.

Fix it with one line:

Fully-loaded affiliate cost = platform commission + samples at landed cost + boost/ads on affiliate content + agency and MCN fees + incentives and prizes + allocated headcount

True affiliate ROAS = Affiliate GMV ÷ fully-loaded affiliate cost

I’ve built the model as a workbook with live formulas — it’s attached to this issue. Plug in one month and you’ll get a number materially below the one in your client deck. That’s the point. December 2025 was our best affiliate month ever on reported ROAS and our worst on blended commission — 14–16%, ROAS down to 6.2 in the peak week. Peak season inverts the economics and reported ROAS hides it.

Two more metrics worth adopting, both of which we now hold ourselves to:

• Sample-to-content rate. Samples shipped ÷ content published within 14 days. Our threshold: green above 50%, red below 30%. This is the single most diagnostic number in affiliate, and almost nobody tracks it, because it requires closing the loop on a tracker that usually stops at “delivered.” Ours stopped there for an entire December.

• Activation rate by rung. Not total affiliates. Affiliates who posted, by how you acquired them. Open plan ~1%. Direct outreach 15–20% acceptance. Targeted top-30 above 50%. MCN 20–30% claimed, materially lower in our actuals.

The diagnostic: is your affiliate programme actually being run?

Seven signs. We’ve been guilty of most of them.

🔴 Your affiliate count goes up and your affiliates-with-sales doesn’t. 16,760 registered, 180 selling. If the deck leads with the pool size, the pool size is the problem.

🔴 Nobody can tell you last month’s outreach number without opening a file. If it isn’t a standing weekly number with a name against it, it’s zero. Ours was.

🔴 Your outreach tracker has follow-up columns that are empty. 612 first touches, no seconds. The cadence was designed and never run.

🔴 You can’t state your cost per activated video. Across every rung, in one currency. If you can’t, you cannot compare an $500 registration fee to an $35,000 seeding package, and you will buy the one with the better deck.

🔴 Your affiliate ROAS excludes samples and boost. Then it isn’t ROAS, it’s a commission efficiency ratio. Useful, but don’t run a budget on it.

🔴 Your top three creators are more than half your affiliate GMV and you have no succession plan. One of ours was 29.3% of a month and dropped 94% the next.

🔴 Affiliate contribution is falling while storewide GMV is rising. This is the one that should genuinely alarm you. One brand we picked up had storewide GMV up 13% YoY while affiliate GMV fell 81% — contribution from 6.3% to 1.0%, one month down 96%. Nothing broke on the platform. The programme simply stopped being run when the person running it left. Affiliate decays to near-zero within two quarters of neglect, and the storewide number hides it the whole way down.

Four or more? The channel is drifting. It won’t announce itself, because affiliate revenue decays slowly enough to look like seasonality.

How to evaluate an affiliate specialist — or the agency pitching you

Our rule of thumb after two years of hiring, firing and being the affiliate agency: ask for four numbers and one artefact. The answers tell you everything in fifteen minutes.

1. “What’s your weekly outreach number, and who owns it?” A real operator answers with a number and a name in one breath — “75 a week, Cindy owns it, we review Friday.” Anyone who answers with a strategy is not doing it. The follow-up that separates the good ones: “and what’s your reply rate?” If they don’t measure replies, they’re sending, not recruiting.

2. “What’s your sample-to-content rate?” The best diagnostic question in this entire discipline. It requires them to have closed the loop between logistics and content, which almost nobody does. Below 30% means samples are being farmed. Above 50% means the briefing and follow-up actually work. “We don’t track that” is a complete answer, and it’s a no.

3. “What does an activated creator video cost you, by channel?” Ringgit per video, across open plan, outreach, targeted, MCN and paid. A specialist who has priced their own supply chain will have a ladder. One who hasn’t will talk about relationships.

4. “Show me a month where affiliate GMV went down, and tell me why.” Everyone can present a peak. The diagnosis of a bad month is where the operating skill lives. Watch for whether they blame the platform.

And the artefact: the weekly tracker itself. Not the monthly client deck — the working file. What you want to see is a weekly row cadence, a commission column that moves (a blended actual, not the plan rate), separate open and targeted lines, samples reconciled to content, and a free-text column where somebody has written what actually happened that week. Our own master template runs 327 formulas across eleven tabs with conditional thresholds baked in — Shopee contribution green above 30% and red below 15%, TikTok green above 35%, sample content rate red below 30%. You don’t need ours. You need theirs to exist.

One thing to stop using as a screen: headcount and affiliate pool size. Both are trivially inflatable and neither predicts output.

On commercial structure, there are two honest models in this market and you should know which one you’re being offered. Retainer plus a small share of affiliate GMV — ours runs around $4,500 a month plus 2% — aligns on building the programme, and is right when the channel needs constructing. Pure commission on affiliate GMV, typically in the low-to-mid teens inclusive of the platform commission, aligns on harvest, and is right when there’s already a pool to work. Be suspicious of anyone who won’t quote either, and of any structure where the agency’s fee doesn’t move when the GMV does.

Where I think this goes

Three predictions, held loosely.

Affiliate budgets will look like media budgets by 2027. Line items for supply (paid creators, MCN packages, seeding), a commission tail, and an incentive pool priced as a percentage of GMV rather than a fixed pot. Our own modelling puts a properly funded incentive layer at around 3.9% of affiliate GMV — an order of magnitude above the 2.6% we actually spent, and I think the higher number is closer to right.

Live gets bought as inventory. The 1,000-hour package we’re testing is either an early signal or an expensive lesson, and either way somebody in this market will be selling affiliate livestream hours the way they sell impressions within eighteen months. Worth noting the split we see: on Shopee, live is 35.6% of affiliate GMV and healthy. On TikTok it’s 7.6% and falling — it hit 0.6% in Q2. The same word, two different channels.

The agency layer consolidates and starts publishing prices. Right now the absence of a public rate card is the intermediary’s main source of margin. That never survives contact with a market where buyers compare notes — and this newsletter exists partly so buyers compare notes.

The close

Affiliate marketing in Southeast Asia is no longer the cheap channel. It’s the channel with the most GMV per dollar and the least honest accounting, and those two facts are related.

You don’t fix that with a new tier structure or another 90-day plan. You fix it by pricing your own supply chain.

This week, do one thing: take last month, and calculate your cost per activated creator video across every rung you’re on. Commission, samples at landed cost, boost spend, agency fees, incentives, the fraction of a salary. Divide by the number of videos that actually went live. One number, in one currency.

Most operators I’ve done this with land somewhere between $50 and $500, and are shocked by the spread inside their own programme — usually because the rung they thought was free turned out to be the most expensive one they own.

Then ask the harder question: at that price, which rung would you buy more of?

Hafiz paid the $1,200. The creator posted twice in July and did $9,400. On a pure cost-per-video basis it’s the second-cheapest thing he bought all quarter.

He’s still annoyed about it. He’s also renewing.

Related reading from the archive:

• The Affiliate Arbitrage — the 90-day build sequence and commission-as-a-lever (Jan 2026)

• Why Affiliate Marketing is Your Secret Weapon — the starter checklist and tier structure (Aug 2025)

• Thinking Full Funnel — orchestrating Shopee, TikTok Shop, Facebook and affiliates into one engine (Jan 2026)

• The Fee Table Is Your New P&L — repricing a SEA catalogue when the take rate moves (Aug 2026)

• Build the System, Not the Headcount — TODO-confirm-url: the 12 July issue isn't in the public archive feed; this link currently points at Q4 Is Decided in July. Replace or drop.

Reply and tell me what your cost per activated video came out to. I’ll tell you where it sits against ours.

Attached: Affiliate Fully-Loaded Cost & Specialist Scorecard.xlsx — live formulas, plug in your own month.

She runs a footwear brand on Shopee Mall Malaysia — about 1,400 orders a month across women’s and men’s shoes, RM139 average order value, enrolled in the Shopee Cashback Programme (SCP) because that’s where the traffic is. Effective 14 August, her SCP commission discount drops from 5% to 2% for fashion. Her all-in rate on women’s shoes moves to 27% + RM0.54 per order + 5% Ads Growth Plan. On a campaign peak day: 29.16%, before she spends a single ringgit of her own on ads.

She opened her price list to work out the damage. The file’s last full revision was March 2024.

She is not alone. Almost every catalogue I look at was priced against 2024 fees. Since then, Shopee Malaysia has revised commission three times in 2026 alone — 1 February, 21 May, and now 14 August. Thailand moved on 4 August. Vietnam’s TikTok Shop went from 2–3% to 12.5% marketplace commission in March. The fees moved. The price lists didn’t.

Every point the platform added and you didn’t answer came straight out of your contribution margin. This issue is the method for answering it — category by category, SKU by SKU, without wrecking your conversion or your ranking.

What Just Landed

Here’s the 14 August package for Malaysia, in order of how much it will cost you:

1. SCP commission discount cut for Fashion. The 5% discount SCP sellers enjoyed on commission drops to 2% for fashion categories. For Mall fashion, that’s a +3.24 percentage point jump in total take after SST. Women’s Shoes with SCP: 23.76% → 27%. Men’s Shoes: 24.84% → 28.08%. (Source: Shopee Seller Education Hub, article 6799)

2. Campaign day fees now run 96 hours. For Double-Digit campaigns (9.9, 10.10, 11.11), the 7.5% SCP campaign fee window extends from 48 hours to 96 hours — from 12:00 AM three days before D-Day through the full 24 hours of D-Day. September’s 9.9: the higher rate runs 6 Sept 12:00 AM to 9 Sept 11:59 PM. Mid-Month Madness and Payday Sale stay at 48 hours. (Article 1096)

3. Automotive and motorcycle commission up 1.5–3 points. With-cashback rates for Automotive Mall move to 14.5–15.5%, Marketplace to 11.5–12.5%. Motorcycle Mall to 15–16%. Without-cashback rates hold.

4. Reverse shipping fee manual claims end 1 October. RIOC and ITRR claims move into the Seller Shipping Fee Saver programme — RM0.28 per order for everyone, RM0.18 for Preferred. You now pay per order, on every order, for something you used to claim per incident. (Article 20439)

5. LiveXtra folds into SCP. At no extra cost — but from 14 August it’s only available to SCP sellers. Livestream reach just became another reason you can’t leave the cashback programme.

Individually, each change reads like housekeeping. Stack them on a fashion Mall seller running campaigns, and the platform’s share of a sale has moved past 29% on peak days — before your own ad spend, before COGS, before fulfilment.

And Malaysia is not an outlier. Shopee Thailand raised commission 1–2 points on selected categories from 4 August. TikTok Shop Vietnam’s March move took marketplace commission to 12.5% and mall to 14.5%. A 5% technical support fee landed on China cross-border sellers into SG/MY/TH/VN in February. The whole region’s fee floor is rising at the same time.

If your prices are from 2024, you have absorbed every one of these increases personally.

Why the Fees Keep Climbing — and Where the Money Goes

We covered the mechanism in The Tightening Squeeze in January: Shopee raises take rates, then recycles a large share of that money into vouchers, cashback, and shipping subsidies that keep buyers on-platform. Fees fund the deals. The deals lock in the buyers. The buyers justify the fees.

The 2026 twist is that the lock-in now has a membership card. ShopeeVIP crossed one million paying subscribers in Malaysia in July — a million people paying RM6.90 a month for extra vouchers, cashback, and free shipping. Shopee says VIP members spend 1.2× more than regular users. (Source: Shopee Seller Summit, July 2026, via SAYS)

Read that from the seller’s side. The commission you pay funds the voucher pool. The voucher pool feeds a paid loyalty programme. The loyalty programme trains the region’s highest-intent buyers to check Shopee first — and to expect a deal when they get there.

You are funding the loyalty programme that keeps your customers loyal to Shopee, not to you.

That’s not a reason to rage-quit the platform — Shopee is still where Malaysian wallets are. It’s a reason to stop treating fee announcements as weather and start treating them as P&L events. Which brings us to the method.

The Repricing Ledger — Five Steps

One named framework this issue, and this is it. Work through the five steps in order. Steps 1 and 2 are arithmetic. Steps 3 and 4 are judgment. Step 5 is the discipline that stops you doing this from scratch every quarter.

Step 1 — Build the real take-rate table, per category

Not the headline commission. The real rate: commission after your actual SCP/programme status, plus the cashback service fee, plus the transaction fee, plus the per-order platform support fee, plus the Ads Growth Plan if you’re on it — all after 8% SST, which applies to Shopee’s fees and which almost everyone forgets to include.

Here’s what it looks like for a Mall fashion seller after 14 August, on a RM100 women’s shoes sale:

One nuance on the Ads Growth Plan before you file it under “fees”: the 5% is enforced, but it comes back to you as Shopee Ads credits. The cash leaves your bank on every order either way — you just get to spend it once, inside Shopee Ads. So treat it as forced marketing budget, not a pure fee. If you were going to spend 5% of GMV on Shopee Ads anyway, it costs you nothing extra. If you weren’t, the platform just made that decision for you. Either way, don’t count it twice: if it sits in this table, take it out of your marketing line in Step 2.

A third of the sale, gone before the shoe leaves the shelf. Build this table for every category you sell, in every market, in both campaign and non-campaign columns. That’s the fee table that is now, functionally, your P&L.

Step 2 — Contribution margin per SKU, not blended

Blended margin is how catalogues die quietly. A 200-SKU catalogue with a “healthy” 12% blended contribution margin usually hides 30–50 SKUs that are underwater — and campaigns push more of them under, because that’s when the take rate peaks and your discounts are deepest.

Run every SKU down the cost ladder:

1. Net selling price — after your funded vouchers and rebates (commission is charged on this number)

2. Platform take — from your Step 1 table, campaign column for campaign forecasting

3. Cost of goods — landed, not ex-factory

4. Warehousing & fulfilment — storage, pick-pack, your share of shipping subsidies, the new SSFS per-order fee from October

5. Returns — at your category’s real rate, not the rate you wish you had

6. Marketing — your ad spend beyond the Ads Growth Plan credits, allocated per order at category level (the AGP itself is already in rung 2 — don’t count it twice)

7. Affiliate costs — your affiliate commission layer, if you run one

8. Operations — your team, or your enabler’s fee if you outsource the channel

What’s left is contribution. Not “profit” — contribution. It still has to pay for your overheads and your salary.

The rule: no SKU keeps its price by default. Every SKU earns its place on the ladder or gets one of the three moves below.

Step 3 — Pick the right move: absolute, bundle, or mix-shift

Repricing is not “raise everything 5%.” That’s how you torch conversion. There are three moves, and each SKU gets the one that fits:

Move 1 — Absolute. Raise the list price. Right for SKUs where you have pricing power: hero products with review depth, low direct comparability, strong repeat rates. Wrong for knife-fight SKUs where you’re one of forty identical listings — there, an absolute increase just hands your ranking to the seller who didn’t move. Check what the top five competing listings did after 14 August before you decide you can’t move; if the whole category absorbed the fee, prices drift up together.

Move 2 — Bundle. The price tag looks the same to the buyer, but the economics underneath improve. Bundles raise AOV, which dilutes the per-order fixed costs (RM0.54 support fee, SSFS, pick-pack) and lets you bury the increase in perceived value. A RM139 pair of shoes becomes a RM159 shoes-plus-care-kit bundle with RM11 better contribution than the pair alone. We covered the mechanics in Boosting Conversion with Virtual Bundles — they convert better, not worse, when the anchor item is strong.

Move 3 — Mix-shift. Change what you sell, not what it costs. Push ad spend and campaign slots toward your high-contribution SKUs. Pull the underwater SKUs out of campaigns entirely — they can sit at full price for organic traffic, but they don’t get boosted, don’t get vouchers, don’t get flash slots. The bottom of the ladder gets retired or reordered in smaller quantities. Your average take rate is fixed by Shopee; your average margin is set by your mix. The mix is yours.

And one lever that costs nothing: stop double-discounting. Part of what those rising fees fund is platform cashback that now runs deep in BAU and deeper on peak days — sometimes bordering 22% off with a minimum spend, funded by Shopee, not you. If the platform is already giving your buyer 20%+ off, your stacked 15% seller voucher is margin given away twice. Trim your own vouchers and rebates and let the platform’s money do the discounting. One guardrail: keep at least a 5% discount on your campaign vouchers — that’s the threshold to qualify for campaign vouchers and keep your spot on the campaign page.

Step 4 — Test without wrecking your ranking

Shopee’s ranking rewards conversion. Price changes move conversion. So a clumsy repricing round can cost you search positions that took two years to earn. Five rules:

• Move in steps, not leaps. Cap any single change at ~10% of list price. Two 6% moves six weeks apart beat one 12% move.

• Test with vouchers before you touch list price. Shrinking a seller voucher from 15% to 10% raises your effective price with zero listing edit and zero ranking event. If conversion holds for two weeks, make it structural.

• Never reprice a hero SKU inside a campaign window. With Double-Digit fee windows now 96 hours, your September calendar has fewer safe days than you think — map them first.

• Watch a 14-day window per change: conversion rate, search position on your top keywords, and add-to-cart. One category at a time, so you can read the signal.

• Tell the platform before you move. Price changes — up or down — can trip Shopee’s price-violation checks and get listings flagged. Brief your account manager (or the platform through Seller Centre if you don’t have one), get the go-ahead, and schedule the changes before or after peak periods — never inside one.

Step 5 — The cadence that stops this recurring

The reason most catalogues are priced against 2024 fees isn’t laziness. It’s that repricing was treated as a project instead of a process. Projects end.

The process: every platform fee announcement triggers a 7-day SOP. Update the take-rate table. Re-run the SKU ladder. Flag every SKU whose contribution fell below your floor. Decide the move — absolute, bundle, mix-shift, or absorb (absorbing is a legitimate decision when it’s made, not when it happens by default). Then a full catalogue review every quarter, timed ahead of the mega-campaign calendar — July before 9.9–11.11, January before Raya.

At current announcement frequency — three revisions in six months — this SOP runs more often than your quarterly review. That’s the point. The sellers being priced out of Shopee right now are not the ones with the worst products. They’re the ones who found out about their own cost structure last.

The Pricing Discipline Behind a 229% Year

A Malaysian footwear brand we operate grew 229% year-on-year on marketplaces. The growth story gets told as a marketing story — campaigns, livestreams, affiliates. What rarely gets mentioned is that none of it would have survived contact with 2026’s fee tables without the pricing discipline underneath:

• The take-rate table was rebuilt within days of every fee announcement — the team never ran a campaign on stale fee math.

• Every SKU had a known contribution figure, updated quarterly. Campaign inclusion was decided from that ladder, not from gut feel about what “usually sells.”

• Underwater SKUs weren’t discounted deeper to “move stock” — they were bundled with high-margin items or pulled from paid traffic entirely.

• Hero SKUs took absolute price increases in two steps across the year. Conversion held, because review depth and content had earned the headroom.

Growth covered the fee increases; discipline made sure the growth was worth having. 229% of negative contribution is just a faster way to run out of cash — we wrote about that trap in The Sales Growth Formula, and about what rising acquisition costs do to undisciplined catalogues in The CAC Crisis.

Download the repricing sheet template here.

The Monday Morning Version

You don’t need a war room. You need the ledger:

1. Pull the current fee tables for your categories — Malaysia’s new rates are live on the Seller Education Hub. Build the take-rate table with SST in it.

2. Run your top 50 SKUs down the cost ladder. Find the ones underwater at campaign-day rates.

3. Assign each one a move: absolute, bundle, mix-shift — or a deliberate absorb. Check what platform cashback is already giving your buyers before you fund another ringgit of your own vouchers (keep the 5% campaign-voucher minimum).

4. Brief your account manager on the changes and time them outside peak windows.

5. Put the 7-day fee-response SOP in writing and give it an owner.

The platforms will keep raising fees — the flywheel demands it, and a million VIP subscribers are proof it’s working. You can’t control the fee table.

Price from the fee table you have, not the one you remember.

Related reading from the archive:

• The Tightening Squeeze — Shopee Malaysia’s 2026 fee changes and the Lock-In Flywheel (Jan 2026)

• The Great Platform Fee Squeeze — the 2025 fee wave and what it did to margins (Jun 2025)

• The Sales Growth Formula — the P&L math that should run your growth targets (Sep 2025)

• The CAC Crisis — when acquisition costs eat the catalogue (Jun 2025)

• Boosting Conversion with Virtual Bundles — the bundle mechanics behind Move 2 (Apr 2025)

A founder I’ll call Amira sent me a screenshot last month at 11:40 on a Tuesday night.

It was the Shopee Indonesia eyeliner category ranking for Q2 2025. She runs a Malaysian colour cosmetics brand — six years old, halal-certified, decent numbers, RM40 average order value, the kind of business that should be winning right now. She wasn’t looking at Indonesia to enter it. She was looking at it because her Malaysian numbers had started doing something she couldn’t explain, and someone told her to look at Jakarta first.

Top of the list: Pinkflash, at 17.71% of category sales value. Maybelline second at 7.64%. Wardah — Wardah, the biggest local beauty brand in the largest Muslim-majority market on earth — third at 4.18%.

Her message was two lines:

“Pinkflash is 2.3x Maybelline. And I have never heard of them. Who are they?”

Fair question. Here’s the harder one, and it’s the reason I skipped last Sunday’s issue rather than publish something half-baked: the four Chinese-owned brands in that top ten — Pinkflash, O.TWO.O, Judydoll and Xi Xiu — total 26.71% of the category. The four Indonesian-owned brands — Wardah, Somethinc, Make Over and Muaq — total 11.89%. And when you go and find out who actually owns the Chinese ones, seven brands collapse into four companies.

(The other two in that top ten: Maybelline is L’Oréal, Sister Ann is Korean. Muaq, for the record, is genuinely local — PT Muaq Berkah Sejahtera, founded in Jakarta by makeup artist Nur Fitriana Qomarry.)

I spent the past ten days pulling trademark registries, Chinese trade press, BPOM enforcement notices, Malaysian NPRA cancellation lists and every marketplace crawl I could get my hands on. What I found is not the story the digests are telling. The growth is real and the numbers are worse than most operators think.

But the model has a shelf life, and it has already started to expire — in public, in the data, in the last six months.

This one’s long. It’s the closest thing to investigative reporting I’ve done in this newsletter, because I don’t think the operator version of this story has been written yet. Everything here is sourced, and where I couldn’t verify something I say so.

Part 1 — The scoreboard is lying to you

Start with the number everyone quotes.

Euromonitor’s 2025 analysis found that in Indonesia’s mass colour cosmetics category, seven leading brands from China went from 2% market share in 2019 to over 15% in 2024. Across Southeast Asian mass skincare, Chinese brands compounded at a 115% CAGR from 2019 to 2024. Colour cosmetics grew at 70% CAGR over the same window.

The seven brands, named: Focallure, Pinkflash, Barenbliss, O.TWO.O, Y.O.U Beauty, Skintific, Dazzle Me.

Now here’s what nobody puts in the deck.

Y.O.U Beauty, Barenbliss and Dazzle Me are all owned by the same company. Hebe Beauty — founded 2018 by a team that came out of the Chinese smartphone brand OPPO — runs Y.O.U, Glamfix, barenbliss, Dazzle Me and lavojoy. That’s Hebe’s own brand roster, from Hebe’s own press release. Its Indonesian operating entity is PT Hebe Beauty Style in North Jakarta. It has raised roughly US$70M, including a US$40M Series C led by Hidden Hills Capital with Gaorong Capital and ATM Capital — both China-based — participating.

Focallure and Pinkflash are also the same company. Both sit under Guangzhou Jizhi Trading Co., Ltd. (广州即至商贸有限公司), which also owns STAGENIUS, and which runs a separate China-facing sister brand called 12Gardens built on “traditional Chinese cultural concepts.”

O.TWO.O is the fourth. Its US trademark — registration 5190742, filed 21 September 2016 — is held by Guangzhou Zoe Biotechnology Co., Ltd. Which entity controls the brand today isn’t established in any public record I could find, but the origin is on the registry.

So: seven brands, four corporate actors. One group with US$70M behind it accounts for three of the seven, and a second group accounts for two.

That would just be trivia, except for what each of those brands says about where it’s from.

Y.O.U Beauty is filed by Indonesian media as a brand kosmetik lokal — a local Indonesian brand. BeautyMatter’s own headline on its Series C called it an “Indonesian Beauty Brand.”

Barenbliss launched in Indonesia as Korean. Not implied — stated. ANTARA, Indonesia’s state news agency, headlined its August 2021 launch: “Brand kosmetik asal Korea Selatan barenbliss resmi hadir di Indonesia” — a cosmetics brand from South Korea officially arrives in Indonesia. The brand’s own copy describes it as a “rising clean beauty brand from Seoul.” Hebe’s internal description of it is blunter: “barenbliss focuses on the k-beauty.”

Dazzle Me‘s September 2022 launch release describes its parent as “Hebe Beauty Cosmetics Inc., an Indonesian beauty group.”

And it still works. In July 2026 — last month — the market-intelligence firm Magpie IQ published its Indonesian ecommerce skincare rankings and wrote that six of the top ten brands are “homegrown.” Its list of homegrown brands includes Skintific, which it classifies as “Local-led.” Skintific’s trademark is registered to Guangzhou Fimedia Network Technology across the US, EU and Vietnam registries. Strip Skintific and Glad2Glow — also Chinese-origin, imported into Indonesia by PT Suntone Wisdom Indonesia — out of that top ten and only four of the top ten are genuinely Indonesian-owned, with Chinese capital holding the #1 and #3 slots.

This is the part I want operators to sit with. The scoreboard itself has been captured. When the analysts, the data vendors and the state news wire all file a Chinese-owned brand as local or Korean, “defend against C-beauty” isn’t even a strategy anyone can brief, because nobody can see the outline of the thing they’re defending against.

Allie Rooke, who runs Clean Beauty Asia, put the structural version of this to BeautyMatter better than I can:

“C-beauty isn’t dominating Southeast Asia as ‘Chinese brands,’ but it is shaping the market structurally [in terms of] product, supply chain, marketing channels, and marketplaces. The brands that win look local, with China capability behind the scenes.“

Her shorter version: China is winning as the backend.

Part 2 — Three timelines, and what each one tells you

The growth-rate headline hides the sequencing, and the sequencing is the transferable part. Let me take the three brands in Amira’s screenshot one at a time.

Focallure: the factory came first

Focallure is the oldest and the most honest about its origins. It’s also the one whose founding date nobody can agree on, and the disagreement is instructive rather than sloppy.

The company’s own line, repeated verbatim across Chinese trade press, is “FOCALLURE菲鹿儿创办于2016年” — founded in 2016. Its US trademark filing swears first use in commerce on 2 January 2015. The legal entity that eventually took its venture round, Guangzhou Feiluer Trading Co., Ltd., wasn’t incorporated until 26 October 2017.

Three dates, three different things: the brand, the mark, the funded company. Founder and legal representative is Fang Xing, holding 30.61% of the funded entity.

The money arrived late and only once. In September 2021, Focallure announced a Series A in the tens of millions of USD, led by 5Y Capital and XVC. The stated use of proceeds, in order: supply-chain platform and own-factory construction; middle- and front-office talent; product R&D; brand building and marketing.

Read that ordering again. A beauty brand raised its first institutional round and spent the first line item on factories.

As at 27 July 2026, 36Kr still lists the company at Series A. No publicly recorded round in nearly five years. The self-declared “next L’Oréal in emerging markets” — that’s the company’s own boilerplate, 目标是打造新兴市场”欧莱雅” — has not needed to raise again.

What it built instead: six factories and an R&D centre in Guangzhou, running an average product life cycle of 3–6 months. The reason it went to Southeast Asia first and China second is almost comically operational:

“By catering to regional markets’ specific needs, such as sweat-proof products for Southeast Asian consumers, Focallure was able to grab a position in emerging markets early on and expand back to its home base, China, later.”

And the catalogue strategy: Euromonitor counts 250+ SKUs in each Focallure Shopee store, in each country, at USD 2–5 — roughly RM8.20–RM20.50. Against Indonesian and Thai Gen Z who spend USD 2–6 per beauty item, with 68% of Indonesian Gen Z spending under USD 6 per eye-makeup item.

That isn’t brand-building. It’s category capture. Flood every price point and every shade in every market simultaneously, and let the search algorithm do the merchandising.

Pinkflash: the price floor, and the bill that came due

Pinkflash launched in September 2020 as the cheaper sibling. Core SKUs on Shopee Indonesia run Rp 9,000–20,000 — call it USD 0.50–1.11, or roughly RM2.05–RM4.55. Three-shade eyeshadow palettes at Rp 30,000–50,000, about RM6.80–RM11.40.

There is no defensible answer to that price if you’re a local brand paying local formulation, local fill, local marketing and local margin. There just isn’t. Which is why Pinkflash sits at 17.71% of Shopee Indonesia’s eyeliner category — 4.2x Wardah.

But this is where the investigation turned, and where I’d push back hard on anyone selling this model as something to admire.

In late November 2024 — covered nationally between 30 November and 3 December — BPOM cancelled the distribution licences for three Pinkflash products as part of a sweep of 55 cosmetics sampled between November 2023 and October 2024. The Pro Touch Eyeshadow Palette PF-E15 #02 contained Red K3 and Red K10. The L01 Lasting Matte Lipcream R04 contained Red K3. The Multi Face Pallet PF-M02 #01 contained Acid Orange 7. Indonesian coverage described these dyes as karsinogenik — carcinogenic — and capable of impairing liver function. All three were registered to PT FCL International Indonesia, manufactured by Guangzhou Duoduo Cosmetics Co. Ltd.

Pinkflash said the contract manufacturer had substituted raw materials without its knowledge or consent, that it had withdrawn and destroyed the products, and that it had terminated the factory contract.

Then it happened again. On 3–6 November 2025, BPOM revoked licences for two more Pinkflash eyeshadows — PF-E23 BR04 for Red K10, PF-E23 BR02 for Acid Orange — out of a batch of 23 dangerous cosmetics drawn from July–September 2025 surveillance. Pinkflash halted distribution online and offline and offered consumers up to double their purchase price as compensation.

Five SKUs pulled in eleven months, across two enforcement cycles, over dyes BPOM prohibits.

That’s the cost structure showing through the paint. A 3-to-6-month product cycle at USD 1 a unit across hundreds of SKUs and multiple contract factories is not a system with room for incoming QC on every dye lot. The speed that wins the category is the same speed that produces this.

Skintific: the one that built a nationality

Skintific is the most sophisticated of the three and the one worth studying hardest.

Its parent, Guangzhou Fimedia (Feimei) Network Technology, was incorporated in 2019. Its founders have never been publicly disclosed — Chinese trade press states this plainly; the team is described only as “tech types from Alibaba and Huawei.” Its Indonesian BPOM licence holder is reported as PT May Sun Yvan, with an office in Pluit, North Jakarta.

It entered Indonesia in August 2021 — not China. It went live on TikTok Shop in October 2021. By full-year 2022 it was TikTok Shop’s Global Sales Champion, plus sales champion in Indonesia and Malaysia, and it swept the Female Daily, Sociolla, Beautyhaul and TikTok Live awards. Cumulative Southeast Asian sales passed RMB 800M — roughly US$110–115M at then-prevailing rates — inside two years of launch.

And the origin story it told while doing it: “Formulated in Canada.”

That claim appeared in the brand’s own LinkedIn copy in October 2021. By 31 July 2023 it was being reproduced verbatim, as fact, by BERNAMA — the Malaysian national news agency:

“Formulated in Canada, their formulas are based on their unique Trilogy Triangle Effect (TTE) Technology which means precise active ingredients that quickly diminish skin problems and effectively protect and improve skin barrier and are particularly suited for Southeast Asian skin tones.”

A Nikkei BizRuptors case study by Universitas Indonesia’s Dr Gita Gayatri catalogued the brand this way: Country of Origin Image — “Canada, Norway.” Actual Country of Origin — “Unclear, Speculated from China.” Year entered SEA — 2022. Note: “Actual founding year unclear.”

That 2022 entry date is one of the disagreements I promised to flag rather than paper over. Skintific’s Indonesian launch and its TikTok Shop activity both date to 2021, so 2021 is what I’ve used.

I could not find a Skintific corporate registration in Canada. I could not find a patent number for TTE. What I could find, in Chinese trade press describing the strategy from the inside, translated:

“It cleverly downplayed its Chinese background; marketing content was entirely planned by an Indonesian local team, and the official website and slogans all used Indonesian… making users mistakenly believe it was an Indonesian local brand.”

The SCMP’s Aisyah Llewellyn reported the consumer consequence in May 2024 — and this is the sentence Malaysian and Indonesian operators should tape to the wall:

“While other Western and Korean skincare products have also flooded the Malaysian and Indonesian markets in recent years, these brands do not obscure their ‘foreign’ origins, meaning consumers are able to make a more informed decision.“

Her piece follows an Indonesian shopper, Farhana, who abandoned a Skintific purchase after discovering the goods were made in China, because she then doubted whether they were halal certified.

That’s the whole mechanism in one anecdote. In this region, origin erasure routes around halal scrutiny and China-safety anxiety at the same time — two objections a local brand has to answer and an erased brand never gets asked.

One thing I want to be precise about, because it matters: no regulator anywhere has ruled on those origin claims. Not BPOM, not KPPU, not KKM, not JAKIM, not Canada’s Competition Bureau. The pushback has been journalistic, academic and consumer-driven. Don’t let anyone tell you Skintific was found to have done something unlawful. What’s documented is a gap between what the marketing implied and what the registries show.

Part 3 — The channel was the moat, and the moat was pre-built

Here’s the part most analyses get backwards. They treat C-beauty’s rise as a pricing story. It’s a channel story, and the channel was already familiar before anyone landed.

K-beauty entered Southeast Asia through the retailer. COSRX launched exclusively via Watsons across Singapore, Malaysia, Thailand and the Philippines in 2017, and Watsons reports a CAGR above 80% off that partnership. Laneige was founded in 1994 and made its first international move — Hong Kong department stores — in 2002. Eight years of domestic incubation before export.

That route requires a category buyer to say yes: GTINs, shelf-ready packaging, safety assessments, pharmacist training assets, a listing committee, a planogram slot, trade terms. Watsons and Guardian don’t publish their listing lead times or margin requirements and I won’t invent numbers for them — but every operator reading this knows the answer is measured in quarters, and knows who holds the leverage.

C-beauty entered around the retailer. Shopee, Lazada and TikTok Shop have operating logic close enough to Taobao, Tmall and Douyin that a Guangzhou team could run them on muscle memory. No buyer. No planogram. Content, inventory, logistics — all three already optimised at home.

And the execution volume is where the gap gets embarrassing. From a peer-reviewed Universitas Indonesia case study on Skintific’s Indonesian TikTok operation:

• IDR 100 billion+ per month in TikTok Shop sales between January and June 2023, rising to IDR 130 billion during Ramadan and Eid.

• 72.28% of May 2023 revenue and 76.51% of June 2023 revenue came from affiliates and live-streamers — not the brand’s own accounts. Of what did come through owned accounts, over 90% came from live streaming.

• Individual live-streamers earning up to IDR 5 billion a month, with more than five affiliates earning billions of rupiah monthly.

• Seven-plus official TikTok accounts in Indonesia alone, a “Skintific Barrier Community” WhatsApp group for affiliates, and monthly top-five revenue rewards.

Chinese trade press adds the regional picture: no fewer than 30 TikTok accounts across Indonesia, Thailand and the Philippines, with the Indonesian account running 24 livestreams in 28 days at an average of IDR 1.169 billion per session.

Meanwhile Skintific had set up local warehousing and logistics in Southeast Asia from the outset and established subsidiaries before launching, rather than exporting into the region. That is a local operator with a Guangzhou spine, not a cross-border seller.

Meanwhile Skintific had set up local warehousing and logistics in Southeast Asia from the outset and established subsidiaries before launching, rather than exporting into the region. That is a local operator with a Guangzhou spine, not a cross-border seller.

If you want the one-line version to take to your board: they didn’t outspend the incumbents. They out-volumed them, on a channel the incumbents were still treating as a secondary sales outlet.

Part 4 — The speed gap is structural, and the “tropical skin” defence is gone

Focallure’s C2M loop is the bit local brands should be most worried about, because it can’t be answered with budget.

Six factories, one R&D centre, an average product lifecycle of 3–6 months, and — in the company’s own words — a big-data system built from day one that connects user-side data to platform-side data and drives both product development and channel analysis.

Put a number on the other side. A local brand in Indonesia notifies a cosmetic with BPOM and waits 14 working days for verification, at Rp 500,000 per item if ASEAN-manufactured or Rp 1,500,000 if not, valid three years. In Malaysia, NPRA notification is RM50 per product and per variant, with the Notification Note generating from the Quest system essentially on payment confirmation. Then add the formulation cycle with a third-party lab, the fill run, the tooling and the retail listing window.

So regulation is not what’s slowing local brands down. Malaysia’s cosmetic notification is a RM50 self-declaration with a near-instant note — a turnstile, not a moat. The gap is the product loop: weeks versus quarters, decisions from live sales data versus decisions from a planning cycle.

And then there’s the uncomfortable one. The standard local-brand answer to all of this is we understand tropical skin and they don’t. Mintel’s senior beauty analyst in Shanghai, Laurie Du, told BeautyMatter:

“Southern China shares similarities with Southeast Asia in terms of hot and humid weather conditions, and Chinese R&D has developed expertise in lightweight, noncomedogenic textures that help manage sebum without leaving the skin feeling stripped. This aligns closely with the needs of consumers in markets such as Indonesia, Thailand, and Vietnam.”

Guangdong is hot and humid too. That defence is thinner than most of us have been pretending.

Part 5 — October 2023: the moment it should have broken

Every good investigation has a point where the thing nearly dies. How C-beauty survived its own tells you more about the model than the growth curve does.

On 4 October 2023, the Indonesian government shut down TikTok Shop. Trade Ministry Regulation 31/2023 banned transactions on social media platforms and required legally separate ecommerce entities, on the stated grounds of protecting MSMEs from cheap imported goods sold direct to consumers.

Think about what that meant for a brand generating IDR 100–130 billion a month on that one channel, with three-quarters of its revenue coming through affiliates and livestreamers on it. It should have been fatal. It wasn’t.

By Q1 2024 — the very next quarter — Skintific was the #1 facial-care brand on Shopee Indonesia, at IDR 64 billion through its official store plus IDR 7 billion non-official, and #1 in beauty bundles at over IDR 70 billion — Compas reports the bundle ranking separately, so read the two as overlapping views of the same quarter rather than additive. Meanwhile TikTok solved the regulatory problem by buying 75.01% of Tokopedia from GoTo in a deal announced 11 December 2023 and completed 31 January 2024, folding TikTok Shop into a licensed local entity.

Then the ratchet kept turning: by Q1 2025 Skintific was the #1 beauty and skincare brand on Shopee Indonesia overall at 4.10% share, ahead of Wardah at 2.97% and Glad2Glow at 2.51%, drawn from Compas’s analysis of 12,900 brands. By Q1 2026 it held #1 in face masks at 11.73% (up from fifth) and #1 in facial serum at 6.4%.

The lesson for anyone building a channel strategy: the brands that survived the shutdown were the ones whose creator relationships and content operation belonged to them rather than to the platform. Skintific’s affiliate roster, WhatsApp communities and agency throughput were portable. It moved them to Shopee Live and Tokopedia and kept selling.

Regulation has kept tightening since, mostly in ways that raise the floor rather than close the door. From June 2025, BPOM added an office-premises audit requirement for cosmetic importers on top of the existing warehouse audit. And the big one is coming: from 17 October 2026, halal certification becomes mandatory for all cosmetics circulating in Indonesia under Government Regulation 42/2024, with BPJPH’s head saying there will be no further extension.

Which brings me to the finding that should worry local operators most.

Hebe Beauty saw the halal mandate coming in 2022 and formed a joint venture with 3INS, a Guangdong OEM/ODM, to build a manufacturing plant in Indonesia — explicitly to secure halal certification and meet BPOM standards. Hebe had already moved its product team to Indonesia in 2019. It won Guardian Indonesia’s Best New Brand award in 2019, within roughly twelve months of Y.O.U’s launch.

The regulatory event that local brands have been quietly counting on to cull the invaders will mostly cull the long tail of small Chinese importers. The best-capitalised attackers are already compliant.

Part 6 — And then they bought the shelf anyway

This is the twist I didn’t expect when I started, and it reframes everything.

The channel-native brands are going offline.

• 31 July 2023: Skintific launches offline in Malaysia as a Guardian Malaysia exclusive, three 5X Ceramide ranges, priced RM11–RM69.

• 11 October 2024: Judydoll — described in the launch coverage as “one of China’s leading beauty brands” — launches exclusively at Watsons Malaysia across 130 stores.

• January 2025: Skintific becomes an exclusive Watsons partner in the Philippines. By end-July 2025 it is in almost 700 Watsons stores nationwide, with a pop-up at SM North EDSA.

• Hebe Beauty is now in over 40,000 physical retail doors across Indonesia, Malaysia, Thailand, the Philippines and Vietnam. And note this: as of H1 2022, online was only 25% of Hebe’s total sales.

The archetypal “C-beauty is a TikTok phenomenon” assumption is wrong for the group that owns three of Euromonitor’s seven brands. Hebe is an offline distribution business with a Chinese supply chain and a local face.

Elsie Zhang of DLG China, speaking to BeautyMatter, explains why the shelf matters even after you’ve won the marketplace:

“Establishing a physical presence signals brand equity beyond the price-sensitive realms of e-commerce and multibrand retailers like Watsons.”

And on where the fight is heading:

“While the core competitiveness of C-beauty remains — driven by price arbitrage, high SKU variety, and transferable digital tactics — the newness is beginning to fade. In 2026, winners will be defined by their ability to move the battle from ‘product’ to ‘brand,’ following the successful trajectory of Chinese giants in other categories like Pop Mart, Vivo, and BYD.”

So the full sequence is: enter around the retailer, win the marketplace, use the marketplace proof to negotiate the shelf, then convert cheap volume into brand equity. K-beauty rented the incumbents’ distribution. C-beauty built a parallel one, then came for the shelf from a position of strength.

If your main asset is “we’re already in Watsons and Guardian,” that moat is being bought out from under you — the first breach was three years ago, and on DLG’s read the product-to-brand shift is what decides 2026.

Part 7 — The five signals the playbook is decaying

Now the part that isn’t in anyone’s doom deck, and the reason I think this is a strategy issue and not an obituary.

The model is showing wear, and it’s showing in transaction data, not opinion.

1. Y.O.U — the original poster child — is softening. Magpie IQ’s April–June 2026 SKU-level tracking of Indonesian ecommerce skincare lists Y.O.U among “legacy value brands” losing share, alongside Pond’s. The brand that pioneered the Indonesian-face/Chinese-backend model is now the one being displaced.

2. Wardah is recovering. In the same dataset, Glad2Glow (the leader) is cooling and Wardah has closed to within a point of the category lead. The incumbent everyone wrote off is climbing.

3. Derma and skin-barrier brands are taking the premium ground. Skin1004 sits at #4 in Indonesian ecommerce skincare, with Skin1004, COSRX, D’Alba and Cetaphil all gaining share — a clear premiumisation and skin-barrier signal. Euromonitor forecast in July 2025 that Chinese brands would cause “the strongest disruption” to local and global players while leaving the premium segment largely untouched. That forecast is holding, and the premium segment is where the growth went.

4. The channel that carried C-beauty is losing share in the category it carried. Magpie IQ, 3 July 2026, on Indonesian beauty and personal care between January and May 2026: Shopee grew from US$109.5M to US$124.9M per month, up 14% and up every single month. TikTok Shop fell from US$86.4M to US$78.0M, down 10%, with the standalone app channel down 27%. The gap doubled from about US$23M to about US$47M in five months. Their hypothesis — flagged as unproven — is that TikTok Shop’s Indonesian headcount consolidated from roughly 5,000 post-merger to a reported 2,500, and beauty live-selling is operationally intensive to support.

Their recommendation is the single most actionable line I read in ten days of research: “Keep your live-selling and creator engine close to home… the beauty brands most insulated are the ones that own their creator relationships and live-selling operation rather than leaning on platform-run enablement.”

5. The quality bill is arriving. Five Pinkflash SKUs pulled by BPOM in eleven months over dyes BPOM prohibits and Indonesian coverage described as carcinogenic, the company’s account that a contract factory substituted raw materials, and an up-to-double refund offer. Run a three-month product cycle at USD 1 across hundreds of SKUs and multiple contract factories long enough, and this becomes predictable. Indonesia’s October 2026 halal mandate lands on exactly that long tail.

None of this means the attack is over. Indonesian ecommerce skincare is now a ~US$77M-per-month market — about RM315M a month — growing 7% year-on-year with roughly 8,900 active brands and no single brand above 7% share — the top five hold 22.1%, the top ten 32.6%. Nobody dominates that market. Distribution and content velocity decide it, permanently.

But it does mean the answer to “how do we compete with C-beauty” is no longer “you can’t.”

Part 8 — The Malaysian picture, and the number that doesn’t exist

I have to be straight with you here, because it’s the biggest gap in this piece.

There is no published figure for C-beauty’s market share in Malaysia. Not for Shopee Malaysia, not TikTok Shop Malaysia, not Malaysian beauty retail, any year. The 2%-to-15% number is Indonesia only. Anyone quoting you a Malaysian C-beauty share is citing something they haven’t shown you, or making it up — and most of the figures circulating, including a widely repeated claim that the market is worth USD 52 billion, are content-farm nonsense off by more than an order of magnitude. The credible market size is about USD 3.7 billion, roughly RM15 billion, projected for 2026, growing 5.6% a year, with online at 27% of revenue.

What we can establish about Malaysia:

• Shopee is used by 71% of Malaysian online shoppers, down from 74% a year earlier; TikTok Shop is at 62% and rising. Beauty and cosmetics purchase incidence jumped to 45% from 29%.

• Beauty and personal care is TikTok Shop’s largest category by GMV across the region, and Skintific and Glad2Glow are among the top-grossing beauty shops on TikTok Shop Malaysia — the same two Chinese-owned brands leading Indonesia.

• C-beauty is already on Malaysian shelves: Skintific at both Watsons and Guardian, Focallure at Guardian, Barenbliss and Colorkey at Watsons, Judydoll across 130 Watsons stores.

• Barenbliss’s Watsons Malaysia brand page, as at 27 July 2026, describes it as “THE NEW TREND KOREAN CLEAN BEAUTY / Initiated by Korean and Chinese Makeup Artist.”

And one finding I did not expect. I read Malaysia’s published NPRA cancelled-cosmetic-notification list for 2024, 2025 and 2026 as at 25 July 2026, and searched it for every Chinese-origin beauty brand I could name — Focallure, Pinkflash, Skintific, O.TWO.O, Barenbliss, Y.O.U, Dazzle Me, Colorkey, Judydoll — plus Guangzhou, Shenzhen, Shanghai and China as manufacturer terms.

Zero matches.

Three cancellations in 2026 to date, fourteen in 2025, twelve in 2024 — and they are almost entirely small local whitening and night creams pulled for mercury, hydroquinone, tretinoin and betamethasone — names on NPRA’s own published cancellation list like Saqa, JLFARA, Molly Care, Seri Ayu and Uyana.

So while BPOM pulled five Pinkflash SKUs across eleven months, Malaysia’s cosmetics enforcement was aimed at kampung-scale local operators. I’m not alleging that anything unsafe is on Malaysian shelves — I have no evidence of that, and the absence of enforcement action is not evidence of a problem. But the asymmetry of attention is worth naming out loud: the enforcement energy is going after the smallest local players while the fastest-growing foreign ones are being onboarded by the two largest pharmacy chains in the country.

One more thing, because it cuts both ways: JAKIM halal certification is voluntary for cosmetics in Malaysia and is not a condition of NPRA notification. I could not find evidence that any of the named C-beauty brands holds JAKIM certification — and I could not find evidence that they don’t. If halal is your differentiator in Malaysia, it is currently a marketing asset, not a legal barrier protecting you. Indonesia is about to make it a legal barrier. Malaysia hasn’t.

Part 9 — Seven signs you’re being out-flanked right now

Run this against your own brand this week. If you’re nodding at four or more, you’re in the trap.

1. Your entry price point is above RM25 (~USD 6) and you have nothing below it. C-beauty’s discovery tier sits at RM2.05–RM11.40. If a shopper can’t try you for the price of a drink, you’re not in the consideration set — you’re in the comparison set.

2. You have fewer than 60 live SKUs per marketplace store. Focallure runs 250+ per country. Search-ranked marketplaces reward catalogue surface area. You don’t need 250 — but at 30 SKUs you cannot cover a shade ladder, and shade gaps are where a competitor’s catalogue does your merchandising for you.

3. Your creator programme is fewer than 30 active posters, or your affiliates need hand-holding before every post. Skintific took three-quarters of its revenue from affiliates and livestreamers it did not manage individually.

4. Your product-to-shelf cycle is longer than one quarter. If your last three launches took two quarters each from insight to listing, you are being iterated around, not outspent.

5. You cannot name your top ten affiliates or reach them outside the platform. If the platform owns your creator relationships, the platform owns your revenue. See what happened to brands leaning on platform-run enablement when TikTok Shop Indonesia’s local headcount reportedly consolidated.

6. You’re relying on halal or “local brand” as your differentiator without clinical substantiation behind it. Origin erasure means the consumer often doesn’t know they have a choice to make. If your provenance claim isn’t loud, specific and paired with efficacy evidence, it isn’t doing work.

7. Your premium line is still “next year’s project.” Euromonitor said the premium segment would be left largely untouched. The Indonesian data now shows Korean derma brands taking exactly that ground. The window is open now and it closes when someone else owns the segment.

Part 10 — The playbook: what a beauty brand can actually do on Monday

This is a guide on how we’d run for a beauty, wellness or personal-care brand on our roster. Five moves, in order, because the order matters.

Move 1 — Rebuild the range for the platform, not the shelf

Build a tiered SKU ladder with a genuine discovery tier at RM9–RM19 (USD 3 to USD 5), a core tier at RM25–RM49 (USD 6 to USD 13) and a hero tier at RM59–RM99 (USD 15 to USD 25). The discovery tier’s job is not margin — it’s trial volume, review velocity and search-ranking data. Model whether you can survive it first: that’s what the Platform-Native SKU Ladder workbook attached to this issue is for. On the default assumptions it puts break-even at RM11.84, which is worth knowing before you promise anyone RM9.90.

The mistake I see constantly: a cheap SKU goes live as a loss leader with nothing built above it, so trial never converts upward. Build the ladder, not the loss leader.

Move 2 — Stand up a content engine, not a campaign

Target 15–20 shoppable videos a month minimum, plus 30+ nano and micro creators posting without individual hand-holding. That means a brief pack, an asset library, a commission structure and a community channel — not a WhatsApp group where your marketing exec approves every script.

Copy the parts of Skintific’s operation that are legal and sensible: multiple owned accounts, a creator community with visible monthly leaderboards, outsourced throughput where in-house capacity runs out. You are not trying to make better videos than them. You are trying to make enough.

We’ve written the operating detail of this before — see the affiliate and creator programme playboo] and the AI-augmented PDP issue for the production-line version.

Move 3 — Compress the product loop to one month

You can’t build six factories. You can build the read.

Once a month: pull PDP search terms, add-to-cart rates by SKU, review sentiment and competitor price movement, and convert them into pack, claim and bundle iterations — not new formulations. Reformulation is where C2M beats you. Packaging, bundling, claim hierarchy and shade extension are where you move in weeks with the lab you already use. Iterate on everything except the chemistry.

Move 4 — Own your creator graph outside the platform

Get every top affiliate’s contact into your own CRM. Run your own community. Pay outside platform commission structures where you can. This is the single highest-return defensive move available to you right now, and the evidence is unambiguous: brands with portable creator relationships survived the October 2023 shutdown and are surviving the 2026 softening. Brands renting platform-run enablement are exposed twice.

Move 5 — Start the premium migration this quarter, not after certification matures

Defend upmarket on the one axis a C2M machine can’t fake quickly: provenance and efficacy.

That means local-origin storytelling that is specific rather than sentimental, and real clinical substantiation — instrumental testing, consumer panels, numbers you can put on a PDP. Halal certification if you’re in Malaysia, and get ahead of Indonesia’s 17 October 2026 mandate if you sell there.

The thing I’d do differently from the incumbents: start now, not when the paperwork clears. The evidence from Indonesia is that the window to own a segment closes before certification matures. Korean derma brands are taking the Indonesian premium tier while local brands are still scoping their premium line.

The 90-day sequencing is in the attached plan, but the short version: Weeks 1–2 audit your SKU ladder and price-band exposure and score yourself on the defence scorecard. Weeks 3–6 recruit the creator pool to 30 and get the discovery tier live. Weeks 7–10 hit content velocity and run the first monthly product-loop read. Weeks 11–13 launch one substantiated premium SKU and migrate your creator graph into your own CRM.

The close

Amira’s question was “who are they?”

The honest answer is that they’re four companies wearing seven flags, running a channel playbook they’d already perfected in a market that looks structurally identical to ours, with factories behind them that iterate faster than our regulators can process a notification. They took 15% of a category in five years while the analysts filed them as local.

The less obvious answer is that the playbook has a shelf life. Y.O.U is softening. Wardah is climbing back. Korean derma owns the premium tier that Chinese brands left alone. TikTok Shop is giving share back to Shopee in the exact category it conquered. And five recalled SKUs in eleven months is what a three-month product cycle at USD 1 eventually costs you.

C-beauty didn’t win because they were Chinese, and they won’t lose because they were found out. They won because they treated the marketplace as the primary channel and the content engine as the product, four years before the incumbents did.

That part isn’t Chinese. That part is just correct. Take it.

Pick one of the five moves above and start it this week. If your entry price point is above RM25 (USD 6) and your creator pool is under 30 people, you already know which one.

See you next Sunday.

— Stan

Downloads

• C-Beauty Defence Scorecard (xlsx) — score your brand out of 100 across SKU ladder coverage, price-band exposure, content velocity, creator pool ownership and product-loop speed. Live formulas; plug in your own numbers.

• Platform-Native SKU Ladder Model (xlsx) — working unit economics for a three-tier ladder. Test whether you can survive an RM9–19 discovery tier before you launch one.

• 90-Day Content-First Defence Plan (docx) — week-by-week execution plan with owners, deliverables and the metrics to check at each gate.

Related reading

• Building With AI: How We Automated Marketplace Chat Without Tanking Account Health — the operating-system approach to platform automation

• The AI-Augmented PDP: Five Pillars — the production-line model for listing content at volume

• The Middle-Market Squeeze — why the RM25–RM60 (USD 6 to USD 15) band is the hardest place to stand

• Health & Wellness Deep Dive and Fashion Deep Dive — the same platform-native dynamics in adjacent categories

Sources

Sources for the principal figures are linked below. A handful of items — trademark filings, regulatory fee schedules and two paywalled datasets — are cited in the text rather than linked. Where sources disagree, I’ve said so rather than picking the more dramatic number.

Market share and category data

• Euromonitor International, The Rise of C-Beauty Brands in Southeast Asia, 2025

• Euromonitor via Fortune, Chinese beauty brands go global, 9 June 2026

• SCMP, Look out, L’Oreal: Dazzle Me, Y.O.U and Chinese cosmetic brands are coming for your market, 14 August 2025

• BeautyMatter, C-Beauty’s New Playbook for Southeast Asia, 19 July 2026

• Compas, Top 10 Brand Perawatan & Kecantikan di Shopee Q1 2025

• Databoks/Compas, 10 Best-Selling Eyeliner Brands on Shopee Indonesia Q2 2025, 27 October 2025

• Databoks/Compas, Best-selling lip tint brands on Shopee Indonesia, 8 April 2026

• GoodStats/Compas, 10 Brand Lip Tint Terlaris di Shopee Indonesia Q1 2026, 6 May 2026

• Magpie IQ, Top Skincare Brands in Indonesia 2026, 24 July 2026

• Magpie IQ, In Indonesian Beauty, Shopee’s Lead Over TikTok Shop Doubled, 3 July 2026

Ownership and corporate structure

• PR Newswire, HEBE BEAUTY: Pioneering Beauty Industry in Emerging Markets, 30 September 2022 — also the source for the 3INS halal-plant joint venture, the Guardian Indonesia Best New Brand 2019 award, the 40,000-door count and the 25%-online figure

• Beauty Journal, Brand lokal MUAQ — Muaq’s Indonesian ownership

• TNGlobal, BGlobal Summit: From “Chinese DNA” to “Local Unicorns”, 18 June 2026

• ANTARA, Brand kosmetik asal Korea Selatan barenbliss resmi hadir di Indonesia, 25 August 2021

• 36Kr PitchHub, Focallure company and funding record

• PingWest, Focallure, SEA’s most selling Chinese makeup brand, secures 8-digit funding, September 2021

• Kompas, Siapa Pemilik Skintific?, 18 May 2025

Skintific

• TMO Group, How Skintific dominates Southeast Asia’s Beauty and Skincare eCommerce, 7 October 2024

• BERNAMA, Skintific Malaysia launch, 31 July 2023

• Atlantis Press / ICVEAST, Skintific TikTok Shop Indonesia case study, Universitas Indonesia, 2023

• Nikkei BizRuptors, Skintific country-of-origin case study, 8 November 2023

• SCMP, Are Chinese-made skincare brands downplaying their origins?, 5 May 2024

• BusinessMirror, Skintific: science meets beauty (Philippines), 18 August 2025

• Compas, Market insight: paket kecantikan Skintific, 2024

Pinkflash recalls

• Suara.com, 3 produk Pinkflash yang ditarik BPOM, 3 December 2024

• Detik Health, Daftar produk kosmetik Pinkflash yang ditarik, December 2024 — recalled-SKU list with registration numbers

• Detik Health, Pinkflash tawarkan ganti rugi, November 2025 — compensation terms

• Suara.com, Produk Pinkflash apa yang dilarang BPOM, 6 November 2025

• Suara.com, BPOM rilis 23 kosmetik berbahaya, 3 November 2025

Channel and platform data

• MarketScreener, TikTok Inc. completed the acquisition of a 75.01% stake in PT Tokopedia, 31 January 2024

• People Matters SEA, Is TikTok Shop gearing up for job cuts in Indonesia? — the reported headcount consolidation

• Metric via VnExpress, TikTok Shop thu hẹp khoảng cách thị phần với Shopee, 25 July 2025

• YouNet ECI, TikTok Shop bành trướng trong nửa đầu 2025, 5 August 2025

• Cube Asia, Shopee, Lazada and TikTok Shop in Southeast Asia: what the data shows in 2026, 2 June 2026

• Momentum Works, Ecommerce in Southeast Asia 2026, 14 April 2026

• Malay Mail, TikTok Shop closes in on Shopee — Ipsos survey, 16 July 2026

Retail and K-beauty contrast

• AS Watson, Watsons’ partnership with K-beauty brands Mise En Scène and COSRX, 21 October 2021

• Product Nation, Judydoll launches at Watsons Malaysia, 11 October 2024

• The Edge Malaysia, Guardian Malaysia store count, 14 May 2025

Regulatory

• NPRA Malaysia, Cosmetic notification guidelines

• NPRA Malaysia, Cancelled cosmetic notifications

• Ministry of Health Malaysia, Are cosmetics in Malaysia certified halal?

• BPJPH, Cosmetics must be halal certified by October 2026

• US ITA, Malaysia Beauty and Personal Care, updated 19 May 2026

Incumbents

• Nikkei Asia, Indonesia beauty brands struggle in battle with Chinese rivals

• BeautyMatter, Thai beauty brand Mistine raises $31 million Series A, 4 April 2022

• DealStreetAsia, ESQA raises $4m Series B

Ecommerce Owner’s Manual is a weekly Sunday newsletter from Stan Chew, founder of Commerceplus — the Southeast Asian eCommerce enabler for brands serious about marketplace growth. Subscribe at ecommercesea.substack.com.

Suraya manages a household and personal-care brand — dishwash, floor cleaner, a laundry range, a couple of RTD health drinks she added last year. Solid business. Shopee, Lazada, a TikTok Shop she’s still figuring out. When we sat down last month she pulled up her own dashboard and turned the laptop around.

“Look at this. My traffic is up. My listing views are up. My ‘add to cart’ is up. Everything at the top of the funnel is up. And my new-buyer count is dead flat. It’s the same number as a year ago.”

She wasn’t wrong, and she wasn’t imagining it. More people were looking at her cleaning products more often — treating a bottle of floor cleaner the way they used to treat a dress they might or might not buy. Browsing. Comparing. Adding to cart, sitting on it, waiting for a coin drop. But the pool of humans buying wasn’t growing.

That’s the whole issue, right there. And the answer is uncomfortable, because it means the thing you were good at — acquiring the next new buyer — matters less than it did, and the thing most essentials brands are bad at — owning the reorder — is now the entire ballgame.

She is not alone. I hear this every week now. So this is that memo.

I wrote a full grocery and food deep dive in May about the supply side — the three operating models, the brutal unit economics, cold chain, who wins. This issue is the other half of the same story: the demand side. What the shopper is actually doing now, why essentials started behaving like fashion, and how you capture routine spend when nobody new is walking in the door.

The reckoning: browsing is at a three-year high, and it isn’t new people

Let me put the numbers on the table, because they reframe everything.

The Ipsos read I’ve been sitting with this month shows a sharp jump in what Southeast Asians say they now buy online in everyday categories, in a single year:

• Grocery purchases: 34% → 47%.
  Nearly half of shoppers now buy groceries online — up thirteen points in twelve months.

• Beauty: 29% → 45%.

• Healthcare and supplements: 15% → 25%.

• Online browsing sits at a three-year high — around 74%.

Read those first three lines and the temptation is to celebrate. Category’s booming, get in. But hold them next to the fact that matters most: the number of people buying online is basically flat. Buyer penetration in the mature SEA markets has plateaued. The easy years — when every quarter delivered a fresh cohort of first-time online shoppers who just needed a platform and a payment method — are behind us. The rails are built. Almost everyone who’s going to shop online already does.

So what’s actually growing? Not the crowd. The behaviour.

The same people are browsing more, across more categories, more often — and they’re moving spend that used to happen at the AEON, the pharmacy, the 99 Speedmart, onto the same apps where they already buy their clothes and their skincare.

That’s the shift. Essentials didn’t win a new audience. They colonised an existing habit.

Browsing at a three-year high with flat buyers means one thing for your P&L: growth is no longer about how many people you acquire. It’s about what share of an existing shopper’s routine spend you can capture and keep. Share of wallet. Frequency. Reorder. That’s the game now, and it’s a completely different game from the one most essentials brands are set up to play.

Why “grocery is the new fashion”

I don’t say this to be cute. The comparison is mechanical, and it tells you exactly what to build.

Fashion was the category that taught Southeast Asians to browse on Shopee and Lazada. You didn’t open the app because you’d run out of anything. You opened it out of habit — to scroll, to see what’s new, to fill a cart you might never check out, to wait for a voucher. Fashion made the marketplace a place you visit, not a place you go to when you need something. High frequency, low intent, driven by discovery and price theatre.

For years, groceries and household essentials were the opposite. Planned. Occasional. You did a “haul” — stocked up during 11.11, bought the big bundle, and disappeared for two months. Low frequency, high intent, driven by need.

What the Ipsos numbers are telling us is that essentials have crossed over. They now behave like fashion:

• Habitual, not occasional. People open the app and browse cleaning products, snacks, vitamins, baby wipes the way they browse tops and shoes. The reorder has become a scroll.

• Discovery-driven, not just need-driven. A cashback banner or a creator’s fridge-restock video triggers a purchase that no shopping list called for.

• Frequency is the frontier. In fashion, the winners learned that the same customer coming back eight times a year beats eight new customers once. Essentials are now exactly that shape — except essentials have something fashion never had: a natural, biological reorder clock. You genuinely run out of laundry pods. You don’t “run out” of a dress.

That last point is the gift. Fashion had to manufacture frequency with newness and FOMO. Essentials have frequency built into the product — if you build the machine to catch it. Most brands don’t. They treat every reorder like a fresh acquisition, pay to reach a customer they already own, and wonder why the maths never works.

The fashion analogy also carries the warning. Fashion on these platforms is a bloodbath of discounting, coin games, and zero loyalty — deal-seekers, not loyalists. If essentials become fashion without you owning the reorder relationship, you inherit fashion’s margins and none of its excuses. Which brings me to the part nobody wants to talk about.

The trading-terms squeeze: the platform is buying your customer’s habit with your margin

Here’s the move happening underneath the growth story.

Shopee and Lazada have worked out the same thing you just read: buyer count is flat, so the war is over frequency and share of wallet. Their answer is to buy the behaviour — to train shoppers to open the app daily and to spread spend into essentials — using cashback coins, scan-and-earn rebates, coin cashback, free-shipping thresholds, and platform-funded voucher theatre. And increasingly they’re funding that land-grab by raising trading terms on sellers: higher effective commissions, more “recommended” seller-funded campaign contributions, more pay-to-stay-visible mechanics.

For a brand running its own branded storefront, this creates a nasty pincer🦀:

• On one side, your cost to play goes up. Effective take rates climb as campaign participation stops being optional — opt out and your search ranking and traffic quietly die.

• On the other side, the platform is actively re-training your customer to chase coins and cashback rather than your brand. Every cashback-driven purchase teaches the shopper that the reason they bought is the platform’s rebate, not your product. You’re paying rising terms to fund the erosion of your own loyalty.

The retail verticals — Shopee Supermarket, LazMart, RedMart — accelerate this. When the platform merchandises the category itself, it controls the promotion, owns the customer data, and decides which brand’s SKU gets discounted to drive its basket, not yours.

I’m not telling you to boycott any of this — you can’t, and I explained why in The Tightening Squeeze when Shopee reworked its fee structure. Marketplaces are non-negotiable distribution. But you have to be clear-eyed about what the platform is optimising for: it wants to own the habit. If your entire relationship with the customer lives inside an app that’s actively converting brand loyalty into coin loyalty, you don’t have a moat. You have a rental agreement, and the rent is going up.

The only durable answer is to own the reorder relationship somewhere the platform doesn’t control — which is the back half of this issue. First, the maths, because it’s the maths that makes the case.

Category economics: low AOV × high frequency is a different sport

Beauty brands obsess over CAC. Fashion brands obsess over return rates and newness. Essentials brands have to obsess over one thing: frequency compounding.

The trap is that essentials look terrible on a single-order basis. Low average order value, thin gross margin, and platform fees that don’t shrink just because your product is cheap. Run the first order in isolation and you almost always lose money. I walked through the full RM100 contribution waterfall in the May deep dive — the punchline was that outright marketplace contribution on a single grocery order is frequently negative.

But essentials aren’t a single-order sport. Watch what happens when you stop looking at the order and start looking at the year.

Take a realistic essentials SKU — call it a $45 basket, 32% gross margin, so about $14 of gross profit per order. Acquisition costs you $18 all-in. On order one, after platform fees, ads and fulfilment, you’re underwater. Classic essentials pain.

Now put the reorder clock on it. This is a consumable on roughly a 35-day cycle. If you do nothing, the customer might drift back once or twice a year on their own, and you’ll re-pay to reach them each time — you’ve built nothing. But if you own the reorder — a reminder at day 28, a subscribe-and-save bundle, a one-tap WhatsApp re-order — that same customer can hit six, eight, ten orders in a year:

• Order 1: below water after CAC. Painful, and where most brands stop reading their own numbers.

• Order 2 (≈day 35): no CAC this time. You’re now positive on the relationship.

• Orders 3–10: almost pure contribution. CAC is long amortised. Fulfilment gets cheaper as baskets consolidate. This is where the money is.

The single most important number in essentials is not AOV and not CAC. It’s orders per customer per year. Lift a cohort from 1.4 orders/year to 4.5 and you haven’t improved the business by a bit — you’ve changed what business you’re in. Low AOV stops being a weakness the moment frequency is high and you’re not re-paying to acquire on every cycle.

That’s why I keep saying the same sentence to essentials founders: stop trying to raise the basket, start trying to raise the frequency. A 15% bundle discount that locks in a 5-week reorder cadence is worth more than any 11.11 hero-deal spike, because it converts a one-night stand into an annuity.

I’ve built a working model for this — the Repeat-Economics Model attached with this issue. Plug in your AOV, margin, CAC, reorder cycle and retention curve and it shows you your break-even order number, your 12-month LTV, and your true LTV:CAC once frequency is doing the work. Most essentials brands have never run this maths on their own SKUs. When they do, the strategy writes itself.

The framework: The Replenishment Calendar

If frequency is the game, you need to stop treating your catalogue as a list of products and start treating it as a set of clocks. Every consumable essential has a natural reorder rhythm. The brands that win map that rhythm and build a machine around it. I call it the Replenishment Calendar, and it has three moving parts.

1. Sort every SKU onto a reorder clock

Not all essentials replenish at the same speed. Put each hero SKU on its clock:

• 7–14 day clock: fast consumables — fresh-ish snacks, pet food for a single pet, baby wipes, certain RTD drinks. Highest frequency, highest lifetime value if held, hardest to hold.

• ~30 day clock: the sweet spot — supplements, coffee, laundry, personal care, household cleaning. One clean monthly cadence you can build an entire CRM rhythm around.

• 60–90 day clock: slower burn — bulk household, vitamins bought in 90-count, refill packs. Lower frequency but very high intent and very predictable.

You don’t need data science for this. You know your product. Look at your actual repeat-purchase intervals in the platform back-end and you’ll see the clock clearly.

2. Size the bundle to the clock

This is where most brands leave money on the table. Your default sale unit should be sized to the reorder interval, not to what’s easiest to pack. If the product is a 30-day consumable, your hero listing should be the pack that lasts about 30 days — or a subscribe-and-save that auto-ships on that cadence. Bundle sizing does three things at once: it lifts AOV honestly (not with a gimmick), it improves fulfilment economics (fewer, bigger orders beat many tiny ones), and it sets the reorder clock for you by controlling when the customer runs out. A well-sized bundle is a scheduling decision disguised as a pricing decision.

3. Build the reorder engine — and put it where the platform can’t tax it

The reminder that lands the day before the customer runs out is the highest-ROI marketing you will ever run, because intent is already at its peak — they’re about to rebuy anyway; the only question is from whom. The engine has three lanes:

• WhatsApp / Chat Broadcast / Meta retargeting Ads Reorder flow. Still the most reliable owned reorder channel in Malaysia, Singapore and the Philippines. A day-28 message — “you’re about due, tap to reorder your usual” — with a one-tap checkout converts at rates paid ads can’t touch.

• Subscribe-and-save. A 10–15% discount for auto-replenishment. Shopee and Lazada’s native subscription tooling is still weak, which is exactly why WhatsApp + a light shopfront integration remains the workhorse. The prize: roughly seven in ten customers who consent to a 4–6 week cadence stay for six-plus orders.

• Platform re-engagement, used deliberately. Follow-up vouchers and Shopee/Lazada CRM tools still have a role — but treat them as the paid lane, not the whole strategy. Every reorder you can move from the platform’s coin economy to your owned WhatsApp flow is margin you keep and a habit you own.

The Replenishment Calendar is the difference between hoping the customer comes back and scheduling it. Fashion never had this lever. You do. Use it.

Where TikTok Shop fits for essentials

A quick, important aside, because founders keep miscasting TikTok Shop in this category.

TikTok Shop is not where the reorder happens. It’s where the trial happens. Its superpower for essentials is discovery-driven trial at the top of the funnel — the fridge-restock haul, the “clean my whole kitchen with this” demo, the supplement routine video that makes someone try a brand they’d never have searched for. F&B, RTD beverages and snacks are among the fastest-growing TikTok Shop categories in Malaysia, Thailand and Vietnam, and live commerce moves genuinely large volume on impulse.

So the right mental model is a relay:

• TikTok Shop = trial and discovery. Content and live drive first purchase and awareness. Cheap attention, high impulse, weak retention.

• Shopee / Lazada = the search-and-restock basket. Where the planned reorder lands when the customer knows what they want.

• WhatsApp / owned = the reorder annuity. Where you convert the trial into a scheduled habit you own.

The mistake is running TikTok Shop as a standalone P&L and judging it on repeat rate — it’ll look terrible, because that’s not its job. Judge it on cost per first trial, then make sure you have a Replenishment Calendar waiting to catch the customer on the other side. Trial without a reorder engine is just paying to fill a leaky bucket faster.

The rise of Shopee Supermarket and RedMart

Two developments worth watching closely, because they change where the essentials shopper lands.

Shopee Supermarket has become the default destination for planned grocery baskets in Malaysia and across much of the region. It’s the retail model I described in May — you vend into the platform, they merchandise and fulfil, they own the customer. For a Tier-1 FMCG brand with case-pack discipline it’s clean, high-volume distribution. For a challenger brand it’s a trade: reach in exchange for relationship. As essentials go habitual, Shopee Supermarket is where a growing slice of that routine spend consolidates — which is precisely why the platform is investing in it and why it’s using cashback to pull shoppers in.

RedMart (Lazada-owned) remains the dominant online grocery destination in Singapore, where penetration is the highest in the region — an estimated 12–18% of grocery spend already online, versus roughly 5–8% across most of SEA. Singapore is the market where essentials-as-fashion is furthest along: high frequency, high AOV (SGD 60–120 baskets), exacting service expectations, and a shopper who genuinely reorders staples online as routine. If you’re building a premium or health-positioned essentials brand, Singapore via RedMart is the cleanest place to prove the replenishment model before you scale it into the messier, lower-penetration markets.

The strategic read on both: the platforms are building the retail rails for the exact behaviour shift this issue is about. You can ride those rails for distribution — but do it with your eyes open about what you’re giving up, and never let them become the only place your customer relationship lives.

Are you renting share, or owning it? Six signs you’re renting

A quick diagnostic. If three or more of these are true, you don’t own your customer’s routine — the platform does, and you’re paying rising rent for the privilege.

🔴 Your reorders come from vouchers, not reminders. If you switched off coin-funded promos tomorrow, would repeat purchases survive? If you don’t know, that’s the answer.

🔴 You can’t state your orders-per-customer-per-year. The single most important number in essentials, and most brands can’t produce it from memory.

🔴 You pay to re-reach customers you already acquired. Retargeting your own existing buyers with paid ads is a tax you’re volunteering to pay because you have no owned reorder channel.

🔴 Your hero listing is a single unit, not a clock-sized bundle. You’re leaving frequency, AOV and fulfilment economics on the table, and letting the customer decide when they run out.

🔴 You have no day-28 message. No WhatsApp flow, no reminder, nothing that reaches the customer at peak reorder intent. The most valuable marketing moment in the category and you’re absent for it.

🔴 Your only growth lever is new-buyer acquisition. In a flat-buyer market, this is a treadmill. The share-of-wallet brands have stopped sprinting on it.

Three or more? You’re renting. The exit is building a reorder relationship the platform can’t tax — which is entirely doable, and which one of our own clients has quietly turned into their whole growth engine.

Mini-case: a wellness brand that stopped chasing new buyers

I’ll keep this one anonymous — it’s a Klang Valley wellness and supplements brand we operate for, and the numbers are theirs, not a composite.

Eighteen months ago they had the exact problem Suraya described: traffic fine, new buyers flat, growth stalling, margins bleeding into 11.11 and 12.12 discounts that spiked revenue and then vanished. Every quarter started from zero. Classic acquisition treadmill.

We stopped optimising for the next new buyer and rebuilt the whole operation around the reorder. Three moves:

• Mapped every hero SKU onto its reorder clock. Their flagship supplement was a clean ~30-day consumable. We made the 30-day pack the default hero listing instead of the single bottle, and added a subscribe-and-save at a modest discount.

• Built the day-28 reorder flow. A single, well-timed reminder with one-tap reorder, plus a light re-engagement sequence for lapsed customers. Owned channel, no platform tax.

• Held the line on discounting. We stopped funding deep coin promos to buy hollow spikes and redirected that budget into the reorder engine and into TikTok Shop trial at the top.

The repeat rate on the flagship SKU climbed from the low-20s into the 40s over roughly three quarters. Orders per customer per year roughly doubled. And because the incremental orders carried no fresh CAC, contribution margin went positive on a cohort that used to lose money — without acquiring a single additional net-new buyer in that period. Same audience. Far more of their wallet.

None of this was exotic. It was picking the reorder clock, sizing the bundle to it, and showing up at day 28. The compounding did the rest — which is the whole lesson of this category.

The playbook — what to do Monday morning

Five moves, in order.

1. Pull your orders-per-customer-per-year, by hero SKU. Before anything else. If you can’t get it cleanly from the platform back-end, that gap is your first project. You cannot manage frequency you don’t measure.

2. Put every hero SKU on a reorder clock. 7–14 day, ~30 day, or 60–90 day. Write it down. This single act reorganises your merchandising, your CRM cadence and your bundle strategy.

3. Re-size your hero listing to the clock. Make the default sale unit the pack that lasts one cycle, and stand up a subscribe-and-save on that cadence. Stop leading with the single unit.

4. Build the day-28 message. Stand up a reorder flow using WhatsApp, Chat Broadcast, Meta Custom Audiences targeting users on a certain purchase window, that reaches the customer just before they run out, with one-tap reorder. This is the highest-ROI thing on this list. Do it this month.

5. Re-cast TikTok Shop as trial, not reorder. Judge it on cost-per-first-purchase, and make sure the Replenishment Calendar is there to catch what it brings in.

Run the Repeat-Economics Model on your own SKUs before you argue with any of this. When you see your real break-even order number and what frequency does to LTV:CAC, the case makes itself.

The close

The old growth story was addition: find the next new buyer, then the next. That story is over in the mature SEA markets — the buyers are already here, browsing at a three-year high, spreading a fixed wallet across more categories and more brands than ever.

The new story is retention and rhythm. Essentials became fashion — habitual, high-frequency, browse-driven — but with one advantage fashion never had: a reorder clock built into the product. The platforms know it, and they’re spending to own that habit before you do.

Own the reorder, or rent your share of the wallet from a landlord who keeps raising the rent.

Pick one hero SKU this week, put it on its clock, and build the day-28 message. That’s the whole game now.

Keep building!

Related reading from the archive:

• Grocery & Food E-commerce Deep Dive — the supply-side companion to this issue: three operating models, unit economics, cold chain (May 2026)

• The 73% Opportunity: Converting SEA’s Offline Shoppers — where the next wave of demand comes from when buyer counts plateau (Nov 2025)

• Health & Wellness DTC Strategy — the category most exposed to the replenishment shift

• Fashion E-commerce Deep Dive — the browse-and-buy habit essentials are now inheriting (the contrast)

• The Tightening Squeeze — Shopee Fee Changes — why trading terms keep climbing

Two weeks ago, in her 7.7 planning meeting, I watched the machinery she’s built work against her. The Shopee lead presented last year’s campaign settings “plus 15%.” The TikTok lead had a livestream calendar that nobody had reconciled against the paid media plan. The performance agency proposed a budget that assumed peak-day conversion rates nobody could actually verify, because nobody had kept them. Five channels, real budgets, capable people — and the campaign plan was, functionally, a photocopy with a stretch target written on it.

Here’s what struck me: Mei Lin’s problem isn’t effort or talent. It’s that her organisation runs campaigns as events instead of running them as instruments. Every double-digit day resets to zero. Nothing compounds. And in four months, when ad auctions are 40–60% more expensive and every decision is playing for the biggest revenue window of the year, her team will be making its most consequential calls on vibes and last year’s folklore.

The thesis of this issue: Q4 is not won in Q4. It’s decided in July — by whoever treats 7.7 and the double-digit days that follow as an instrumented rehearsal, and builds the operating cadence to make each campaign inherit the last one’s lessons.

Subscribe now

The Setup: Mid-Year Is Now Mega-Sale-Sized

For years, the mental model was simple: double-digit dates from January to August are warm-ups, and the “real” campaigns are 9.9, 11.11, and 12.12. That model is dead, and the logistics data killed it.

J&T Express just reported its Double 6 (6.6) numbers for Southeast Asia: 213 million parcels during the core campaign window, up 83.5% year-on-year. Daily average volume hit 35.51 million parcels, and on the peak business day — 2 June — single-day volume crossed 37 million. Indonesia grew 112% campaign-over-campaign; Thailand grew 89%.

Put that in context: the entire Southeast Asian ecommerce market ships roughly 43.6 million parcels a day in normal conditions, across all couriers. One courier, during one mid-year campaign, was moving 35 million a day on its own. Mid-year is no longer a warm-up. It is a full-dress production with mega-sale volumes — running through the same infrastructure, ad auctions, and consumer wallets that will face the real test in Q4.

And the market it sits inside keeps compounding. SEA platform GMV reached US$157.6 billion in 2025, up 22.8%, with Shopee holding roughly 53% share and growing Q1 2026 GMV 30.2% to US$37.3 billion. Three platforms — Shopee, TikTok Shop, Lazada — now take 98.8% of platform GMV. For a brand at Mei Lin’s scale, the campaign calendar is not a promotion schedule. It is the operating rhythm of the entire market she competes in.

Right now, as you read this, Shopee Malaysia’s 7.7 Mid-Year Sale is live: 26 June to 10 July, peak day 7 July, with peak-day mechanics unlocking from 12 AM on 6 July — daily 22%-off vouchers at RM40 minimum spend, RM800 vouchers at RM1,600 minimum, daily free shipping, Shocking Sale time-drops, and a packed livestream calendar.

If you’re reading this on Sunday 5 July: you have roughly 48 hours before peak day. This issue is sequenced accordingly — the first three frameworks are things your team can still do before Tuesday; the rest is what you build in the weeks after.

Framework 1: The H2 Arc — Test, Refine, Scale, Harvest

Stop treating the five big H2 dates as five separate campaigns. They are one campaign in five acts, and each act has a different job:

The discipline this demands is uncomfortable: at 7.7, you deliberately accept a slightly lower peak-day number in exchange for clean data. A team that runs three voucher structures across three comparable SKU groups learns which threshold moves AOV. A team that runs one blended “everything on” campaign learns nothing except its topline.

Why does the sequencing matter so much? Cost. Meta’s average CPM ran US$14.19 in 2025, up 20% year-on-year — and within the year, the November peak hit US$25.22 against a January low of US$15.74. Global median CPC peaked at US$1.32 in November 2025 versus US$0.85 in January. Across markets, Q4 CPMs run 30–60% above the rest of the year. The identical A/B test costs materially more to run in November than in July — and a failed November test costs you your best inventory during your biggest window. July failure is cheap tuition. November failure is a write-off.

Framework 2: The Peak-Day Baseline — Backdate Before You Forecast

Before you can test anything, you need to know what “normal” looks like on an abnormal day. This is the exercise I’d put Mei Lin’s team on this week, and it’s the first thing we build when we take over a multi-channel P&L: the backdated peak-day baseline.

The build is unglamorous and takes a channel lead about half a day per channel. Pull every double-digit campaign from the last 12–18 months — 9.9, 10.10, 11.11, 12.12, CNY, 6.6 — and for each one, per channel, record the peak-day and campaign-window numbers for the variables that actually run your revenue:

Revenue = Sessions × Conversion Rate × AOV — minus the variable costs that decide whether the revenue was worth having: ad spend, platform commissions, affiliate commissions, voucher subsidy.

This is the Sales Growth Formula applied to campaign days — the same decomposition we run every week on managed brands, pointed at the calendar. Lay it out as a grid: channels down the side, campaigns across the top, the formula’s variables in each cell.

Three things happen the moment this grid exists, and I’ve watched it happen across automotive, fashion, personal care, and FMCG accounts:

First, the channels stop being folklore. “TikTok converts better during campaigns” becomes “TikTok Shop peak-day CVR runs 1.8× baseline but AOV drops 12% because livestream skews to entry SKUs.” Those are different facts with different implications.

Second, forecasting becomes arithmetic instead of ambition. Your 7.7 forecast per channel is last mid-year’s actuals × your observed campaign-over-campaign growth × any structural change you can name (new SKU lines, a livestream programme that didn’t exist last year). “Last year plus 15%” is not a forecast; it’s a wish with a percentage on it.

Third — and this is the real prize — the teams synchronise. The grid forces the Shopee lead, the TikTok lead, the performance agency, and the livestream team to reconcile their numbers in one place, against one formula, before the campaign. That reconciliation meeting is where the coordination problems surface: the paid plan that assumes traffic the livestream calendar can’t support, the voucher plan that will cannibalise the bundle test. In our experience, the baseline exercise does more for team cadence than any number of alignment meetings — because it replaces opinions with a shared denominator.

Do this for 7.7 even if it’s rough. Then repeat for 8.8 and 9.9 with one campaign’s more data each time. By 11.11, your forecast isn’t a stretch target — it’s a model with three campaigns of calibration behind it. That is operational muscle, and it compounds exactly the way ad spend doesn’t.

Here’s a downloadable template for your reference. Link to H2 Planning Template

Framework 3: The Growth Formula Test Grid — What to Actually A/B

“Instrument your campaign” is useless advice without specifics. Here is how we choose experiments for a mid-year campaign: every test must move a named variable in the growth formula, and every variable gets at most one test per campaign. If you change the bundle and the voucher and the livestream slot on the same SKUs, you’ve run zero experiments, not three.

The levers, mapped to the variable they move:

Sessions (traffic) levers:

• Paid channel mix — shift a measured slice of budget between Meta, TikTok, and platform ads against the same SKU set, and compare cost per session and downstream CVR by source. Campaign weeks compress the feedback loop; you’ll see in days what takes weeks in BAU.

• Ad creative — campaign traffic volumes give you statistically usable creative reads fast. Two or three concepts per hero SKU, winners banked for Q4. Your 7.7 top performers are your 11.11 starting lineup.

• Shoppable video — short-form video with product anchors is now a first-class traffic source on TikTok Shop and increasingly on Shopee Video. Test posting cadence and video-to-live handoffs; measure sessions and attributed orders per video.

Conversion rate levers:

• PDP images — the first three images on your hero listings are a conversion asset, not decoration. Test a benefit-led first image against the plain product shot on comparable SKUs. When we run performance-merchandising audits on client catalogues, the image stack and title are consistently the highest-leverage CVR fixes — and campaign traffic gives you the sample size to prove it quickly. One rule: freeze listing content during a test window. If titles and images change mid-campaign, your CVR reads are garbage.

• Livestream slots and formats — during Malaysia’s 11.11, Shopee Live sellers saw a 6× sales uplift versus normal days, but that average hides enormous variance by slot, host, and format. Run your streams across two distinct windows (8 PM prime versus lunch; peak-day midnight versus evening), hold format constant, and track GMV per stream hour, new-follower rate, and voucher claims per viewer. Slot economics learned in July become host bookings in October — when the good hosts and agency slots are gone.

AOV levers:

• Bundle configuration — not discount depth, configuration: hero + hero, hero + accessory, hero + trial-size of the SKU you’re seeding for Q4. Measure AOV delta and margin per order. A bundle that lifts AOV 25% while giving away 18% margin loses to one that lifts 15% at 6% cost. A fashion accessories client of ours ran exactly this test in a mid-year window — a two-piece hero bundle against a hero-plus-small-leather-goods pairing. The pairing won on margin-adjusted AOV, carried near-zero incremental returns, and shaped every bundle submission for the rest of their year. The campaign’s topline was unremarkable; the decision it produced was worth more than the campaign.

• Virtual bundles — platform-native virtual bundling (combine listings without re-warehousing combined SKUs) lets you test pairings with zero inventory risk. Test aggressively here first; physically kit only the proven winners for Q4, when packing bench time is your scarcest resource.

• Voucher thresholds — every category has a threshold cliff: the minimum-spend level where the voucher stops stretching baskets and starts subsidising orders that would have happened anyway. If your AOV is RM95 and your RM80-minimum voucher redeems heavily at RM82–88 order values, you are paying customers to do what they were already doing. Test thresholds above natural AOV until redemption behaviour visibly stretches basket size — then lock that threshold for 11.11, when voucher subsidy at scale can quietly eat 3–5 points of margin.

Variable cost levers:

• Affiliate commission structure — test tiered commission (base rate plus campaign-window bonus) against a flat elevated rate on matched SKU sets. Measure affiliate-attributed GMV per commission ringgit, not raw affiliate GMV. More on why this matters urgently in the next section.

• Co-funded mechanics — where the platform is matching your voucher or traffic investment, your effective subsidy cost per order changes. Track subsidised versus unsubsidised cohorts separately or your unit economics will flatter you.

The output of the grid isn’t a report. It’s a decision list: which channel mix, which creatives, which bundle architecture, which threshold, which commission structure — each one locked, written down, and inherited by the 9.9 brief.

Framework 4: The Relationship Moat — Affiliates and Platforms

Here’s the part of H2 preparation that doesn’t show up in any dashboard, and it’s where operators quietly separate from everyone else: the second half is a relationship market, and the price of relationships rises every week you wait.

Affiliates first. As Q4 approaches, every seller in your category will discover affiliates at the same time and respond the same way — throwing higher commissions and paid placement fees at the same finite pool of creators. By late October it’s a commission war, and in a commission war the affiliate’s rational move is to promote whoever pays most this week, with zero loyalty and zero category understanding.

The counter-move happens now, in July and August:

• Lock your top performers early. Identify the 10–20 affiliates who actually drove your H1 GMV (per-affiliate GMV, not follower counts). Offer them something the commission war can’t: guaranteed campaign-window rates agreed in advance, early samples of Q4 hero SKUs, first access to your best voucher stock, and a named person on your team who answers them within hours. We’ve built affiliate programmes across categories, and the pattern holds everywhere: the brands that win Q4 affiliate share are the ones whose top creators already know the products, already have content drafted, and already trust the payout — before the bidding starts.

• Use 7.7 as the audition. The campaign-window commission test from the grid above doubles as talent scouting: the affiliates who over-index this week at moderate commission are the ones worth locking for November at premium rates.

• Feed them like a channel, not a coupon. Content briefs, product education, exclusive angles. An affiliate armed with your differentiation story converts; one armed only with a commission rate promotes your competitor tomorrow.

Platforms second. At the USD 5–50M scale, you are big enough for the platforms to invest in — if you show up as a partner, not a listing. Two mechanics matter most for H2:

• Co-funded vouchers. During mega campaigns, platforms co-invest in voucher pools for brands they nominate — effectively subsidising your discount depth. Access runs through your account manager, campaign submission deadlines, and your performance history. The brands that get the good co-fund allocations for 11.11 are having those conversations in July and August, with a 7.7 performance story in hand — not emailing their AM in the last week of October.

• TikTok Shop’s BXP and platform co-investment programmes. TikTok’s brand co-investment mechanics reward committed campaign participation with matched incentives and traffic support. The structures shift campaign to campaign — sit down with your account team now and map what committing to the full H2 arc unlocks versus showing up date by date. (As always with platform programmes: get the current mechanics in writing from your AM; they change faster than any newsletter can track.)

None of this is a hack. It’s the compounding return on being a professional counterparty — forecasts you hit, submissions on time, co-fund budgets honoured. Platforms and affiliates both allocate their scarce resources to brands that make them look good, and that reputation is built across campaigns, not inside one.

Framework 5: The 3-Number Capacity Plan

Fulfilment is the section most heads of ecommerce skip, and it quietly decides whether Q4 makes or loses money. The brands that break in November aren’t the ones that under-ordered stock — they’re the ones that had stock and couldn’t ship it inside the ready-to-ship SLA, then ate the late-shipment penalties and search-ranking demotion that follow.

Your mid-year campaign hands you the three numbers you need:

1. P — Peak-day orders: your actual 7 July count, per channel.

2. T — Packing throughput: orders packed per person-hour, measured during the 7.7 rush. Put someone on the bench with a stopwatch.

3. M — Q4 multiplier: your own 11.11-to-mid-year ratio from the baseline grid. No history? Plan at 2.5–3× until your data says otherwise. Hope is not a capacity plan.

P × M ÷ T = person-hours needed on peak day.

A brand doing 3,000 orders on 7.7 peak day, at 18 orders per person-hour, planning for a 3× 11.11: 9,000 ÷ 18 = 500 person-hours — thirty-plus packers on long shifts, or a decision made now, in July, to split the load: inbound your top-velocity SKUs into platform fulfilment (SPX already delivers 50% of Shopee’s SEA parcels in under two days — use that network), keep the long tail on your own bench, pre-book 3PL overflow for the peak-72-hours. Every one of those options is cheap and available in July. Every one is expensive or gone by late October — FBS inbound cutoffs ahead of 11.11 typically land mid-to-late October, and warehouse slots fill early.

Framework 6: The 48-Hour Debrief

This is the SOP we hold ourselves to on every managed campaign, across every category we run: within 48 hours of peak day, the campaign gets a written debrief. Not a month later in the quarterly review, when the ops pain is forgotten and the numbers have been rounded into folklore. Within 48 hours, while the packing bench is still sore.

Why 48 hours? Decay and the calendar. The qualitative detail — which voucher confused customers, which stream slot died, which courier pickup got missed — evaporates within days. And your 9.9 submission windows open within weeks. The debrief is the 9.9 brief.

Five blocks:

Block 1 — The Numbers vs. The Forecast. GMV, orders, AOV, sessions, CVR, ad spend, blended ROAS — each against the baseline-grid forecast. Variance against a forecast is a lesson; variance without one is just a number.

Block 2 — The Funnel Read. Sessions → CVR → AOV per channel and per hero SKU, campaign window versus the 30 days prior. This tells you whether the campaign created demand or merely time-shifted it — a pre-campaign traffic crater means your customers have learned to wait for the sale, which is a pricing-strategy signal, not a marketing win.

Block 3 — The Test Grid Verdicts. For every lever tested: hypothesis, result, decision. One line each. “RM120 threshold beat RM80 on incremental AOV; lock RM120 for 9.9.” The decision column is the whole point — a debrief that ends in observations instead of decisions is a diary.

Block 4 — The Ops Log. Peak-day orders, packing throughput, RTS SLA hit-rate, courier pickup failures, stockouts by SKU, CS ticket spike and its top three causes. This block feeds the capacity plan directly.

Block 5 — Money. Voucher subsidy, affiliate commissions paid, campaign-period margin versus normal-period margin, and the number almost nobody computes: contribution margin per campaign order versus business-as-usual. Some campaigns are worth running near zero margin for buyer acquisition — but that should be a decision made in the plan, not a surprise discovered at month-end close.

An automotive brand we work with ran this loop across their 2025 calendar, and the compounding was visible campaign-over-campaign: the mid-year debrief flagged a voucher threshold set below natural basket size and an underperforming livestream slot; both were corrected for 9.9, which converted measurably better on the same traffic; by 11.11 the mechanics were locked and the team’s entire energy went into inventory depth and execution rather than decisions. Same brand, same products, same platforms — the difference was that each campaign inherited the previous one’s lessons in writing. That’s the arc working: 11.11 isn’t where you figure things out. It’s where you cash in what July taught you.

This issue ships with a working Excel version — the H2 Campaign Planner — carrying the peak-day baseline grid, the debrief template with live variance formulas, the capacity calculator, and the budget pacer from the next section. Put your own 7.7 numbers into it on Thursday morning.

Framework 7: The H2 Pacing Curve — Why Front-Loading November Wastes Money

The instinct most teams bring to Q4 budgeting is to hoard budget for November and detonate it across the 11.11 window. It feels prudent — save the powder for the big battle. In auction terms, it is the most expensive possible way to spend the same money.

Recall the cost curve: November CPMs peaked at US$25.22 against a January trough of US$15.74, and Q4 rates across markets run 30–60% above baseline. Auction prices spike exactly when every brand in your category shows up with the same hoarded budget. A ringgit of prospecting spend in November buys roughly 60–70% of the impressions the same ringgit bought in July–September — aimed at consumers who are, that week, the most deal-saturated they will be all year.

The pacing logic:

• July (7.7) — spend to learn. Modest budget, deliberately spread across the test grid. CPMs are near their H2 low; this is the cheapest data you’ll buy all year.

• August — the quiet build. Evergreen retargeting pools, creator and UGC production at off-peak prices, and the actual implementation of your debrief fixes: PDP rebuilds, bundle restructures, FBS inbound planning, affiliate lock-ins.

• September (9.9) — refine at moderate cost. Re-run winners with tightened settings. The 9.9 debrief locks the Q4 playbook.

• October — buy your November audience before November. The counterintuitive move: your heaviest prospecting weeks should be early-to-mid October, before full CPM inflation. Fill the retargeting pools, run the lead mechanics, grow the store followings and Live audiences. You are pre-buying November’s conversions at October’s prices.

• November (11.11) — bottom-funnel only. Peak-window spend goes overwhelmingly to retargeting warm pools, brand search, and Live traffic — audiences you already paid to build. Prospecting at 11.11 auction prices is charity to the platform.

• December (12.12) — harvest and retain. Lowest paid intensity of the arc. CRM, WhatsApp and email flows against the buyer file the arc built — the customers acquired at 10.10 and 11.11 hit their first replenishment window right as 12.12 lands. (The retention machinery that makes this work is its own discipline.)

As an illustrative starting split across July–December: 10 / 10 / 20 / 30 / 20 / 10 — with October, not November, as the single heaviest month. Your curve should flex with your category and your baseline data, but the shape is the point: the spend peak sits one campaign before the revenue peak. Front-loading November is really late-loading — by the time the money moves, the cheap impressions are gone.

This is also the argument for treating your media plan as one H2 document rather than five campaign briefs — a single phased budget where each debrief can trigger a reallocation. When we build media plans for clients, the phasing sheet isn’t an appendix; it’s the control surface. The full cross-channel logic — how Meta and TikTok create the demand that platform search and Google brand terms capture — is in Thinking Full Funnel; the H2 arc is that funnel with a calendar strapped to it.

The Diagnostic: 7 Signs Your H2 Is Running on Folklore

1. Your 7.7 plan is last year’s settings “plus a growth percentage.”

2. Nobody can state, per channel, what peak-day CVR and AOV did in the last three campaigns.

3. Your channel leads and agencies built their campaign plans separately and reconciled them never.

4. You cannot name a single hypothesis this campaign is testing.

5. Your top 10 affiliates by H1 GMV haven’t heard from you about Q4 — but they’ll hear from your competitors in October.

6. Nobody has measured packing throughput; you’d have to guess orders per person-hour.

7. Your Q4 budget plan is a single number labelled “11.11.”

Three or more, and your organisation is running on folklore — capable people executing an uninstrumented calendar. The fix isn’t more effort in November. It’s cadence, starting Tuesday.

The Playbook: The Next 10 Days

Before peak day (today and Monday, 5–6 July):

• Start the backdated baseline grid — even a rough version covering the last two campaigns per channel. Assign one owner per channel, half a day each.

• Write the one-page forecast off the grid: sessions, CVR, AOV, GMV, ad spend, commissions, voucher subsidy — per channel. This single document converts Tuesday from an event into an experiment.

• Confirm the test grid is clean — one lever per variable. Where levers are tangled on the same SKUs, untangle what you can and note the contamination for the debrief.

• Put a stopwatch owner on the packing bench for Tuesday.

• Screenshot hero PDPs (images, titles, prices) so test conditions are documented, then freeze listing content until the window closes.

Peak day (Tuesday, 7 July):

• Execute. Capture hourly GMV if your dashboards allow — the intraday curve shows which Live slots and voucher drops actually moved the needle.

• Log ops failures in real time in a shared note. Memory lies by Thursday.

Debrief (by Thursday, 9 July):

• Run the five blocks. Ninety minutes, all channel leads and agencies in the room, decisions written down.

• Extract P, T, and your best-estimate M into the capacity math for 11.11. If the answer scares you — good. You have 15 weeks, which is exactly enough.

• Book what October will make scarce: FBS inbound slots, livestream hosts for the 11.11 window, 3PL overflow for the peak-72-hours.

By end of July:

• Convert the debrief’s decision column into the 8.8/9.9 brief.

• Lock your top 10–20 affiliates: campaign-window rates, Q4 samples, a named contact. Before the commission war, not during it.

• Open the H2 conversation with your platform account managers: co-fund voucher nominations, TikTok co-investment commitments, submission calendars — with your 7.7 performance story in hand.

• Draft the H2 pacing curve against your own numbers and lock the October prospecting budget — the line item your future self will thank you for.

The Close

Mei Lin’s 7.7 will be fine. Five channels’ worth of vouchers will redeem, the number will land somewhere near the stretch target, and the machinery will roll on to the next fire drill. That’s exactly the problem — “fine” is what an uninstrumented calendar produces, and uninstrumented July becomes folklore-driven November, priced at November auction rates.

The brands that win Q4 in this region don’t out-spend anyone. They out-sequence everyone: baseline before forecasting, test at 7.7 when mistakes are cheap, lock affiliates and platform partnerships before the market bids them up, refine at 9.9, scale at 11.11 with nothing left to figure out, harvest at 12.12 off the buyer file the arc built.

The rehearsal is Tuesday. Start the baseline grid today, put a stopwatch on the bench, and hold the debrief by Thursday. Q4 is decided in the next ten days — quietly, in a spreadsheet, while everyone else watches the countdown clock.

Related reading from the archive:

• The Sales Growth Formula (That Actually Runs Your P&L) — the visitors × CVR × AOV decomposition the baseline grid and test grid are built on (Sep 2025)

• Thinking Full Funnel — orchestrating Meta, TikTok Shop, and Shopee as one funnel (Jan 2026)

• The Ecommerce Cold Start, Part 2 — fulfilment build-vs-partner math and FBS economics (Feb 2026)

• The DTC Cold Start: Why Your Shopify Store Fails in Month 3 — the retention engine your 12.12 harvest depends on (Jun 2026)

• Building Beyond Marketplaces — standing up the owned channels that turn campaign buyers into a customer file (Oct 2025)

A reader wrote back with the obvious question. Fine. So what does the Indonesia playbook actually look like?

Fair. I spent a whole issue telling you what Indonesia is not the right template for. I owe you the other side of it.

So here’s a different brief, from a different client. This one runs a mid-sized local snack and beverage brand out of Kuala Lumpur. Good business — profitable in Malaysia, decent in Singapore, just cracked the Philippines. He’d been staring at the map and decided Indonesia was next. Of course it was. Ten times the population, the biggest ecommerce market in the region, right next door, similar language. “It’s basically a bigger Malaysia,” he told me. “We’ve already proven the model. We just scale it across the Strait.”

I asked him how he planned to handle the islands. Long pause. “What do you mean, the islands?” That’s the whole issue, right there. He was about to make the exact mistake I’d just warned the Jakarta brands about — only in reverse. He’d taken a playbook that works in a compact, urban, three-platform, English-tolerant peninsula and assumed it would scale into a 280- million-person archipelago of 17,000 islands, where the second-largest sales channel is a live video stream and half the growth is happening in cities he couldn’t name.

Indonesia isn’t a bigger Malaysia. It isn’t a bigger anything. It’s the market that makes its own rules — literally, it has banned business models that work everywhere else — and it punishes brands that arrive with a borrowed map.

This is what the Indonesian map actually looks like.

The Setup: the gravitational centre of Southeast Asian commerce

Let’s start with scale, because scale changes the physics of everything that follows.

Southeast Asia’s platform ecommerce hit US$157.6 billion in GMV in 2025, up +23% year-on-year and nearly triple what it was in 2020 (Momentum Works, Ecommerce in Southeast Asia 2026). Indonesia alone is roughly 44% of that — call it a US$65 billion-plus platform market sitting inside one country. It is not the biggest market in the region by a nose. It is the biggest by a landslide, several times the size of Malaysia and larger than the next two or three SEA markets combined.

Here’s what that market looked like through 2025:

• ~US$65B+ in platform GMV — about 44% of all of Southeast Asia (Momentum Works, 2026)

• TikTok Shop Indonesia drove US$13.1 billion in GMV, up +111% year-on-year — making Indonesia TikTok Shop’s second-largest market on earth, behind only the United States (Momentum Works, 2026)

• Shopee holds roughly 52% of the Indonesian market; the Shop | Tokopedia combine sits around 35%; Lazada has slipped to single-to-low-double digits (industry estimates, 2025)

• Video commerce reached ~20% of online GMV in 2025, up from under 5% in 2022, and 60% of online buyers now purchase through live sessions — an estimated 2.6 billion livestream and short-video orders in a single year (Mordor Intelligence)

• Tier-2 and tier-3 cities are on track to generate half of national digital GMV, up from around 40% in 2021 (Mordor Intelligence)

None of that says “scale up your Malaysia plan.” It says you are walking into the single largest, most concentrated, most livestream-native, most geographically dispersed ecommerce market in the hemisphere — one that has already deported a US$1.5 billion business model for breaking its rules.

Because here is the thing that makes Indonesia genuinely different, and not just bigger: in late 2023, Indonesia banned social media platforms from processing sales. TikTok Shop — at the time the fastest-growing channel in the country — went dark overnight. It only came back by buying 75% of Tokopedia for US$1.5 billion and folding its commerce operation inside an Indonesian company. No other market in Southeast Asia has done anything like that.

So when I say Indonesia is its own beast, I mean it structurally. It has its own scale, its own rules, its own geography, its own dominant channel, and its own winners. Get those five forces right and it’s the prize of the region. Get them wrong — arrive with the bigger-Malaysia map — and you’ll spend a fortune learning each one the hard way.

The Five Forces of the Archipelago

This is the framework. Five forces that make Indonesia structurally unlike Malaysia, Singapore, or anywhere else you’ve won — and what each one changes about how you operate.

Force 1: Scale changes the physics — distribution beats brute force

Most operators read “biggest market in the region” as “biggest version of what I already do.” Spend more, run more campaigns, scale the engine. That instinct is exactly wrong, and Indonesia’s scale is the reason.

In a compact market like Malaysia or Singapore, you can win with concentrated spend and tight operation — the country is small enough that brute-force marketing reaches most of it. Indonesia is the opposite shape. It’s a 280-million-person country spread across 17,000 islands and dozens of distinct regional markets. There is no single “Indonesian consumer” to brute-force. Jakarta is not Indonesia any more than KL is Malaysia — except the gap between the capital and the rest is far wider, and the rest is far bigger.

The lever that wins Indonesia isn’t spend intensity. It’s distribution depth — how many of those regional markets you can actually reach, stock, ship to, and convert. A brand that pours its whole budget into a slick Jakarta-and-Java campaign is fishing in maybe half the pond. The growth — the part that’s compounding fastest — is in the other half, in cities most regional planners can’t place on a map.

The implication: in Indonesia, your constraint is reach, not budget. Plan around how far your distribution and content travel across the archipelago, not how loud you can be in the capital.

Force 2: The regulator sets the board — and it’s playing for the home team

Indonesia is the only Southeast Asian market where the government has actively redrawn the commerce map, and it did so to protect domestic sellers. You cannot operate here without understanding the rules of the board, because they are not the same rules as anywhere else.

The headline event: Trade Regulation No. 31 of 2023, effective 26 September 2023. Two pieces matter most.

First, it banned social commerce platforms from processing transactions — they can host content and promotion, but not the checkout. This is what killed TikTok Shop overnight in late 2023 and forced the US$1.5 billion Tokopedia acquisition to bring it back inside an Indonesian entity. The integration concluded in April 2025. The lesson isn’t trivia — it’s that the channel structure you’re planning around in Indonesia exists because of a regulatory workaround, and it could move again.

Second, it set a US$100 minimum price floor on imported goods sold directly cross-border (with a permitted-goods exception list). Translation: you cannot dropship cheap imported product into Indonesia the way you might cross-border into other SEA markets. Jakarta has also repeatedly slow-walked or blocked the ultra-cheap cross-border players that have flooded other markets. The explicit, stated objective is to protect domestic MSMEs — local sellers — from foreign competition.

Read those two moves together and the strategy writes itself: Indonesia rewards brands that localise and punishes brands that import. Local stock, local entity, local sellers, local content. The regulator has put its thumb firmly on the home team’s side of the scale.

The implication: plan for a local operating footprint — local fulfilment, a compliant selling structure, and a content engine that runs inside the platforms rather than around them. Cross-border-only, import-light strategies that work elsewhere hit a regulatory wall here.

Force 3: Geography is the moat — the fulfilment network, the tier-2/3 surge and the payment methods

This is the force my snack-brand Ecommerce client had never thought about, and it’s the one that separates the brands that scale in Indonesia from the ones that stall at the Jakarta city limits.

Indonesia’s growth is decentralising fast. Tier-2 and tier-3 cities are on track to produce half of national digital GMV, up from about 40% in 2021 (Mordor Intelligence). Jakarta, Surabaya, and Bandung still lead, but the momentum is in Medan, Makassar, Yogyakarta, and a long tail of provincial capitals where households are coming online and getting banked for the first time. That’s where the next decade of growth lives — and it’s the part a Jakarta-office plan systematically under-serves.

Serving it is a logistics problem before it’s a marketing problem. The archipelago is the hardest last-mile geography in Southeast Asia, and the brands winning it are the ones plugged into the dense local courier network — J&T, JNE, SiCepat, Ninja Xpress, Pos Indonesia — and the micro-fulfilment hubs pushing stock closer to provincial consumers. That hub network has been cutting shipping costs 12–18% and unlocking next-day delivery beyond Java, helped along by roughly IDR 400 trillion (US$25.8 billion) in government infrastructure spend (Mordor Intelligence).

And then there’s how those consumers pay. Indonesia is not a card market. The payment mix runs digital wallets (~35% — GoPay, OVO, DANA, ShopeePay), bank transfer (~26%), then a long tail including BNPL and cash-on-delivery at ~8% (industry data, 2025). COD share is falling as QRIS and wallets spread — QRIS transaction volume jumped 175% year-on-year in 2024 — but in the tier-2/3 and rural markets that are driving growth, COD is still how a meaningful slice of first-time online buyers transact. And COD carries a tax: it correlates with materially higher return and refusal rates, which lands straight on your unit economics.

The implication: Indonesia is won at the last mile, not the campaign. Pick your logistics partners deliberately, push stock into regional hubs, plan for COD’s return-rate drag in your margins, and treat tier-2/3 reach as a primary growth strategy — not an afterthought once Java is “done.”

Force 4: Live is the default channel, not a tactic

In most markets, livestream is a thing you add. In Indonesia, livestream is increasingly the thing itself.

Video commerce reached around 20% of all online GMV in 2025 — up from under 5% three years earlier — and 60% of online buyers now say they purchase through live sessions, totalling an estimated 2.6 billion livestream and short-video orders in a year (Mordor Intelligence). Indonesia is the most live-commerce-native large market on the planet. The TikTok-Tokopedia machine is built around exactly this: short-video discovery and live checkout fused into one funnel, now embedded into Tokopedia’s catalogue post-integration.

This is a different operating model, not a content add-on. Winning it means always-on live — not a campaign-day broadcast, but daily or near-daily streaming, often for hours, run by hosts who know how to demo, bundle, handle comments, and close in real time. It means an affiliate and creator army at the nano and micro tier, sized for a country of 280 million, not the tight creator roster you’d run in Malaysia. It means production capacity — studios, hosts, scripts, offers — as a core operational function, not something you outsource per campaign.

A brand that ports its Malaysia content plan into Indonesia — polished static creative, a handful of mid-tier creators, the occasional live event — is bringing a campaign to a channel war. The Indonesian buyer is sitting in a live session waiting to be sold to. If you’re not in the room, your local competitor is.

The implication: build live commerce as core infrastructure — daily streaming, a deep creator/affiliate base, in-house or dedicated production. In Indonesia, the live stream isn’t a campaign tactic. It’s the storefront.

Force 5: Local wins — domestic brands, halal trust, and the price floor

The last force is the one imported brands find hardest to accept: in Indonesia, the local champions are not the warm-up act. They’re the headliners.

Look at beauty, the most active and most-watched category on TikTok Shop Indonesia (beauty and personal care is around 14% of online retail). The brands winning it are largely Indonesian — Skintific, Wardah, Somethinc — not imported Western or Korean labels. Skintific in particular has run the exact playbook this whole issue is describing: aggressive pricing (hero products at US$8–15 where international equivalents sit at US$25–35), a creator-led discovery engine, and live commerce as the spine. Local brands win because they understand the price psychology, the cultural register, and the trust signals — and because the regulator, as we saw, is actively tilting the board toward them.

Two trust signals matter more in Indonesia than almost anywhere else in the region. Halal certification is close to non-negotiable in food, beverage, and cosmetics — it’s a baseline trust factor, not a niche feature. And price runs to a floor: the Indonesian value-tier consumer anchors low, and a price ladder built for Malaysian or Singaporean wallets will read as expensive and foreign. You’re often competing against a domestic brand that has engineered its cost base specifically for that floor.

This doesn’t mean an outside brand can’t win. It means an outside brand has to stop behaving like an import. Local manufacturing or co-packing, halal certification, locally engineered price points, a local creator base, content made by people who are visibly from there. The brands that treat Indonesia as a place to sell into lose to the brands that treat it as a place to be from.

The implication: localise to the point of being indistinguishable from a domestic brand where it counts — halal, price floor, local content, local stock. In Indonesia, “imported” is not a premium signal. It’s a handicap.

The Diagnostic: are you running Indonesia like a bigger Malaysia?

Six signs your Indonesia plan is a scaled-up peninsula plan. Be honest.

🔴 Sign 1: Your Indonesia plan is your Malaysia plan with a bigger budget. Same structure, same channel mix, same content model — just more spend. You didn’t build an Indonesia plan. You inflated one.

🔴 Sign 2: You can’t name your tier-2/3 city strategy. If your growth plan is “win Jakarta and Java, then see,” you’re planning around half the market and ignoring the faster-growing half.

🔴 Sign 3: Livestream is a line item, not the spine. A few creators and a campaign-day broadcast, in a market where 60% of buyers purchase on live and video is 20% of GMV. You brought a campaign to a channel war.

🔴 Sign 4: You’re planning cross-border, import-light. If the plan leans on dropshipping cheap imported stock without a local footprint, you’re on the wrong side of the US$100 floor and the regulator’s MSME agenda.

🔴 Sign 5: No halal, no local price floor, no local content. You’re bringing a Malaysian or Western price ladder and English-ish creative into a market where the local champions have engineered for the floor and certified for trust.

🔴 Sign 6: You’re running it from a Jakarta (or KL) desk with no last-mile plan. No deliberate courier strategy, no regional hubs, no plan for COD return rates. The hardest last-mile geography in SEA, run as an afterthought.

Three or more? You’re running Indonesia on the bigger-Malaysia map. The fix isn’t a bigger budget. It’s a plan built from the Indonesian seat — archipelago-first, live-first, local-first.

Two From the Field

Both anonymised — composites drawn from work in and around the Indonesian market, not named clients.

The imported beauty brand that stopped being an import

A regional beauty brand — real product, genuine following in its home market — pushed into Indonesia the way most do: shipped stock cross-border, ran its existing creative with Bahasa subtitles, lined up a handful of mid-tier creators, and waited. It didn’t fail outright. It just sat there, quietly outsold by local brands it had never heard of, on a channel (live) it was barely using, at a price point that read as expensive next to the domestic shelf.

The turnaround was structural, not cosmetic. The brand moved to local co-packing and got its hero range halal-certified — clearing the trust bar. It re-engineered the price ladder down to the Indonesian floor instead of converting its home-market prices across. And it rebuilt the content engine around always-on live commerce and a broad nano/micro creator base sized for the market — not a tight roster, an army. Same products, same brand. A completely different operating model — built to look domestic, not imported.

The lesson for the reader: in Indonesia, the imported-brand instinct — ship in, translate, sprinkle creators — is the losing move. The winning move is to become functionally local: local stock, local certification, local price, local live.

The Java-only brand that found the other half of the market

A brand doing well across Jakarta, Surabaya, and Bandung had quietly concluded that was the Indonesian market. Growth had flattened, and the read was “we’ve saturated.” They hadn’t saturated. They’d hit the edge of the geography they’d bothered to serve.

The unlock was distribution and last-mile, not marketing. They plugged into the regional courier and micro-hub network to push stock closer to provincial consumers, brought delivery times and costs down beyond Java, and built for COD properly — pricing in the higher return rates instead of being surprised by them. Then they briefed regional creators who actually spoke to tier-2/3 audiences, rather than recycling Jakarta content nationally. The “saturated” brand found a second growth engine in the cities it had been treating as rounding errors.

The lesson: in Indonesia, flat growth often isn’t saturation — it’s the edge of your distribution. The next half of the market is real, it’s growing faster than the first half, and it’s a logistics problem before it’s a demand problem.

The Market-Entry Checklist

If you’re planning Indonesia — or fixing an Indonesia plan that’s quietly stalling — here’s the Monday-morning version.

1. Plan for distribution depth, not spend intensity. Map the regional markets you can actually reach and stock, and treat that reach as your primary growth constraint. Indonesia is won by how far you travel, not how loud you are in Jakarta.

2. Build a local operating footprint. Local fulfilment, a compliant selling structure, local stock. Assume the regulator favours the home team — because Trade Regulation 31/2023 explicitly does — and get on the right side of the US$100 import floor.

3. Win the last mile. Choose your couriers deliberately — J&T, JNE, SiCepat, Ninja, Pos — push inventory into regional micro-hubs, and price COD’s return-rate drag into your margins from day one.

4. Make tier-2/3 a primary strategy. Half the GMV and the faster-growing half. Brief regional creators, tailor content beyond Jakarta, and serve the provincial cities as a core market — not a phase two.

5. Build live commerce as infrastructure. Always-on streaming, a deep nano/micro creator and affiliate base sized for 280 million people, dedicated production capacity. In Indonesia the live stream is the storefront, not a campaign.

6. Localise until you stop looking imported. Halal certification where it counts, a price ladder engineered to the Indonesian floor, content made by people visibly from there, local manufacturing or co-packing where the economics allow. Be from there, don’t sell into there.

The Close

My snack-brand client went quiet when I asked about the islands — and that pause was the most useful thing that happened in the whole conversation. It was the moment “it’s basically a bigger Malaysia” fell apart in his head.

Indonesia is the biggest prize in Southeast Asian ecommerce, and it is the one market that most punishes the borrowed map. It has its own scale that rewards distribution over spend, its own rules written to favour the home team, its own geography that turns logistics into the moat, its own channel where the live stream is the storefront, and its own winners who are local by design. Scale your Malaysia plan into it and you’ll grow — Indonesia is growing for almost everyone — but you’ll grow slower than the market while local operators and better-prepared regionals take the cities you never reached.

Two weeks ago I told you Malaysia deserves its own playbook because it’s not a smaller Indonesia.

Indonesia deserves its own playbook for the opposite reason. It’s not a bigger anything. It’s its own beast — and the brands that win it are the ones that build for the archipelago, not the ones that arrive with a map drawn for somewhere else.

Build the Indonesian plan from the Indonesian seat. The biggest market in the region is worth its own.

Related reading from the archive:

• Malaysia is not Indonesia — the companion to this issue: why the peninsula needs its own playbook (Jun 2026)

• Thinking Full Funnel — connecting discovery to conversion across platforms, the discipline live commerce demands (Jan 2026)

• Cross-Border Without Borders — using Malaysia as a launchpad into the rest of SEA, including the regulatory walls (Mar 2026)

• How Shopee Malaysia’s 2026 Fee Changes Will Reshape Your Margins — the platform-economics layer beneath all of this (Jan 2026)

Thanks for reading Ecommerce Owner’s Manual — your weekly strategic guide to winning in Southeast Asian eCommerce. Free for the first 1,000 subscribers. Share it with someone planning their Indonesia entry.

That was the AI chat assistant we’d just switched on for one of our brands, having burned through its monthly credit allocation. It wasn’t a disaster — it was a signal. Within weeks of going live, the thing was doing so much work that it ran out of fuel. For an enabler drowning in customer chats across dozens of marketplace storefronts, “the AI ran out of credits because it was busy” is a champagne problem.

But getting there was not a plug-and-play story. Anyone who tells you that you can switch on an “ecommerce chatbot,” walk away, and watch your numbers improve is selling you something. What actually happened is that we spent weeks mapping SLAs, rebuilding our entire product knowledge base, writing escalation rules scenario by scenario, and learning — sometimes the hard way — how each marketplace silently grades you.

This issue is the executional version of that journey.
If you operate a brand on Shopee, Lazada, or TikTok Shop in Southeast Asia and you’re staring down the same chat volume we were, this is the playbook. I’m going to show you exactly how we did it for one client — a battery brand we manage across Shopee and Lazada in multiple SEA markets — including the real before-and-after numbers, and the parts most case studies leave out.

Part 1 — Every marketplace and country grades you differently

When people think about customer chat, they think about being helpful. That’s the romantic version. The operational reality is that on a marketplace, chat is a scored compliance metric that quietly controls your visibility and your seller privileges.

Here’s what makes it genuinely hard at scale:

Each platform measures a different thing. Shopee’s headline metric is Chat Response Rate (CRR) — the percentage of conversations you respond to within a 12-hour window. Shopee recommends keeping it at 90%+, and you need at least 85% to qualify as a Preferred Seller. Let it collapse and seller privileges start getting suspended. Lazada and TikTok Shop each have their own response-time and service expectations. So “good customer service” isn’t one target — it’s three or four different scorecards, each with its own clock.

You can have Shopee’s native assistant cheerfully answering customers all day, and your CRR can still be sitting at a number that gets you penalised, because the platform doesn’t credit its own automated replies as you responding. If you’re going to automate, the automation has to happen in a layer the platform recognises as you — not the marketplace’s own bot.

Multiple markets means multiple languages and rulebooks. The same brand on Shopee Philippines and Lazada Malaysia is effectively two different support operations. Customers in the Philippines message in Taglish (”Paki-ship na po,” “Legit ba?“). Malaysian customers ask about LazPayLater and whether you ship to Sabah and Sarawak. The refund mechanics differ (Shopee’s flow vs. LazMall’s 15-day window), the courier rules differ, and even warranty answers differ by product line. A single generic script is wrong in at least three places the moment it crosses a border.

Cross-platform fragmentation. Every storefront is its own login, its own notification, its own place to miss a message. When you’re managing many accounts, “just check the chats” stops being a real instruction. There’s no single screen, so something always slips — and on Shopee, a slipped message inside that 12-hour window is a direct hit to the metric that governs your account.

That’s the box we were in: high volume, multiple platforms, multiple markets, multiple languages, and a scoring system where the platform’s own shortcuts don’t count. The decision wasn’t whether to bring in tooling. It was how to do it without making account health worse.

Part 2 — The decision: integrate first, automate second

The instinct most operators have is to lead with the AI. We deliberately did the opposite. We led with integration, and treated AI as something we earned the right to switch on only after the foundations were in place.

The first move was consolidating every storefront’s chat into one unified inbox — pulling Shopee, Lazada, and TikTok Shop conversations across all the accounts we run into a single workspace. Before any automation, this alone changed the game: the team stopped switching between apps and logins, and for the first time we had one view of every incoming conversation and which ones were ageing toward an SLA breach.

Only then did we layer AI on top. The reason for the sequencing matters: an AI agent is only as good as the knowledge and the rules you give it, and the inbox is where you watch whether it’s actually working. Switch on automation before you can see and audit it, and you’re flying blind on the exact metric you’re trying to protect.

For the tool itself — we tested an AI chat platform built specifically for Southeast Asian marketplaces. I’m deliberately not turning this into a tool review, because the tool is the least transferable part of the story. The platform you pick matters far less than the operating system you build around it. Everything that follows works regardless of which vendor you land on.

Part 3 — The playbook: how to actually implement AI chat support

This is the part I wish someone had handed me at the start. Eight steps, in the order we did them.

Step 1 — Map the SLA you’re actually being graded on, per platform

Before touching any AI, write down — for each platform and each market — the exact metric, the exact window, and the exact threshold. For us that meant Shopee’s CRR at a 12-hour window with a 90% target, plus each platform’s response-time expectations. You cannot automate toward a target you haven’t defined. This document becomes the spec the entire system is built to satisfy.

Step 2 — Turn OFF the platform’s native AI assistant

Counter-intuitive, but essential. On Shopee we switched the Shopee AI Assistant off, on purpose. Because the platform doesn’t count its own assistant’s replies toward your Chat Response Rate, leaving it on actively works against you — it answers the customer (so you think you’re covered) while your CRR quietly stays exposed. Routing every conversation through the integrated tool instead means the responses are logged as you responding, and the metric finally moves. This single setting is the difference between automation that helps your account health and automation that silently sabotages it.

Step 3 — Build the product knowledge base in a structured template

The AI is only as smart as your knowledge base. We rebuilt every brand’s FAQs into a strict, repeatable template — Product Name → Details → FAQs — so the AI retrieves clean, unambiguous facts. This is where the unglamorous domain knowledge lives: the rule that certain chargers need a minimum of two batteries to function, which product lines carry a 12-month warranty (for this brand, only the rechargeable series) and which don’t, expiry and freshness handling, authenticity guarantees. Reformatting messy FAQs into this structure was genuinely the most labour-intensive part of the whole project — and the highest-leverage.

Step 4 — Design scenarios, not scripts

A chatbot reads a script. A useful AI agent runs scenarios, each with four parts: a trigger (what the customer is really asking, including the messy real phrasings), an action/response logic, a negative constraint (when not to fire this scenario), and an escalation rule. We built roughly twenty of these per storefront — order status and expedite requests, failed deliveries, missing items, damaged goods, wholesale and discount requests, freshness/FIFO assurance, authenticity checks, address and courier changes, invoice requests, and more.

The negative constraints are what separate a system from a toy. “Customer wants to rush their order” and “customer wants to cancel because shipping is slow” use almost the same words but require opposite responses — one is reassurance with no escalation, the other is a careful retention/cancellation path. Spelling out do NOT trigger this if… is how you stop the AI from confidently doing the wrong thing.

Step 5 — Localise per market, not per brand

Each market got its own knowledge source and scenario set. The Philippines scenarios handle Taglish triggers and Shopee PH’s refund flow. The Malaysia knowledge source covers LazPayLater, the LazMall 15-day return window, the “we do not ship to East Malaysia” rule, courier-specific answers, and market-specific warranty terms. Same brand, same products — three different support brains. Trying to serve all markets from one generic configuration is the fastest way to give confidently wrong answers.

Step 6 — Define the human–AI handover precisely

The goal is never 100% automation — it’s letting the AI clear the routine load so your humans handle what actually needs a human. We wrote explicit escalation triggers: damaged or defective items, wrong items, invoice/tax requests, excess or missing items, system-cancelled orders, and edge cases like a customer insisting on the exact printed expiry date. Everything else, the AI resolves end to end.

Two rules made the handover clean: humans only reply to chats that are assigned or escalated to them, and agents do not reply if the AI is already handling it successfully — that “double-reply” is how you confuse customers and waste your team’s time. When a case is beyond the AI, it routes to a live agent rather than guessing.

Step 7 — Automate after-hours with a flow builder, and set honest operating hours

We set business hours (9:00am–5:30pm) and used a flow builder so that any message arriving outside those hours gets an immediate, useful automated response instead of sitting unanswered overnight — which on Shopee is exactly how a 12-hour window gets breached while everyone’s asleep. On public holidays we’d switch the AI on to cover the gap and have a team member log in once in the afternoon and once in the evening just to confirm nothing was stuck or wrongly routed. Automation with a human spot-check, not automation and hope.

Step 8 — Run a weekly knowledge-gap loop

This is what turns a launch into a system. Chat logs get reviewed regularly for two things: is the AI holding the brand’s voice (natural and friendly, not robotic), and where did it stumble? Every stumble is a knowledge gap, and every knowledge gap becomes an FAQ update. The automation rate climbs week over week not because the model gets smarter, but because your knowledge base does. Skip this loop and your AI quietly degrades as products, promos, and policies change underneath it.

Part 4 — The results: what happened to account health

We ran a clean before-and-after on the battery brand: three weeks before the AI agent went live, three weeks after. Same storefront, same products. Here’s what moved (figures in MYR, Shopee Malaysia):

The headline is the 60% cut in response time on the metric the platform actually watches. Faster, more consistent responses inside the SLA window is precisely what protects Chat Response Rate and, by extension, your seller standing and visibility. Across the wider portfolio of brands we run on this setup, the AI now fully resolves the large majority of conversations without a human ever touching them — which is what frees the team to spend its hours on the escalations that genuinely need judgement.

A note on honesty, because this newsletter is an owner’s manual and not a pitch deck: the platform and the tool measure response time differently, so those two rows aren’t directly comparable — read each against itself. The sales and order lift happened over the same window that automation went live, but visitor traffic also rose in that period, so I’d treat chat automation as a strong contributor to the commercial lift, not the sole cause. And CSAT on a single storefront over three weeks is too small a sample to lean on. The response-time and SLA improvements, though, are direct and unambiguous — that’s the line the automation owns outright.

Part 5 — What we’d tell our past selves

The knowledge base is the product; the AI is just the interface. Eighty percent of the value came from the unglamorous work of structuring FAQs and writing scenarios. Budget your time accordingly.

Watch the platform’s own settings, not just your tool. The single highest-impact move — turning off Shopee’s native assistant so responses count — was a platform setting, not a feature of the AI. Know how each marketplace scores you before you automate.

Don’t chase 100% automation. The win is a clean split: AI clears the repetitive volume, humans own the escalations. Design the handover as carefully as you design the automation.

Localise or lie. One generic configuration across markets will confidently tell a Sabah customer you’ll ship to them. Per-market knowledge isn’t a nice-to-have; it’s correctness.

It’s a loop, not a launch. The weekly knowledge-gap review is the difference between an AI that improves and one that rots.

Part 5 — What we’d tell our past selves

The knowledge base is the product; the AI is just the interface. Eighty percent of the value came from the unglamorous work of structuring FAQs and writing scenarios. Budget your time accordingly.

Watch the platform’s own settings, not just your tool. The single highest-impact move — turning off Shopee’s native assistant so responses count — was a platform setting, not a feature of the AI. Know how each marketplace scores you before you automate.

Don’t chase 100% automation. The win is a clean split: AI clears the repetitive volume, humans own the escalations. Design the handover as carefully as you design the automation.

Localise or lie. One generic configuration across markets will confidently tell a Sabah customer you’ll ship to them. Per-market knowledge isn’t a nice-to-have; it’s correctness.

It’s a loop, not a launch. The weekly knowledge-gap review is the difference between an AI that improves and one that rots.

Your 8-step implementation checklist

1. Document the SLA for every platform and market — metric, window, threshold.

2. Turn off the marketplace’s native AI assistant so your tool’s replies actually count.

3. Rebuild your knowledge base in a structured Product → Details → FAQs template.

4. Write scenarios, not scripts — trigger, action, negative constraint, escalation.

5. Localise knowledge and scenarios per market and language.

6. Define the human–AI handover with explicit escalation triggers and a no-double-reply rule.

7. Cover after-hours with a flow builder and set honest operating hours, with a human spot-check.

8. Run a weekly knowledge-gap loop to feed every AI stumble back into the FAQs.

If you’re an operator trying to get AI chat live across Shopee, Lazada, or TikTok Shop without putting your account health at risk, this is exactly the kind of build we run for the brands we manage. Reply to this email or reach me at stan@commerceplus.asia — happy to walk through what this would look like for your storefronts.

— Stan

I asked her one question: what’s the plan?

She sent me the deck. It was a good deck. Clean unit economics, a TikTok-first content engine, a clear creator tiering model, a festival calendar built around Harbolnas and 12.12. The whole thing was built — carefully, intelligently — for Indonesia. And then it had been copy-pasted into a tab marked “MY,” with the currency changed from IDR to RM and the population number swapped out.

That was the strategy. Same playbook, smaller country, change the money.

She is not alone. I hear this constantly. Malaysia is the market regional brands think they understand and consistently get wrong — because it looks familiar enough that nobody bothers to build a separate playbook for it. It’s the market that gets the ID strategy with the serial numbers filed off.

And here’s the part that should bother every regional planner reading this: Malaysia grew +47.6% in 2025. It was the second-fastest-growing ecommerce market in the entire region, behind only Thailand. TikTok Shop Malaysia grew +132%, crossing US$5 billion in GMV.
(Source: Momentum Works, Ecommerce in Southeast Asia 2026 report.)

The fastest-growing market is being run on a borrowed playbook.

This is the playbook it should be running on instead.

The Setup: Why “small and developed” is the wrong mental model

Regional brands treat Malaysia as a smaller, tidier Indonesia — same dynamics, fewer zeros. So they prioritise it accordingly: Indonesia gets the bespoke plan and the senior operator, Malaysia gets the spillover and a junior account manager who runs the ID model on autopilot.

The numbers say that’s backwards. Here’s what the Malaysian market actually looked like through 2025 and into 2026:

• +47.6% GMV growth in 2025 — second only to Thailand’s +51.8% across SEA-6 (Momentum Works, 2026)

• Total Malaysian ecommerce market around US$10.6 billion in 2025, projected to clear US$12.2 billion in 2026 (Mordor Intelligence)

• TikTok Shop MY +132% to over US$5 billion — the platform’s fastest-growth market in the region alongside Indonesia and Thailand

• Nearly two-thirds of Malaysians now shop online; TikTok Shop’s share of online shoppers jumped from 35% in 2023 to 52% in 2025 (Ipsos, November 2025)

• Livestream now drives up to 22% of TikTok Shop GMV in Malaysia — content converting three-to-five times faster than static ads (Mordor Intelligence)

None of that says “secondary market.” It says one of the highest-velocity ecommerce environments on earth, sitting inside a country small enough that you can actually win it with focused operation rather than brute-force spend.

So why does the borrowed playbook fail? Because Malaysia differs from Indonesia — and from the rest of SEA — on four specific axes that the ID model gets structurally wrong: platform balance, price psychology, language, and category economics.

Get those four right and Malaysia stops being your underperforming tab. Get them wrong and you leave the fastest-growing market in the region growing slower than it should — for you, while it accelerates for everyone else.

The Four Differences: A Country-Specific Playbook

This is the framework. Four axes where Malaysia is genuinely not Indonesia — and what each one changes about how you operate.

Difference 1: Malaysia is genuinely tri-platform. Indonesia is not.

This is the single biggest planning error I see, and it cascades into everything else.

In Indonesia, the practical reality is a two-horse race tilting hard toward TikTok Shop and Tokopedia under one ByteDance roof, with Shopee defending. Your content engine, your creator strategy, your budget — they all bend toward the TikTok-Tokopedia gravity well. That’s a rational ID plan.

Malaysia doesn’t work that way. Malaysia is the most genuinely balanced three-platform market in Southeast Asia.

Shopee remains the transactional anchor — seven in ten Malaysians have bought on it in the last six months, and its search-led, high-intent demand is non-negotiable for consistent volume. But TikTok Shop has closed to 52% of online shoppers, fuelled by short-video and live discovery, and it grew +132% in 2025. And Lazada — written off in much of the region — holds a real, defensible niche in Malaysia through LazMall as the trust-and-premium environment for established brands.

The roles are distinct, and you need all three:

• Shopee — bottom-of-funnel. Search demand, campaign volume, the reliable transaction engine. Your floor.

• TikTok Shop — top-of-funnel made shoppable. Discovery, livestream, impulse, new-product launch. Your growth.

• Lazada / LazMall — brand-equity protection. The premium, trust-focused storefront that shields your pricing and your story from the marketplace knife-fight.

A brand that runs Malaysia as “TikTok-first, Shopee-backup, ignore Lazada” — the ID-shaped plan — is leaving the transactional floor under-tended and the brand-equity ceiling unbuilt. You don’t pick a platform in Malaysia. You orchestrate three. We’ve written before about thinking full-funnel across channels rather than per-platform — Malaysia is the market where that discipline pays the most.

The implication: budget and staffing should reflect a three-platform split, not an ID-style two-platform tilt. If your MY plan has Lazada at a rounding error and Shopee on autopilot, you’ve imported the wrong country’s map.

Difference 2: Price psychology runs in RM, and the festival calendar is its own animal

Currency is not the only thing that changes when you cross the Strait of Malacca. The psychology of price changes.

Malaysian shoppers anchor in Ringgit, and Ringgit anchoring is tighter and more legible than Rupiah anchoring. The mental math is different. An RM30–RM100 beauty bundle reads as a clean, considered price point — and that band is exactly where TikTok Shop’s best-converting Malaysian beauty offers sit. Convert your IDR price ladder straight across and you’ll land on amounts that feel off to a Malaysian buyer — odd, untidy numbers that don’t match how they think about value.

Free-shipping thresholds matter more here than the regional model assumes. The Malaysian shopper is conditioned to the platform free-shipping minimum, and pricing your hero SKU just below the threshold — forcing a second item into the basket to unlock free shipping — is a basket-size lever that’s specific to how MY platforms structure their subsidies.

Then there’s the calendar. The ID plan is built around Harbolnas and the double-digit dates. Malaysia shares the 6.6, 9.9, 11.11, 12.12 mega-sales — but the peak that the borrowed playbook completely misses is Raya.

Hari Raya isn’t a sale. It’s the commercial centre of gravity for the Malaysian year — a weeks-long ramp of gifting, fashion, home, beauty, and food that reshapes demand the way nothing on the ID calendar quite does. Brands that treat Raya as “a Q1/Q2 promo” instead of the tentpole — that don’t build inventory, content, and creator activations months ahead of it — are planning Malaysia on the wrong calendar. The festival calendar isn’t a localisation detail. It’s the spine of the MY demand curve.

The implication: rebuild the price ladder natively in RM around clean anchor points and the free-shipping threshold, and rebuild the promo calendar around Raya as the anchor tentpole — not as a translated Harbolnas.

Difference 3: Bilingual BM/EN content that actually travels

Here’s where “everyone speaks English” quietly costs regional brands the most.

Yes, English works in Malaysia. That’s exactly why brands get lazy — they run English creative, it doesn’t fail, and they conclude the language question is solved. It isn’t. English-only content underperforms its bilingual ceiling, and the gap is real: mid-tier local creators producing bilingual Malay/English content have been shown to lift engagement by around 32% versus single-language content (industry data via Accio, 2025).

And the language mix shifts by region and platform. TikTok’s pull is strongest among Malays in the East Coast and northern states — where roughly 60% report recent TikTok Shop purchases — markets where BM-forward content isn’t a nice-to-have, it’s the price of entry. The KL/urban Shopee shopper tolerates more English. One national content strategy papers over both.

The implication: content is not a translation line item. Brief local creators to make natively bilingual content from scratch, calibrate the BM/EN mix to platform and region, and stop measuring “we localised it” as a box ticked when the English script merely got a Malay subtitle.

Difference 4: Category economics — beauty and fashion are not electronics

The last axis is the one finance gets wrong: the assumption that platform economics are uniform across categories. They aren’t, and Malaysia’s category mix has its own shape.

Beauty and personal care is the leading GMV category on TikTok Shop Malaysia, and it behaves nothing like electronics. Beauty carries the margin to absorb platform take rates, ad spend, and creator commissions — it’s discovery-led, bundle-friendly, repeat-purchase, and it lives or dies on content. The RM30–RM100 bundle, the livestream demo, the affiliate-driven impulse buy — that’s a beauty playbook, and it works because the gross margin gives you room.

Fashion and accessories share the discovery-led, content-hungry shape but with their own returns and sizing dynamics. F&B converts hard on livestream — Malaysians buy snacks and packaged food in volume off live sessions — but on thinner margins that demand tighter unit economics.

Electronics is the trap. Lower margin, higher consideration, warranty-and-trust-driven — which is precisely why electronics over-indexes on Shopee’s search-and-reliability environment and under-performs the TikTok livestream impulse model. Run the beauty content engine against an electronics margin profile and you’ll burn the contribution margin to zero chasing GMV that doesn’t pay.

The implication: your platform mix, your content intensity, and your discount tolerance should be set by category margin profile — not applied as one uniform “MY plan” across a portfolio. A beauty SKU and an electronics SKU need different platforms, different content budgets, and different promo depth, even inside the same company.

The Diagnostic: Are You Running Malaysia on Indonesia’s Map?

Six signs your MY plan is a borrowed plan. Be honest.

🔴 Sign 1: Your MY deck is your ID deck with the currency changed. If the structure, the platform priorities, and the calendar are identical and only the numbers moved, you didn’t build a Malaysia plan. You renamed one.

🔴 Sign 2: Lazada is a rounding error in your MY budget. In Indonesia, fine. In Malaysia, you’re skipping the brand-equity-protection layer that LazMall actually provides here.

🔴 Sign 3: Your festival calendar leads with Harbolnas and treats Raya as a minor promo. The single clearest tell that the calendar was built for the wrong country.

🔴 Sign 4: Your “localisation” is English creative with Malay subtitles. You ticked the box without doing the work. The 32% engagement lift lives in natively bilingual content, not translated content.

🔴 Sign 5: You run one content intensity across beauty, fashion, and electronics. Uniform content budget across category margin profiles means you’re overspending on the low-margin SKUs and underspending on the ones that can carry it.

🔴 Sign 6: Your senior operator is on the ID account and MY gets the junior. You staffed Malaysia as the spillover market. The +47.6% growth market is being run by your least experienced person.

Three or more? You’re running Malaysia on Indonesia’s map. The fix isn’t more budget. It’s a plan built from the Malaysian seat.

Two From the Field

I’ll keep these anonymised — both are composites drawn from work in the Malaysian market, not named clients.

The fashion accessories brand that finally planned for Raya

A regional fashion accessories brand ran Malaysia for two years on a calendar built around the double-digit sales — 9.9, 11.11, 12.12 — because that’s what the regional plan tracked. Raya showed up in the plan as a generic “festive promo” with a modest budget, slotted in like any other campaign.

The shift was simple and it was structural: treat Raya as the anchor tentpole of the Malaysian year and build backward from it. Inventory committed months ahead. Creator activations briefed for the Raya ramp specifically — gifting angles, baju and accessory pairing content, the cultural register that Raya demands. Bundles priced in clean RM bands around the free-shipping threshold. The double-digit sales stayed in the plan — but as supporting acts, not the headline.

The lesson for the reader: the Malaysian demand curve has a different peak than the regional one. Plan the year around the peak that’s actually there, not the one your regional template assumes.

The wellness brand that stopped translating and started code-switching

A wellness brand was running English-language content across all three Malaysian platforms — clean, professional, on-brand. It wasn’t failing. It was just quietly leaving engagement on the table, and nobody could see it because there was no failing number to point at.

The change was a content rebuild, not a translation pass. Mid-tier local creators briefed to make natively bilingual content — code-switched BM/English the way the audience actually speaks, calibrated more BM-forward for the East Coast and northern TikTok audience, more English-tolerant for the urban Shopee buyer. Same products, same claims, same brand — different content, built from the Malaysian ear out.

The lesson: in a market where English “works,” the cost of lazy localisation is invisible. You don’t see the engagement you’re not getting. Bilingual-native content closes a gap you didn’t know was open.

The Market-Entry Checklist

If you’re a regional brand planning Malaysia — or fixing a Malaysia plan that’s quietly underperforming — here’s the Monday-morning version.

1. Build the platform mix as a genuine three-way split. Shopee for the transactional floor, TikTok Shop for discovery and livestream growth, Lazada/LazMall for brand-equity protection. Budget and headcount should reflect three real platforms, not an ID-style two-platform tilt with Lazada as an afterthought.

2. Rebuild the price ladder natively in RM. Clean anchor points in the RM30–RM100 band where it fits your category. Price hero SKUs against the platform free-shipping threshold deliberately. Don’t convert the IDR ladder straight across — rebuild it for how Malaysians read price.

3. Anchor the calendar on Raya. Make Hari Raya the tentpole and build inventory, content, and creator activations months ahead. Keep 6.6 / 9.9 / 11.11 / 12.12 as supporting peaks. Plan the MY demand curve, not the ID one.

4. Commission natively bilingual content. Brief local creators to make code-switched BM/EN content from scratch — not translated English. Calibrate the mix by region and platform: BM-forward for East Coast and northern TikTok, more English-tolerant for urban Shopee.

5. Set content intensity by category margin. Beauty and fashion carry the content engine and the livestream model. F&B converts on live but on thinner margins — tighten the unit economics. Electronics belongs on Shopee search, not the TikTok impulse model. Stop running one uniform plan across margin profiles.

6. Staff Malaysia like the growth market it is. Put a real operator on the +47.6% market. The spillover-staffing model is how the fastest-growing market in the region quietly underperforms for you while it accelerates for everyone else.

The Close

Malaysia looks like the easy market. Small, developed, English-speaking, familiar. That familiarity is a reason why many cross border sellers and overseas brands entering Malaysia get trapped — it’s why it gets the borrowed plan and the junior operator while it grows faster than almost anywhere else in the region.

It is not a smaller Indonesia. It’s a genuinely tri-platform market, with its own price psychology, its own calendar built around Raya, its own bilingual ear, and its own category economics. Run someone else’s playbook against it and you’ll grow — Malaysia is growing for everyone — but you’ll grow slower than the market, and you’ll never be able to see exactly how much you left behind.

Build the Malaysian plan from the Malaysian seat. The fastest-growing market in Southeast Asia is worth its own playbook.

Related reading from the archive:

• Thinking Full Funnel — connecting discovery to conversion across platforms (Jan 2026)

• Cross-Border Without Borders — Malaysia as a launchpad into the rest of SEA (Mar 2026)

• How Shopee Malaysia’s 2026 Fee Changes Will Reshape Your Margins — the platform-economics layer beneath all of this (Jan 2026)

In May 2025, putting a 15–30s video on a PDP was the differentiator. In May 2026, it’s the entry ticket. The new differentiator is the AI stack you wrap around it — the tools that pull keywords, write the title in BM and EN, generate the lifestyle hero shot, animate the still into a vertical clip, draft the review reply, and audit the listing weekly without a human touching the Seller Centre.

This week’s piece is a refresh and a deepening. Same five pillars from the original video-first PDP framework. But each pillar now sits on three AI layers — platform-native, third-party, and human QC — and the brands pulling ahead in 2026 are the ones who’ve figured out which layer does which job.

If you run a brand, manage an enabler team, or are building a marketplace operation, this is the playbook for the next twelve months. Three downloadable templates at the bottom, including a PDP audit checklist you can drop a SKU list into tomorrow morning.

What Changed Between June 2025 and May 2026

Three platform shifts you need to internalise before any of the tactical stuff matters:

1. Shopee turned the Seller Centre into an AI workbench. ChatGPT Go (powered by GPT‑5.1) rolled out to all Malaysian sellers in January 2026, embedded directly in Seller Centre — free. Product AI Optimiser now generates titles, descriptions, attribute fills, and size recommendations from a single hero image. Shop AI Assistant handles the chat tier underneath.

2. Lazada finished its GenAI suite for sellers. AI Smart Product Optimization auto-rewrites titles, regenerates images on white background, and produces multilingual descriptions across BM, ID, TH, VN, EN. AI Lazzie is now embedded in shopper search — and it reads your listing the way a human shopper does, so listings written for keywords-only are getting deprioritised.

3. TikTok Shop closed the content gap. Symphony Creative Studio + AI Fashion Video Maker now generate vertical product video from a single still in under two minutes. “List with AI” populates titles, descriptions, and category from one product photo. The bottleneck has moved from “can we make video” to “can we make video that’s on-brand and on-message“.

The net effect: the cost of producing a baseline-quality listing has collapsed. Which means the floor has risen for every category. A generic AI-written PDP from a competitor is now better than a manually-written PDP from most brands twelve months ago. To pull ahead, your stack has to be more disciplined, not just more enabled.

The Refreshed Framework: 5 Pillars × 3 AI Layers

Same five pillars I’ve been using on Carlo Rino, Goodyear and our other accounts. What’s new is the layered AI treatment underneath each.

The principle: let platform-native do the volume work, use third-party where it does a job better, and protect every output with a human QC layer. Skipping the QC layer is what produces the AI-slop listings you can spot immediately on Shopee right now — over-promising titles, hands with six fingers in lifestyle shots, descriptions that contradict the size chart.

Pillar 01 · Title: Keyword Clusters, Not Copywriter Instinct

What changed: Shopee’s Product AI Optimiser and Lazada’s Smart Product Optimization both now write titles from a hero image. Quality is good for generic categories, mediocre for fashion, weak for nuanced wellness or technical electronics.

The 2026 process:

1. Pull the top 200 search terms in your category from Shopdora or Shopee Insights. Don’t skip this — the platform AI doesn’t know what’s trending in your sub-category this week.

2. Cluster keywords by intent: brand, material, use-case, BM phrasing, occasion. Use ChatGPT/Claude for clustering — it takes 60 seconds.

3. Run the platform-native generator on a hero image. Treat the output as a first draft.

4. Rewrite the title using your cluster — brand · BM phrase · material · style · colour · use-case · size cue. One title, every search lane covered.

5. A/B test two title variants per SKU for two weeks. Move winners into the template.

Metric to watch: organic CTR on the search results page. A well-titled listing should sit at 3–6% organic CTR for fashion and accessories on Shopee MY.

Pillar 02 · Imagery: Templated Slot Order, AI-Assisted Production

What changed: AI image tools have collapsed the cost of producing lifestyle and scene variants. A one-product shoot used to need a half-day studio booking; now it needs Photoroom or Pebblely and 20 minutes. Cost per image: $0.10–$0.50 vs the $25–$150 average for traditional product photography.

The 2026 image stack — same slot order, AI-augmented production:

• Slot 01 — Hero (white BG): Shoot once on white. Use Photoroom or Lazada’s auto-clean for clean cutouts.

• Slot 02 — Lifestyle: Pebblely generates 5 scene variants from your hero in under a minute. Pick the one that matches your brand. Always check for AI artefacts — six fingers, melted backgrounds, wrong textures.

• Slot 03 — Video (covered in Pillar 03).

• Slot 04 — Detail / size: Macro shot of stitching, hardware, zip. Plus a dimension diagram — Canva still wins for this.

• Slot 05 — Social proof / UGC: A grid of three real customer photos pulled from reviews, with permission. AI cannot fake this layer credibly yet, and the algorithms increasingly reward listings that have it.

Metric to watch: add-to-cart rate. A well-staged gallery should lift ATC by 30–60% off the same traffic.

Pillar 03 · Video: From “We Don’t Have a Video” to “We Have Six”

This is the pillar that’s been most fundamentally rewritten in twelve months.

What changed: TikTok Symphony Creative Studio + AI Fashion Video Maker can take a single product still and generate a 5–15s vertical clip — with voiceover, captions, music, and a digital avatar holding the product. Production time per video has dropped roughly 70%.

The 2026 video doctrine:

• One hero video per SKU is no longer enough. Generate three to six variants — different hooks, different captions, different soundtracks — and let the platform’s content algorithm pick the winner.

Hook in the first 3 seconds. Show the “wow” — the texture, the open compartment, the before-after. Static product spinning on white is what your competitor uploaded in 2024.

Vertical 9:16 only. Square is for Shopee gallery slot 3 only.

Captions baked in. 85% of SEA marketplace video views are sound-off mobile. No captions = no message.

• End on a CTA frame. A 1-second still showing the bundle, the price, or the variant range. Symphony will generate this if you prompt it.

Where third-party still wins: Runway and Dreamina (Seedance 2.0) for higher-fidelity creative — model walking with the bag, hands assembling the product, before/after scene transitions. Symphony is faster and cheaper for the volume layer; Runway/Dreamina for the hero assets you’ll cut campaign ads from.

Metric to watch: PDP video view-through rate, then ATC rate from video-engaged sessions. The Dr. Squatch benchmark (9.9% CVR on video-engaged vs 3.2% on non-engaged on the same page) is a useful directional target.

Pillar 04 · Attributes: The Pillar That Quietly Kills CPAS

What changed: Shopee Product AI Optimiser now auto-fills attribute fields from your title, description, and image. Most sellers leave it on default. Most sellers are leaving filter visibility, suggested-product placement, and Meta CPAS catalogue feed quality on the table.

The 2026 process:

1. Run Product AI Optimiser’s attribute autofill across the catalogue. This is the volume work.

2. Spot-check 10% manually. The AI gets material wrong on textiles roughly 30% of the time in our experience, and use-case wrong on bags about 20% of the time.

3. Map every attribute that exists in Shopee/Lazada/TikTok Shop’s filter taxonomy — not just the ones the AI suggests. If the platform offers “office bag” as a filter and the AI didn’t pick it up, fix it.

4. Push attribute updates to your Meta product catalogue feed (CPAS). A clean attribute set means cleaner DPA targeting and a higher Catalogue Quality Score on Meta.

Metric to watch: filter-visibility coverage and Meta CPAS Catalogue Quality Score. Internal target on our accounts: every SKU appears in at least 8 filter combinations.

Pillar 05 · Reviews & Q&A: The Layer Where Brands Lose Trust the Fastest

What changed: Shopee Shop AI Assistant now handles 50%+ of pre-sale chat queries automatically. Average response time on Shopee MY accounts using it has dropped to under one minute. Lazada AI Lazzie does the same on the shopper-facing side.

The 2026 reviews & Q&A Approach:

• Turn on Shop AI Assistant on day one. Free, in Seller Centre, immediate response-time lift. Customise the FAQ block — don’t run defaults.

• Use ChatGPT Go in Seller Centre to draft review replies in BM and EN. Tone calibration is critical — AI defaults too formal for Shopee Malaysia. Add a tone instruction: “reply in casual but respectful BM-English mix, like a small business owner”.

• Shopee provides a Review Summary. Shopee algorithmically reads pinned-review freshness as a quality signal. This is the single highest ROI 5-minute task in the playbook — set a Monday recurring reminder to reply to reviews matters.

• Escalation rules. Anything below 4 stars, anything mentioning “fake”, anything mentioning logistics — human reply within 4 hours, no AI. A bad AI reply on a 1-star review is what kills a listing.

Metric to watch: review velocity (reviews per 100 orders) and average review rating. Listings sitting at 4.8+ with 100+ reviews dominate the search results page on Shopee MY

The Tooling Stack at a Glance

I’ve put the full tool-by-pillar matrix in the downloadable AI Tool Stack Matrix (xlsx) — pricing, free tiers, BM support, recommended use case for each. Headlines:

• Free or near-free, in this order of impact: Shopee Product AI Optimiser, Shopee Shop AI Assistant, ChatGPT Go in Seller Centre, Lazada AI Smart Product Optimization, TikTok Symphony Creative Studio’s free tier, Photoroom (250 exports/month free).

• Paid but worth it for an enabler running 5+ brands: Shopdora (SEA keyword data), SplitDragon (Lazada keyword + competitor), Pebblely (lifestyle scene generation), Runway/Dreamina (premium video), Epsilo (ads + catalogue at scale).

Avoid: any “magic AI listing” tool that promises a one-click optimization without showing you keyword data. The cheap ones are wrappers around GPT‑4 with no SEA-specific search data, and you’ll just get generic English output.

Field Test: What Shopee’s Product AI Optimiser Actually Does (And Where It Stops)

Most “AI tools for sellers” articles are written from a press release. This one isn’t. This week we ran Shopee’s Product AI Optimiser across a live Commerceplus-managed account — a Malaysian footwear brand with roughly 1,200 active SKUs — to see what it does when you actually open it in Seller Centre. Here’s the honest read.

What it does well. The Optimiser dashboard is genuinely useful and it’s free. On the account we tested, it reported 484 products AI-optimised, contributing 8.6m impressions and answering 495 buyer questions automatically. It works across three jobs: a size-recommendation engine (it turns your size chart into a “what size fits me” answer in chat), AI-powered product showcasing (it spins your selling points into images, short video, and smart replies), and image poster polish (it drops your product onto an enhanced background). For the size and smart-reply layer especially, it’s a real time-saver — on a 1,200-SKU catalogue, manually answering “which size” for every buyer is impossible, and the AI handles it well.

The catch — and it’s the whole point of this newsletter. The Optimiser splits your catalogue into “To Be Optimised” and “AI Optimised”, and the column that decides which bucket you’re in is labelled Tasks to Empower AI. On the account we tested, that meant 706 products waiting on the seller to do two things: add a size chart, and upload selling points. Open any product’s Optimise flow and the first thing it asks for is your selling points — with a good-versus-bad example to coach you. In other words, the AI doesn’t know what’s good about your product. It amplifies what you tell it. Give it a sharp, specific, benefit-led selling point and it produces good showcasing. Give it nothing, or give it “Long Lasting”, and it produces nothing useful — or worse, generic filler.

The gaps an enabler needs to plan around:

Two Anonymised Case Studies

Case A — Malaysian fashion accessories brand (Commerceplus managed). Top 30 SKUs went through the full 5-pillar AI-augmented refresh over 6 weeks. Result: organic search impressions +180%, ATC rate from 2.1% to 6.4%, conversion rate from 0.9% to 2.8%.

Case B — Regional electronics brand (Commerceplus managed). Different starting point — listings were technically fine but flat on video. We layered Symphony Creative Studio for volume video plus one Runway-rendered hero video per SKU for the top 10. PDP video view-through rate climbed from 18% to 47%; ATC rate from video-engaged sessions doubled.

The 2-Week Rollout Plan (For One Brand, ~30 SKUs)

• Week 1: Audit the 30 SKUs against the 5-pillar AI checklist (download below). Pull keywords. Lock the title template, image template, and video brief. Approve brand voice for AI replies.

• Week 2: Roll out titles, regenerate galleries, deploy 3 video variants per SKU, autofill + audit attributes, switch on Shop AI Assistant with custom FAQs.

• Week 3 onward: Weekly 30-minute review — ranking, ATC, CR, review velocity. Iterate on the worst-performing 5 SKUs.

Most brands will see an inflection by week 4, full impact by week 8. If you’re not seeing movement in week 4, the issue is almost always either pricing or stock availability, not the listing itself.

Tactical Tip of the Week

Before you do anything else with AI listing tools, write a brand voice instruction prompt and save it in a doc. Every AI tool you use — ChatGPT Go, Shopee Product AI Optimiser, Lazada AI Smart, Symphony — should be primed with the same 2-paragraph brand voice instruction every time. This is the single biggest difference between brands whose AI output looks like a brand and brands whose AI output looks like AI.

I’ve put a starter prompt in the downloadable AI Prompt Library.

Industry Pulse: The AI-Slop Backlash Is Coming

Worth flagging: shopper trust in obviously AI-generated listings is starting to crack. We’re seeing review comments calling out “this looks AI” on lifestyle images with rendering artefacts, and platforms are quietly down-weighting listings that use stock-feeling generic descriptions. The brands that win the next 12 months will use AI heavily but invisibly — the human QC layer is what makes the difference between AI-augmented and AI-slop. Lean into the layered model.

Personal Insights from Stan

The temptation in 2026 is to lean entirely on platform-native AI because it’s free and embedded. Don’t. The free tools get you a baseline-quality listing — which is the same baseline-quality listing your competitor is producing with the same tool. Differentiation now lives in the third-party layer and the human QC layer. The brands that treat AI as a production line — with a brand voice prompt, a slot-order template, a weekly QC pass, and a metric for every pillar — are the ones we’re seeing pull 2–3× ahead of category averages on the accounts we manage.

The 5-pillar framework hasn’t changed. What’s changed is the speed of execution. A year ago, a full PDP refresh on 30 SKUs was a 6-week project. Today it’s 2 weeks if you have the stack. If you’re still spending 6, that gap is your competitor’s moat.

Three downloadable templates are linked below — start with the audit checklist this week.

Keep testing, keep optimising, see you next Sunday.

— Stan

Downloads

• PDP Audit Checklist (xlsx) — 5-pillar scorecard, weighted scoring, AI tool column per pillar

• AI Tool Stack Matrix (xlsx) — every tool by pillar, pricing, free tier, BM support, use case

• AI Prompt Library for Listings (docx) — copy-paste prompts in BM + EN for titles, descriptions, image briefs, video scripts, review replies

Ecommerce Owner’s Manual is a weekly Sunday newsletter from Commerceplus — the Southeast Asian eCommerce enabler for brands serious about marketplace growth. Subscribe at ecommercesea.substack.com.

Hope you had a restful day off on Labour Day. Am back after a week’s break to dive into a topic I’ve spent the past 2 months doing research on.

If you’ve been reading our Beauty, Fashion, and Health & Wellness deep dives, you know the playbooks I’ve been writing about. Premium positioning. K-beauty halo. Brand archetypes. Authority moats. All of those categories — even the toughest of them — let you imagine a world where you build a 70% gross margin brand and pay the platform their take and still go home with money.

Then I sat down with three different grocery and food brand operators in the last six weeks, and the conversation went somewhere completely different.

The first was the digital lead at one of Malaysia’s most established biscuit houses — a brand your parents grew up on. “Stan, our gross margin on the SKUs that move volume on Shopee is 22%. After platform fees, ad spend, fulfilment, and the discount we have to absorb during 11.11, we’re losing money on those orders. But if we don’t run those campaigns, our search ranking dies. So we’re stuck.”

The second was a founder of an emerging RTD beverage brand in Klang Valley. Healthy formulation, beautiful pack, listed on Shopee, Lazada, and TikTok Shop. “We can hit RM200K monthly GMV when we push hard. But our cost-per-bottle delivered is RM4.20 against an RM12 retail price, and our reseller channel is undercutting us by RM2 on Shopee. I don’t know how Mayora makes this look easy.”

The third was the CEO at Beacon Mart — a Malaysian frozen food brand we worked with for three years. They’re a different story. They went from RM300K annual GMV in 2022 to RM7.6M in 2025. November alone hit RM1.6M. They figured out something the other two haven’t. (More on that later.)

What ties these three conversations together is the question every grocery and food brand in Southeast Asia is wrestling with right now: can you actually make money on ecommerce in this category, or is it just expensive distribution?

This deep dive is my best attempt at answering that.

The State of Grocery & Food E-commerce in Southeast Asia

Let me set the scale before we get into the mechanics.

Southeast Asia’s overall e-commerce market crossed US$160B in 2025 and is on track for ~$215B by end of 2026 (16.5% CAGR). Within that, packaged food, beverages, household groceries, and fresh — collectively the FMCG / grocery basket — represents one of the fastest-growing but lowest-margin segments.

Online grocery in SEA grew at a 97% compound rate from 2020 onwards. But base penetration is still small: across the region, online grocery is only ~5–8% of total grocery spend, versus 20%+ in markets like China or the UK. The runway is enormous; the margin pressure is real.

A few specifics that frame the rest of this issue:

• Local and regional FMCG manufacturers now command over 50% of SEA FMCG market value (Bain / NielsenIQ, Nov 2025). This is a meaningful shift from a decade ago when multinationals dominated the shelf. Mayora, Indofood, Aik Cheong, Khong Guan, Wings Group, Munchy’s, F&N, Yeo’s — local incumbents are ascendant.

• Shopee, TikTok Shop, and Lazada own 84%+ of regional platform GMV combined. For grocery specifically, Shopee dominates with its Shopee Supermarket vertical. Lazada has positioned LazMart and RedMart (in Singapore) as the premium grocery destination. TikTok Shop’s food and beverage vertical is exploding — F&B is one of the fastest-growing TTS categories in MY, TH, and VN.

• Quick commerce is a parallel battlefield — GrabMart, Pandamart, Foodpanda’s grocery vertical, and FairPrice On (in Singapore). These are different unit economics from marketplace ecommerce: 30-minute delivery, smaller baskets, higher delivery cost, but enormous frequency.

• Bain projects FMCG category growth in H1 2025 was weaker than prior years, with consumers tightening discretionary spending and trading down within categories. This is the squeeze: more channels demanding shelf, more discounting pressure from consumers, less category growth to share around.

Groceries operate on 8–25% net margins at the brand level. On ecommerce, the platform plus marketing plus fulfilment can eat 25-40% of revenue before you’ve paid for the product itself. So the question isn’t just “how do we grow” — it’s “can the unit economics ever work, and under what model?”

The Three Operating Models — Pick the Right One Or Lose Money Faster

This is the section that took me the longest to write, because most market reports skip it entirely. They tell you to “go on Shopee” without telling you how to go on Shopee — and the how changes everything.

In grocery and food, there are three meaningfully different operating models for an ecommerce presence. Each has dramatically different unit economics, working capital requirements, and team capabilities. Most brand owners I speak to have only thought about one or two of these — usually the wrong one for their business.

Model 1: Retail (e.g., Shopee Supermarket, Lazada Mart, RedMart)

How it works: You sell your products to the platform at wholesale or distributor pricing. The platform takes ownership of the inventory, lists and merchandises the SKUs, runs their own promotions, manages fulfilment from their warehouse, and pays you on payment terms (typically 30–60 days). You’re effectively a vendor to a retailer that happens to be online.

Who runs it well: Established FMCG brands with proven volume velocity, supply chain capacity to deliver case-pack into platform DCs, and pricing discipline across modern trade. Think Nestlé, Unilever, F&N, Yeo’s, Mayora — they’re already vendors to Tesco/AEON/Lotus, so vending into Shopee Supermarket is structurally similar.

The unit economics: You give up 18–28% margin to the retailer (similar to MT discounts), but you don’t pay platform commissions, ad fees, or fulfilment costs as line items. You don’t manage CRM. The platform owns the customer relationship — you’re just a brand on their shelf.

Where it breaks: You have zero brand control on the listing experience. Promotions are decided by the retailer (and they will discount your SKU to drive their basket, not yours). You’re not building any customer-level data. And you’re at the mercy of category buyers who can delist you tomorrow.

Model 2: Outright Marketplace (Brand-Run Storefront)

How it works: You run your own Shopee/Lazada/TikTok Shop storefront. You own the listings, the pricing, the promotional mechanics, the ads, the customer service — everything. The platform takes a commission on every sale (Shopee ~15% Mall fees, ~9% Marketplace, plus campaign fees and shipping subsidies). You handle fulfilment yourself or via a 3PL (or via Shopee FBS / Lazada FBL if you opt in).

Who runs it well: Most brand-owned grocery operations sit here, and most struggle. You need an internal team — or an enabler partner — managing storefront, ads, content, livestream, and campaigns weekly. Beacon Mart is the textbook example of this model executed properly. Most established FMCG brands are actually under-investing in this model because they treat it as “nice to have” alongside MT, when it should be a P&L center.

The unit economics: Higher gross take than retail, but you absorb all the variable costs. On a typical RM100 grocery SKU, you're looking at:

That negative number is what every grocery operator in SEA is staring at. Until repeat rate climbs (more on that below), volume promotions can’t fix this — they make it worse.

Where it breaks: You burn cash if you’re not at scale. Below RM200K monthly GMV, fixed costs (team, content, agency fees) eat your margin. Above that, you can amortise. The brands that win run this model with discipline: SKU-level P&L tracking, campaign math before they participate in 11.11, pricing parity with reseller channels.

Model 3: Service Fee (Enabler-Operated)

How it works: You retain ownership of the inventory and the storefront entity, but a specialist enabler operates the day-to-day — listings, ads, content, customer service, campaigns, fulfilment integration. You pay either a fixed monthly retainer plus a percentage of net GMV (typical: RM 10–25K base + 8–15% of GMV after platform fees), or a pure rev-share model. The enabler is on your side of the table — it’s not the platform.

Who runs it well: Brands at RM500K-3M monthly GMV scale who don’t want to build a 6-8 person internal ecommerce team but need professional operations. Also new entrants who want to learn the channel before committing to in-house infrastructure.

The unit economics: Same as outright on the platform side, plus the enabler fee. The math only works if the enabler can deliver materially better velocity than your in-house team would — typically 30-60% better GMV at the same or better contribution margin, achieved through better ad spend efficiency, better campaign positioning, better content velocity. If the enabler isn’t moving those numbers, you’re paying for distribution you could have done yourself.

Where it breaks: Misaligned incentives kill this model. If the enabler is paid on GMV without margin accountability, they’ll burn ads to grow top-line at your expense. The contracts that work tie at least 30–50% of variable comp to contribution margin, repeat rate, or new buyer acquisition — not just gross sales.

How to Pick

There is no “best” model — there’s the model that fits where your brand is.

A useful diagnostic:

• You’re already a Tier 1 modern trade brand with national distribution and category share?

  Retail (Shopee Supermarket, LazMart, RedMart) makes sense as the entry — you’re already operationally aligned.

• You’re a challenger brand or hero-SKU player who wants to control your story and your customer data?

  Outright is the only way you’ll build the moat.

• You’re somewhere in the middle — some scale, but not at the level where you can justify a 6-FTE in-house ecommerce team?

  Service fee with a competent enabler beats both extremes.

• Hybrid is real.
  Many of the best operators I see run outright on Shopee and TikTok Shop (where the brand-building leverage is highest) and retail through Shopee Supermarket and LazMart (where their core SKUs already have distribution discipline). Same brand, two models, different objectives.

The Margin Reality — Why Grocery Math Is Different

In FMCG, brand-level gross margins typically sit between 30% (low — generic biscuits, basic RTD beverages) and 50% (high — premium chocolates, niche health-positioned drinks). Compare that to beauty (60–85% gross margin) or supplements (50–70%) and you immediately see why grocery operators can’t absorb the same platform fees.

A few realities that follow from this:

• Discounting is structurally dangerous. A beauty brand running a 20% promo gives up 14 points of contribution margin. A grocery brand running the same 20% promo on a 35% gross margin SKU is left with 15% to cover everything — fees, fulfilment, returns, ad spend. That’s why so many FMCG brands look “active” on Shopee but quietly post negative contribution.

• Repeat rate is the lever, not AOV. In beauty, you can lift AOV with bundles. In grocery, the cleanest economics come from getting the same customer to buy the same SKU again — and again, and again. Beacon Mart’s repeat rate hit 40-42% in 2025, and that’s the engine. CAC amortises over 3-5 purchases. The unit economics of customer #1 doesn’t have to work if customer #1 becomes customer #1, #2, #3, and #4.

• Hero SKUs do the heavy lifting. In every successful grocery brand on ecommerce, one or two SKUs drive 40-60% of revenue — and they’re usually consumables with natural reorder cycles (essence, RTDs, snacks, instant noodles). The rest of the catalogue rides the halo. Trying to sell your entire SKU range with equal effort is the fastest way to dilute margin.

• Subscription is the unfair advantage. A 10-15% subscription discount in exchange for auto-replenishment every 4-6 weeks is the single highest-leverage promotion in grocery. You acquire once, you collect 6-12 orders, your CAC amortises, and your forecasting gets dramatically easier. Most SEA brands haven’t built this yet — Shopee and Lazada’s subscription tooling is improving but still weak.

Cold Chain vs Ambient — The Logistics Question Most Brands Underestimate

Grocery is the only category where logistics architecture materially changes the brands you can build. Beauty brands ship in jiffy bags. Fashion brands ship in poly mailers. Grocery brands ship in everything from foam-insulated cold packs to bulk pallet drops.

Three distinct fulfilment realities in this category:

Ambient packaged food (biscuits, canned goods, instant noodles, packet drinks, dry seasonings, snacks):

• Standard 3PL fulfilment works. Pick/pack RM1.50–2.00, last-mile RM3–6 within Klang Valley.

• Volumetric weight matters — packaged food is bulky relative to value, so per-order economics get expensive on small baskets.

• The fix is bundle minimums. Most successful ambient brands enforce a minimum spend (RM50-80) before free shipping kicks in, and lean into multi-pack SKUs.

• This is also where case-pack on retail (Shopee Supermarket / LazMart) genuinely shines — fulfilment per unit drops dramatically when the platform picks from a pallet vs picking single jiffy bags from your warehouse.

Chilled and frozen (RTD dairy, frozen meats, fresh produce, ice cream):

• Cold chain 3PLs are a different cost category — RM8–18 per order in MY, even higher in SG.

• Coverage is limited. West Malaysia mostly works; East Malaysia is largely impossible without partnering with platform-native cold chain (FBS Cold, Lazada Cold) where available.

• Beacon Mart owns its own cold chain fleet — that’s a moat, not a cost line. Most brands don’t have this option, which is why frozen DTC is a small but defensible niche rather than a mass play.

• Cold chain lead time forces tighter forecasting. Stockouts are catastrophic (you can’t pull from a backup warehouse). Overstock is catastrophic (you can’t slow-sell expiring inventory).

Quick commerce (GrabMart, Pandamart, Foodpanda Grocery):

• Different math entirely. The platform takes 22-28% commission, but they handle the dark store, the picker, the rider, and the 30-minute SLA. You ship case packs into their dark stores periodically (typically 1-2x weekly).

• Best for convenience SKUs: drinks, snacks, tobacco, baby essentials, ready meals, household basics. Not great for hero SKUs that consumers buy on planned shopping.

• A growing share of impulse F&B revenue in SEA cities is migrating here. If you’re a beverage or snack brand and you’re not on at least Pandamart or GrabMart in your home market, you’re missing an entire shopping mission.

The decision tree most operators don’t draw:

A practical rule of thumb: if your gross margin is below 35% and your average ticket is below RM50, the platform’s own fulfilment layer (FBS, FBL, GrabMart) will almost always beat what you can do yourself. Above those thresholds, hybrid (your warehouse for low-velocity SKUs, FBS/FBL for hero SKUs) usually wins.

Country-Specific Realities

The same brand cannot run the same grocery playbook across SEA. Here’s how I think about each major market.

Malaysia: The Volume Sweet Spot

Shopee Supermarket dominates online grocery in Malaysia, with Lazada Mart as the secondary vertical. TikTok Shop is gaining share fast in F&B (especially RTD beverages and snacks). GrabMart and Pandamart are real for impulse SKUs in Klang Valley, Penang, JB.

Halal isn’t optional. ~62% of the population is Muslim and halal certification (JAKIM) is a baseline trust signal across food and beverage. If your packaging doesn’t carry it visibly, you’re locking yourself out of a significant share of the market — even for products that don’t need certification (e.g., vegetarian foods).

Bilingual SEO matters. Your Shopee listing should carry English and Bahasa keywords, and for Chinese-leaning brands, simplified Chinese as well. Beacon Mart’s chicken essence listing carries all three — that’s not vanity, it’s how their search ranking compounds.

Voucher dependency is structural. Malaysian consumers wait for 11.11, 12.12, 9.9, 6.6, Ramadan and Hari Raya promotions. Your annual P&L needs to absorb a 60-30 split between baseline months and campaign months — and your campaign-month math has to work without panic.

Unit economics target: AOV RM45-90, contribution margin 8-15% on outright marketplace, repeat rate 30-40% by Month 9, LTV:CAC 6-10x.

Indonesia: The Local Brand Dominance

Indonesia is the largest SEA grocery ecommerce market by absolute size — and the one where local brands have the deepest moat. Mayora, Indofood, Wings Group, and Garudafood collectively dominate shelves both physical and digital. Foreign brands struggle here unless they have proper local distribution (or buy a local brand).

TikTok Shop Indonesia has been particularly strong for F&B — viral RTD launches and snack brands routinely break through. Shopee remains the basket platform for planned grocery shopping. Tokopedia is third but credible for premium positioning.

Cash on delivery is still 30-40% of orders in tier 2/3 cities — your fulfilment partner needs to handle COD reconciliation cleanly or you’ll bleed on failed deliveries.

Halal certification (MUI) is rigorously enforced for any Muslim-targeted SKU. Plan 90-180 days for new product certification.

Philippines: The Group-Buy Market

The Philippines has a structurally different grocery ecommerce dynamic from MY/ID. Group buy and bulk-pack purchases dominate — driven by family-size households and price sensitivity. Lazada and Shopee both run group-buy mechanics aggressively. TikTok Shop is growing fast for snacks and beverages.

COD is 50%+ of orders, which means logistics complexity and failed delivery rates (typically 10-15%) are real costs, not edge cases.

The Philippines is also the SEA market where local supermarket chains (SM Markets, Robinsons, Puregold) have extended into ecommerce credibly, creating a separate channel that operates somewhat like the modern trade extension Shopee Supermarket plays in Malaysia.

Thailand: The Mature Market

Thai consumers are sophisticated, price-aware, and brand-loyal. The grocery ecommerce market is more fragmented across Shopee, Lazada, LINE Shopping, and supermarket-owned channels (Big C, Tops). 7-Eleven Thailand’s online grocery extension is a meaningful channel.

LINE is non-negotiable. Your CRM strategy in Thailand must include LINE — broadcast and 1:1 — or you’re missing the highest-engagement channel for this market.

Premium and health-positioned SKUs perform well. Thai consumers will pay for quality, but you have to show your work — ingredient stories, certifications, founder credibility.

Singapore: The Premium Ecommerce Grocery Market

Singapore has the highest grocery ecommerce penetration in SEA (estimated 12-18% of grocery spend) but a small absolute market. RedMart (Lazada-owned) is the dominant player, with FairPrice On as the supermarket-extension challenger. Shopee Supermarket exists but is less central than in MY/ID.

Singapore is the right market to test premium grocery propositions and clean-label/health-positioned SKUs. AOV is high (SGD 60-120), customer expectations are exacting (next-day or same-day delivery, perfect picking, no out-of-stock), and the market is small enough that you can learn fast without enormous spend.

Vietnam: The Fastest-Growing TTS F&B Market

TikTok Shop has reshaped F&B ecommerce in Vietnam. Live commerce sells massive volume of snacks, RTD beverages, and packaged foods — often with creator-driven hauls in a format that’s harder to replicate in MY or SG.

Shopee Vietnam remains strong for planned baskets. Lazada is third. Local platforms (Tiki, Sendo) have specific use cases but are not where most volume sits.

Vietnam’s new E-Commerce Law (effective July 2026) requires foreign sellers to have legal representation or operate through platform programs. If you’re a regional brand looking at Vietnam, the regulatory side now matters more than it did 18 months ago.

Marketing for Grocery — What Actually Works

A grocery brand cannot run a beauty brand’s content playbook, even if the platforms reward similar formats. Three things make grocery different:

1. Repeat triggers beat aspiration. A consumer doesn’t aspire to your biscuit. She runs out of it. The most effective grocery content reminds the customer that the product is in their life, with a low-friction reorder. WhatsApp post-purchase flows (”Your last delivery was 28 days ago — most customers reorder around now”), Shopee follow-up vouchers, email “stock up before 11.11” cycles.

2. Social proof beats founder story. Beauty consumers want to know your founder. Grocery consumers want to know that 47,000 other people bought it last month and 92% would buy again. Surface review counts, “best-seller” badges, repurchase rates, and platform-native trust signals. You can layer founder authenticity in for premium plays, but volume grocery wins on confidence, not romance.

3. Live commerce works disproportionately well. Live cooking demos, ingredient education, real-time bundling, host-led Q&A — for F&B specifically, livestream converts at multiples of the static-listing rate. Beacon Mart hit a 488x ROI on Shopee Live during November 2025. That’s not a typo. Live commerce for grocery is one of the most under-utilised levers in this category right now.

A few specific tactics that compound:

• Bundles as the default unit. Single-SKU listings are rarely the optimal sale unit in grocery. Build “starter packs,” “monthly stock-up bundles,” “family bundles” as your hero listings. They lift AOV, improve fulfilment economics, and reduce decision friction.

• Subscription via WhatsApp Commerce. This is currently the highest-leverage growth channel for SEA grocery brands that have figured it out. ~70% of customers who consent to a 4-6 week reorder cadence stay for 6+ purchases. The economics speak for themselves.

• Geo-targeted CPAS. Beacon Mart’s Facebook CPAS campaigns ran only in Klang Valley, Penang, JB, and Ipoh — the cities where they had cold chain coverage. Cutting wasted impressions out of unservable geographies is a 30-50% ad efficiency lever.

• Account manager relationships are multipliers. Platforms reward responsiveness. Shopee and Lazada AMs control co-fund voucher allocation, mega campaign placements, surprise box mechanics, and category banners. Brands that are reliable promotional partners get a disproportionate share of platform-funded growth. This is operational, not transactional — it requires someone on your team (or your enabler) who shows up week after week.

Case Study: Beacon Mart — From RM300K to RM7M+ Annual

Beacon Mart is a Malaysian frozen and chilled food brand that we worked with for three years. They own their cold chain — organic seaweed-fed chicken farming through to last-mile frozen delivery across West Malaysia. Hero products: chicken pure essence, premium chicken cuts, marinated ready-to-cook meats, and seafood.

When the engagement started, the brand was doing 5-figure monthly revenue primarily on Lazada. Shopee was on the verge of being shut down — performance had been inconsistent and the team was about to redirect the budget. The platform AM relationships were thin. The product listings were generic. The hero SKU — chicken essence — was facing pricing pressure from resellers who were happy to undercut on the same platforms.

Three years later, the brand crossed RM7M+ in combined Shopee + Lazada revenue in 2025, with November alone clearing a 7-figure month. Roughly half of revenue now sits on Shopee, the platform they almost shut down. Repeat rate hit ~40% on both platforms. Twenty-thousand-plus new buyers were acquired in 2025 alone.

The strategy compressed into one diagram:

What actually moved the numbers:

• Bet on the hero SKU. Chicken essence had higher gross margin and higher repeat probability than any other product in the catalogue. We shifted promotional dollars and ad inventory aggressively toward driving first-time trial — Buy 3 Free 1 became the workhorse mechanic. The product quality did the rest. By 2025, chicken essence accounted for ~50% of total revenue.

• Don’t abandon a platform too early. Shopee was nearly cut. We argued for keeping it, increasing budget, and committing to consistent livestream presence. Three years later, Shopee revenue now exceeds Lazada in peak months. Platform dynamics shift — the cost of leaving and re-entering a year later is far higher than the cost of staying through a soft year.

• Platform AM relationships compound. We rebuilt direct lines with Shopee and Lazada category and account managers — supporting their campaign requests, accepting promotional asks, attending review sessions. The return: co-fund vouchers, surprise box placements, category banners, priority access to mega campaigns. None of these are buyable individually; they’re earned by being a reliable partner over many months.

• Geo-targeted Facebook CPAS focused only on the cities where the brand could deliver frozen reliably (Klang Valley, Penang, JB, Ipoh). Customers who couldn’t be served weren’t paid to be impressed.

• Storefront discipline. Every listing was rebuilt with bilingual SEO titles, 8-9 professional images plus video, and conversion-focused descriptions that justified premium pricing against the dominant incumbent in the category.

The full case study — with the full numbers and the full tactical breakdown — lives on the Commerceplus blog. The reason I include it here, in summary form, is that it’s the cleanest illustration I have of what disciplined outright marketplace operation looks like in grocery. None of these moves were exotic. The compounding came from running them consistently, over multiple years, with proper measurement.

Other Brands Worth Studying

A few more public-record examples I think about often.

Mayora Indah (Indonesia). I wrote about Mayora in The Middle Market Squeeze as an example of how distribution becomes brand. The grocery-ecommerce specific learning: when Vincenzo drove the Kopiko coffee candy moment in 2021, Mayora’s ability to be on every relevant platform, in stock, at the right price across SEA was what turned a viral moment into compounding revenue. Distribution discipline is grocery’s version of brand equity.

Khairul Aming’s KA. Sambal (Malaysia). A creator-founder brand that built RM-tens-of-millions in DTC sambal sales without ever fitting neatly into any traditional grocery model. Direct-to-consumer through their own webstore, marketplace presence on Shopee, with content velocity on TikTok carrying the demand. The lesson: cult product + direct relationship + relentless content cadence can outflank retail distribution entirely for the right SKU profile.

Aik Cheong (Malaysia). A 60-year-old coffee brand that has built credible ecommerce presence on Shopee and Lazada by treating their digital storefront like a flagship retail outlet — premium SKU sets, bundle architecture, gifting-positioned packaging. Old brand, new channel, deliberate execution.

The pattern across these three plus Beacon Mart: grocery ecommerce winners are not necessarily the brands with the biggest marketing budgets. They are the brands that pick a clear operating model, execute it with discipline for two-plus years, and treat platform relationships as a long-term asset rather than a quarterly transaction.

Old Established vs DTC — Can Incumbents Win?

Short answer: yes, but most won’t.

The question is structural. Established FMCG brands have advantages a DTC challenger cannot replicate quickly: factory cost base, distribution scale, retail relationships, established consumer trust, proven SKU performance data, working capital. These are real moats.

The disadvantages are also structural: organisational designs built for modern trade (where the buyer at AEON or Lotus is the customer), promotional discipline calibrated to retail margins (not platform margins), legal/finance teams unfamiliar with platform contracts, marketing teams optimised for above-the-line, and an internal political reality where the head of digital is two reporting layers below the head of retail.

The incumbents who win on ecommerce solve four organisational problems before they solve the marketing ones:

1. Move ecommerce to a P&L center, not a cost center. The team running Shopee should own a P&L line that’s measured separately from MT — and incentivised on contribution margin, not just GMV.

2. Resolve channel conflict explicitly. Pricing parity policy, reseller controls, MAP enforcement. The single biggest time-sink in established brands going digital is internal politics about whether ecommerce is allowed to undercut MT. Resolve it at the top of the org or every campaign becomes a litigation.

3. Build (or rent) ecommerce-native operational capability. Storefront ops, ad ops, livestream production, and platform AM management are specialist skill sets. Most incumbents need to either hire 3-6 specialists or partner with an enabler. Re-skilling traditional trade marketing isn’t fast enough.

4. Treat platform relationships as senior-level strategic accounts. Your sales lead at Shopee should be at the same internal seniority as your sales lead at AEON. Most established brands haven’t done this yet — and the platforms notice.

DTC challengers, by contrast, often arrive ecommerce-native — but they hit different walls: working capital, supply chain reliability at scale, regulatory complexity in second/third markets, and the brutal economics of converting a digital community into national distribution.

The honest truth is that the strongest grocery brands of 2030 will mostly be incumbents who got their organisational act together in the 2026-2028 window, augmented by a small number of breakout DTC challengers (mostly health-positioned, mostly cult-product driven). The rest will continue treating ecommerce as a side channel and quietly losing share each year.

What I’d Do If I Were Starting Today

A few archetypes — pick the one that fits.

If I were a Tier 1 incumbent FMCG brand (national distribution, RM 50M+ annual revenue):

• Start with Shopee Supermarket and LazMart for hero SKUs — retail model, vendor relationship, clean ops.

• In parallel, set up brand-owned outright storefronts on Shopee Mall, Lazada Mall, and TikTok Shop for the SKUs you want to control narrative on (premium tier, new launches, gifting bundles, regional/halal-specific variants).

• Build a 3-person internal ecommerce team or engage a service-fee enabler for outright operation.

• Quarterly P&L review on each model separately.

• Settle channel conflict policy at the C-level before campaign planning.

If I were an emerging DTC food/beverage brand (RM 200K – 2M monthly GMV):

• Outright on Shopee, Lazada, and TikTok Shop. Don’t bother with retail model until you have proven SKU velocity.

• WhatsApp Commerce as your CRM and subscription channel.

• Heavy investment in livestream — minimum 3 sessions per week, with a product-knowledgeable host who can talk to the camera about what makes your formulation different.

• Bundle architecture as your default sale unit, not single-SKU.

• Geo-target your Facebook CPAS to the markets you can actually fulfil into.

• Subscription mechanic via WhatsApp + a basic shopfront integration.

• Invest in your storefront like it’s a retail flagship — no jpg-looking images, no missing variants, no out-of-stocks.

If I were a frozen / fresh / cold chain player: Like Beacon Mart’s path.

• Outright marketplace on Shopee and Lazada (TikTok Shop selectively, depending on category fit).

• Invest in cold chain logistics as a core competency, not a 3PL line item — because the brands that own this can defend a moat.

• Hero SKU strategy: pick the two products with the highest gross margin and highest natural repeat probability, drive trial aggressively with volume promotions, let the product quality do the rest.

• Geo-target marketing to your serviceable footprint.

• Build platform AM relationships as a senior-level strategic priority.

• Repeat rate is the whole game.

If I were a niche / premium / health-positioned grocery brand (clean label, organic, halal-premium, etc.):

• Singapore first if your AOV is high (SGD 30+ effective basket).

• RedMart for category credibility, Shopee for volume, your own site for margin.

• WhatsApp + email CRM aggressively from day one — premium grocery customers are the most subscription-receptive segment in this category.

• Content strategy that emphasises ingredient credibility, founder/maker authenticity, and certification trust signals.

• Bundle architecture for both reorder convenience and gifting (gifting is materially under-served in SEA premium grocery).

Final Thoughts

Grocery and food ecommerce in Southeast Asia is the toughest unit-economics category any of us are operating in right now. The platform fees compress what was already a tight margin structure. The discounting expectations are brutal. The fulfilment costs eat orders below a certain ticket size. And the consumer is more price-aware, more comparison-fluent, and more impatient than they were two years ago.

But — and this is what I keep coming back to — there is no other category where consistent, disciplined operation compounds as powerfully. A beauty brand can have a hit launch that creates a year’s worth of growth. A grocery brand grows the way Beacon Mart grew: not by getting everything right in one campaign, but by getting the same things slightly less wrong every month for thirty-six straight months.

If you’re reading this and you operate a grocery, FMCG, or food and beverage brand in Southeast Asia, the most useful thing you can do this week is sit down with whoever owns your ecommerce P&L (or should) and ask three questions:

1. Which of the three operating models are we running on each platform — and is it the right one for that platform?

2. What is our true contribution margin per SKU per platform, fully loaded for fees, ads, fulfilment, and returns?

3. What is our hero SKU’s repeat rate at 30, 60, and 90 days — and how are we driving it up?

If you can’t answer those three with confidence, the work is in front of you. The good news is that the brands who do this work are seeing meaningful operational and financial returns. The discipline pays off.

As always, I’m here learning alongside you. If you’re a grocery, food, or FMCG operator working through any of this — whether you’re an incumbent figuring out organisational design, a DTC challenger figuring out unit economics, or a cold chain player figuring out how to scale fulfilment — reach out. Happy to compare notes.

Keep building.

Previous Deep Dives You Should Read:

• Beauty & Skincare E-commerce Deep Dive

• Fashion E-commerce Deep Dive

• Health & Wellness DTC Strategy

• The Middle Market Is Being Hollowed Out

• Building Beyond Marketplaces

• The Tightening Squeeze — Shopee Fee Changes

Welcome back, Ecommerce Operators and Brand Owners

If Part 1 was the “why this matters” conversation, Part 2 is the “what do I actually do on Monday morning” conversation.

Because the forecasting theory is one thing. The operational reality of managing inventory across Shopee, Lazada, TikTok Shop, and maybe your own site — across multiple Southeast Asian markets — is another thing entirely.

Let’s get into it.

The formulas that actually matter — explained in plain English first

Every time I’ve seen someone talk about supply chain and approaches in front of a brand team, the room glazes over in about 90 seconds. Z-scores, standard deviations, square roots — it all starts to feel like a statistics class nobody signed up for.

So before I show you the formulas, let me tell you what they actually are in normal human language. Because these three numbers are not complicated concepts. They’re just three questions, and every operator — whether you’re selling skincare on Shopee or running a distributor’s marketplace channel — needs answers to all three.

Think of it like driving a car.

• Safety stock is your spare fuel can in the boot. You don’t use it every day. It’s there for the moments when demand suddenly spikes, or your supplier calls to say the shipment is two weeks late, or a TikTok creator accidentally puts your product on blast. If you never have to tap into it, great — it just sat there doing its job. If you ever need it, you’ll be very glad it was there.

• Reorder point is your low-fuel warning light. When your tank drops below a certain level, the light comes on and you know: time to refuel, now, before you get stranded. The reorder point is exactly that — the stock level at which you must place your next order, or you’ll run out before the new shipment arrives.

• Order quantity is how much fuel you actually pump. Not too little (you’ll be back at the pump every three days). Not too much (you’ll spill, or tie up cash you could have used elsewhere). Just enough to cover the distance until your next planned fill-up, accounting for what’s already in the tank.

That’s the whole concept. Safety stock, reorder point, order quantity — spare fuel, warning light, pump. Now let’s walk through each one with the numbers, using a real-world worked example you can copy.

1. Safety Stock — your spare fuel can

The formula looks intimidating, but let’s unpack it one piece at a time:

Safety Stock = Z × σd × √L

Three things you need, and what they mean in practice:

• Z is basically a confidence level. How often are you willing to run out? If you want to stock out only 1 in 20 reorder cycles (95% of the time you’re covered), Z = 1.65. For hero SKUs during mega sales, you might want 98-99% confidence (Z = 2.05 or 2.33). For long-tail stuff that nobody will riot about if it’s temporarily unavailable, 90% (Z = 1.28) is fine.

• σd (sigma-d) is the standard deviation of daily sales. Plain language: how much your daily sales bounce around. If you sell 30 units most days, but some days 20 and some days 40, your σd is roughly 8. The more erratic your sales, the bigger σd, and the more safety stock you need. You can calculate this in Excel with =STDEV(your daily sales column). Takes 10 seconds.

• √L is the square root of your lead time in days. Why the square root and not just lead time? Because demand risk grows slower than time — doubling your lead time doesn’t double your risk, it increases it by about 1.4x. Don’t worry about why statistically; just know that longer lead times need bigger safety buffers, but not linearly.

Worked example — step by step:

Quick reference table for Z-scores:

(Source: Nventory)

2. Reorder Point — your low-fuel warning light

Once you know your safety stock, the reorder point is trivial. It’s just the stock level at which you must place your next order, so that the new shipment arrives right around the time you’re about to dip into your safety buffer.

Reorder Point = (Average Daily Sales × Lead Time) + Safety Stock

Continuing the worked example:

This is the single most under-used number in ecommerce operations. Most brands react to stockouts; the disciplined ones react to reorder points. The difference is weeks of cover and thousands of dollars in lost sales.

3. Order Quantity — how much fuel to pump

The textbooks will throw Economic Order Quantity at you — a formula that tries to balance holding costs against ordering costs to find the mathematically optimal batch size. For most SEA marketplace brands, that’s overkill and a little disconnected from reality. Your real constraints are usually supplier MOQs, container/shipment sizes, working capital, and warehouse space — not abstract inventory-theory tradeoffs.

Here’s the version that actually works for operators:

Order Quantity = (Forecast for next cycle + Safety Stock) − (Current Stock + In-Transit Stock)

In plain language: Work out how much you expect to sell until the next time you'll place an order, add your safety buffer, and subtract whatever you already have on hand or in transit from your previous order.

Three numbers. Three questions. One discipline.

What about MOQs that don’t match? If your supplier’s minimum order is 1,000 units but your math says 480, you have a real decision to make: order 1,000 and accept 5+ weeks of excess stock, negotiate a smaller MOQ, or find a supplier whose minimums match your velocity. Don’t pretend the math says 480 and order 1,000 anyway “just to be safe.” That’s how dead stock happens.

How to allocate inventory across multiple marketplaces

This is where it gets tricky. You have one pool of stock and three (or more) channels that need it.

Option A: Unified pool, real-time sync. (RECOMMENDED)
Keep all inventory in one warehouse. Use an inventory sync tool (OneCart, BigSeller, SiteGiant, or similar) to update stock levels across Shopee, Lazada, and TikTok Shop in real time. When an order comes in on Shopee, stock decreases everywhere.

• Pros: Simple, less capital tied up, no risk of one channel sitting on excess while another stocks out.

• Cons: If sync breaks during a mega sale (and it does), you oversell. Need reliable middleware.

Option B: Allocated pools by channel.
Pre-assign specific quantities to each marketplace. Shopee gets 40%, Lazada 30%, TikTok Shop 20%, own site 10%.

• Pros: Protects each channel from the others. Easier to manage platform-specific fulfilment (e.g., FBL for Lazada).

• Cons: Capital-inefficient. One channel may stock out while another sits on surplus.

Option C: Hybrid — unified base + channel buffers.
This is what most mature operators land on. Keep 70% of inventory in a unified pool with real-time sync. Allocate 30% as channel-specific buffers for platforms that need pre-positioned stock (like Lazada FBL warehouses) or for planned campaigns.

The right answer depends on your scale, your warehouse setup, and which platforms require pre-positioned inventory. But the principle is: synchronise what you can, buffer what you must.

What happens when goods don’t sell

Let’s talk about the thing nobody wants to admit: dead stock.

Dead stock is inventory that has stopped moving and has little likelihood of selling at full price in the future. It’s the silent killer of ecommerce cash flow.

For context: fashion brands have 44% of retailers reporting excess stock, with unsold merchandise accounting for 17-20% of total inventory. In beauty, roughly $50 billion goes to waste annually.

The real costs of dead stock:

• Storage costs — warehousing isn’t free, and in SEA where 3PL costs are per cubic metre per month, dead stock is paying rent for doing nothing

• Opportunity cost — capital locked in dead stock can’t be invested in hero SKUs or marketing

• Markdown erosion — when you eventually discount to clear, you train customers to wait for sales

• Platform impact — slow-moving SKUs hurt your store metrics and search ranking on Shopee and Lazada

What operators actually do to manage dead stock:

1. Bundle it. Pair slow-movers with hero SKUs in “value sets.” This works especially well for beauty and personal care.

2. Gift with purchase. Turn dead stock into a perceived benefit. A slow-selling travel-size product becomes a compelling GWP.

3. Flash clearance. Dedicate a campaign slot to clearing old stock, but do it on a separate listing or under a “clearance” banner to protect your main brand pricing.

4. Liquidation channels. Platforms like Komento in Asia help brands move excess inventory to non-core markets discreetly. In beauty, Highstock serves a similar function globally. The key is not letting dead stock leak into your primary marketplace at discounted prices.

5. Donate and write off. If the cost of holding exceeds the recovery value, sometimes the smartest financial move is to donate (for tax benefit where applicable) and free up the capital and warehouse space.

Planning for mega sales without blowing up your cash flow

This is the section every SEA operator needs to read twice.

In Southeast Asia, the double-digit sales (9.9, 10.10, 11.11, 12.12) can represent 40-60% of annual revenue. That concentration creates enormous pressure on inventory planning — and the brands that win these sales have already done most of the hard work eight weeks before the sale actually starts.

The 30-50% rule: For TikTok Shop during mega sales, keep 30-50% buffer stock above your baseline forecast. Viral moments during live selling can blow through your forecast in hours. Better to have stock left over than to miss the spike.

Cash flow hack: Negotiate supplier payment terms that align with your mega-sale cash cycle. If you pay suppliers net-30 and mega-sale payouts arrive in 14-21 days, you can use sale proceeds to cover a significant portion of your inventory cost.

Templates for your team

Here’s what a working forecasting and inventory planning setup looks like for an operator managing across SEA marketplaces. I’ve built four templates with sample data so you can see how they work in practice.

Before we go through them one by one, here’s how they connect. Each template answers a different question, but they feed into each other.
Template 1 tells you what you expect to sell.
Template 2 tells you when to reorder.
Template 3 tells you how to scale for mega sales.
Template 4 tells you what’s not selling and needs to be cleared.

Template 1: SKU-Level Demand Forecast

Fill by SKU × Channel × Market. Granularity = real decisions.

How to read this: For every SKU, every channel, every market, take last year’s monthly average, apply a realistic growth rate, then multiply by your campaign uplift factor (for example, 2.5× during 11.11). That gives you your adjusted monthly forecast. Do this once per quarter, review monthly.

Template 2: Reorder Point Tracker

“Days of Cover” = (Current Stock + In Transit) ÷ Avg Daily Sales. If it drops below Lead Time days — order now.

How to read this: Every SKU has a reorder point calculated from the formulas above. The tracker tells you at a glance which SKUs are safe, which are approaching the trigger, and which you need to order today. The Action column is what your ecommerce lead looks at first thing every Monday morning.

Template 3: Mega-Sale Inventory Planner

Work backwards from sale date. If 11.11 is target & 6-week lead time → order by late September.

How to read this: Start with what you normally sell per week, multiply by your expected mega-sale uplift (usually 2–4×), add a 30–50% buffer, then calculate your order-by date by counting backwards from the sale date using your supplier lead time. The Status column tells you whether you're on track, behind, or already ordered.

Template 4: Dead Stock Audit

Run monthly. 0 sales in 60+ days = on this list. <1 unit/week in 90 days = action candidate.

How to read this: Every month, list every SKU that hasn't moved in 60+ days, how much capital is locked in it, and the recommended action — bundle, gift-with-purchase, liquidate, or donate. This is the hardest discipline for most brands because it forces you to admit which bets didn't work. Do it anyway.

How to operationalise this across your team

Here’s the part that separates brands that talk about inventory management from brands that actually do it well.

You can have the best forecast model in the world, but if nobody reviews it, nobody acts on the reorder triggers, and nobody reconciles forecast versus actual — it’s just a pretty spreadsheet.

Most ecommerce teams I work with especially the ones at Commerceplus are built small because we want everyone to be directly involved in making decisions. Maybe 3-8 people covering everything from content to ads to customer service to logistics. There’s no dedicated “demand planner.” So the question becomes: who does what, how often?

Here's the cadence that works for most lean marketplace teams:

The golden rule: One person owns the inventory number. Not a committee. Not a shared responsibility. One person who can see across all channels, all markets, all SKUs — and who has the authority to place a purchase order without waiting for three rounds of approval.

In practice, this is usually the ecommerce lead or the operations manager. In smaller teams, it might be the founder. The title doesn’t matter. What matters is that when a hero SKU hits reorder point on a Tuesday, someone places the order on Tuesday — not the following Monday after a team meeting.

If you’re working with an enabler or agency: make sure they have visibility into your inventory levels. I’ve seen too many situations where the marketing team is scaling ads on a product that’s already at 5 days of cover. The enabler doesn’t know because they only see the ad dashboard, not the warehouse dashboard. That disconnect costs money every single time.

The honest bottom line

Here’s what I’ve learned from watching brands navigate this:

The brands that manage inventory well are not smarter. They’re more disciplined.

• They review their forecast weekly, not monthly.

• They track by SKU, not by category.

• They plan by channel, not in aggregate.

• They kill losers early instead of hoping they’ll sell.

• They hold their nerve on hero SKUs and keep them in stock even when cash is tight.

• They treat mega sales as inventory events, not just marketing events.

• And they have someone — one person — who owns the inventory P&L and can override both the marketing team’s optimism and the finance team’s conservatism.

That’s it. It’s not complicated. It’s just hard to do consistently when everything else is on fire.

If you’re running a brand across SEA marketplaces and this resonated, do one thing this week: pull your top 20 SKUs. Calculate their days of cover. Check if any hero SKU is below reorder point. If it is, place the order today.

This two-part deep dive is part of the Ecommerce Owner's Manual series.
For the 2026 Southeast Asia eCommerce Marketing Calendar, go here: Marketing Calendar.
For SKU economics and portfolio management, revisit The Great Middle Market Hollow-Out: Middle Market Hollow-Out.
For full-funnel orchestration across platforms, see Thinking Full Funnel: Thinking Full Funnel.

Not because it’s boring. Because it’s a really difficult and sometimes embarrassing topic.

Inventory.

Specifically: forecasting demand, planning stock, and managing the painful reality of what happens when you get it wrong. And in Southeast Asia — where you’re selling across Shopee, Lazada, TikTok Shop, maybe your own site, across Malaysia, Singapore, the Philippines, Thailand, and sometimes Indonesia — “getting it wrong” is almost the default.

In Southeast Asia’s marketplace-first world, the brands that win are not the ones with the best ads. They are the ones that never run out of the thing the customer came to buy.

We’ve sat in discussions where the marketing team is celebrating a 5× ROAS campaign while the ops team is quietly panicking because the hero SKU sold out in 48 hours and the next shipment is six weeks away. We’ve watched brands launch into 11.11 with half their best-sellers already on backorder. We’ve seen teams order three months of stock for a TikTok-viral product that stopped trending in nine days. We’ve just onboarded a new client and see their Hero SKU going out of stock and the next batch of inbound only comes 3 weeks later but the growth team is pressed for topline growth still.

This is the stuff that actually determines whether a brand makes money or just makes noise.

Sound familiar? It should. As I explored in The Great Middle Market Hollow-Out, most brands don’t even track SKU-level economics properly, let alone forecast at the SKU-by-channel-by-market level. And in Building Beyond Marketplaces, we talked about the 80/20 hybrid model — but that model falls apart if you can’t keep the right products in stock across the right channels.

So let’s dig into it.

The real cost of getting inventory wrong

Before we talk about how to forecast, let’s talk about what happens when you don’t. These are some publicly available case studies that I can pick up. (If you have a more local example, feel free to share it over, but these cases do help make a point)

These aren’t edge cases. They are what happens when demand planning is treated as someone else’s problem.

• Nike (2001): Installed a demand-planning system without proper testing. The software overordered slow-selling shoes and underordered Air Jordans. Direct result: $100 million in lost sales. Stock dropped 19.8% in a single day. The lesson wasn’t “don’t use software.” It was “don’t deploy a system you don’t understand at a scale you can’t manage.” (Source: Henrico Dolfing)

• Adidas (2022-2023): The Yeezy partnership collapse left Adidas sitting on an estimated $500 million worth of unsold inventory. Potential revenue hit: $1.3 billion. Gross margin declined 47.3% in 2022 after supply chain cost increases and aggressive discounting to clear stock piles. This wasn’t just a celebrity problem — it was an inventory concentration risk that the demand planning function didn’t hedge. (Source: CNBC)

• ASOS (2019): A new inventory system was successfully sending out orders but wasn’t recording returns — for an entire week. With 85,000 SKUs, the compound error was massive. Profits dropped 68% year-over-year. (Source: Khaos Control)

• Peloton (2021-2022): Invested heavily in manufacturing capacity to meet pandemic-era demand, then found itself with warehouses full of equipment nobody wanted. Wrote down $9.1 million in Guide inventory alone. Stock price cratered. The demand forecasting failure: they planned for a trend, not for the reversion. (Source: Pelo Buddy)

• Ralph Lauren (2016-2018): A 15-month inventory lead time meant the brand couldn’t react to trends, held too much excess stock, and saw inventory grow 26% over three years while sales grew only 7%. Result: profits fell 50%. The fix required cutting product lines, closing stores, and rebuilding the planning function from scratch. (Source: Khaos Control)

And here’s the broader picture: stockouts cost retailers nearly $1 trillion worldwide annually, according to HBR. The fashion industry produces around 100 billion garments per year, and an estimated 20-30% — up to 30 billion pieces — go unsold. In beauty, only 50% of products produced in a given year actually sell in that year, with roughly $50 billion going to waste annually. (Sources: HBR, BeautyMatter)

The common thread? Overstock and stockout look like opposite problems. They’re not. They are the same failure — a failure of forecasting — showing up at different points in time.

Why forecasting is harder across multiple marketplaces

Now layer on the complexity of Southeast Asia.

You’re not forecasting for one channel. You’re forecasting for Shopee, Lazada, TikTok Shop — possibly across four to six countries — each with different demand patterns, promotional calendars, fee structures, and buyer psychology.

• Shopee is a high-velocity, promo-driven beast. Flash sales, mega campaigns (9.9, 11.11, 12.12), and aggressive free-shipping promotions create massive demand spikes that can drain your inventory in hours. Between campaigns, velocity drops. The pattern is: spike, trough, spike, trough. Forecasting based on averages is useless here.

• Lazada positions itself more premium, with a heavier emphasis on voucher mechanics and fulfillment (Fulfilled by Lazada). The demand pattern is steadier but driven by different levers — basket-building rather than impulse. Inventory allocation here needs to account for warehouse lead times, not just marketing calendars.

• TikTok Shop is the wildcard. Demand is creator-driven, viral, and fundamentally unpredictable. Nearly 70% of sellers report experiencing stockouts or shipping delays due to unanticipated demand surges. You can go from 10 units a day to 1,000 in 48 hours because a creator made a video. And then back to 10 a week later.

Each platform has different demand DNA. One forecast cannot serve three channels.

And then you add the regional calendar.

As I outlined in the 2026 Southeast Asia eCommerce Marketing Calendar, success in this region means planning across three overlapping calendars: platform mechanics (mega sales, payday cycles), cultural triggers (CNY, Ramadan, Hari Raya, Deepavali, Christmas), and business realities (inventory lead times, cash flow cycles). If you’re only planning for 11.11, you’re leaving months of compounding growth on the table — and you’re probably either overstocked or understocked for every other moment.

Two approaches to forecasting: what you have versus what you don’t

Here’s a simplified but usable thought process and approach for you when it comes to demand planning and forecasting.

If you have historical data (12+ months)

This is your foundation. Quantitative forecasting uses past sales, seasonal trends, campaign performance, and lead-time data to build forward-looking projections.

The framework I like — borrowed from Common Thread Collective and adapted for marketplace operators — is what they call the “Three-Layer Revenue Cake”:

• Layer 1 — Existing customers (the base). How many customers from last year will buy again? At what frequency? At what AOV? This is your most predictable revenue.

• Layer 2 — Owned audiences (the middle). Email, SMS, WhatsApp, organic social. These are people who know you but haven’t converted or haven’t repurchased. What’s the conversion rate per touch? How many touches are you planning?

• Layer 3 — Paid acquisition (the top). New customers from ads, affiliates, creators. This is the most variable and the most expensive. What’s your planned spend? What’s your expected CAC and first-order AOV?

Each layer produces a demand signal. Roll them together by SKU, by channel, by market — and you have a working forecast.

But here’s the operator truth: the model is only as good as its inputs. And in Southeast Asia, the inputs are messy. Shopee and Lazada backend data exports are inconsistent. TikTok Shop analytics dashboards are limited. Cross-platform data rarely reconciles cleanly. Most brands I talk to are still forecasting in spreadsheets — and honestly, for brands doing under $5M in annual GMV, a well-structured spreadsheet might still be the right tool.

If you’re a new brand (no historical data)

This is where it gets interesting — and where most advice falls apart, because it assumes you have data you don’t have.

For new brands, forecasting is qualitative. You’re working with:

• Category benchmarks. What does a comparable brand sell per month on Shopee MY in your category? Talk to operators. Ask in founder groups. The data exists in aggregate form through syndicated tools,

  or you can also talk to enablers like Commerceplus to get a sense of the size of the marketplaces sales volumes.

• Competitor sell-through signals. Look at competitor listings on Shopee and Lazada. The “X sold” counter gives you a rough volume proxy. Check multiple times over a few weeks to estimate velocity.

• Small-batch testing. This is the Shein model scaled down. Start with 50-100 units per SKU. List across your target channels. Let real demand teach you what to reorder. Don’t commit capital to 2,000 units of something you’ve never sold before. Shein launches 2,000-5,000 new SKUs daily but only produces 50-100 units per SKU initially. If it sells, they reorder instantly. If it doesn’t, they kill it. The entire design-to-shelf cycle: 7-14 days. (Source: Logistics Navigators)

• Social proof and creator signals. If you’re doing a TikTok-first launch, the engagement on seeding content is a leading indicator of demand. High save rates, comment-to-view ratios, and “where to buy” comments are demand signals before they become orders.

• Pre-launch waitlists. Simple but effective. A landing page or WhatsApp group collecting interest before launch gives you both a demand proxy and a first-day conversion pool.

The honest truth: for new brands, your first forecast will be wrong. That’s fine. What matters is the speed at which you learn and adjust. Build a 4-week review cycle. After each cycle: what sold faster than expected? What didn’t move? What’s the reorder velocity? Adjust and repeat.

The real inventory challenge: your factory, your distributor, your lead time

I know the Shein and Zara supply chain stories are popular. And yes, Zara’s 2-4 week design-to-shelf cycle and Shein’s 50-unit micro-batches are genuinely impressive. But I’m not going to spend time on them here — because if you’re reading this newsletter, you’re probably not running 300 factories in Guangzhou or a vertically integrated European production machine.

You’re more likely dealing with one of these situations:

• You’re a brand that manufactures your own products — maybe skincare, supplements, or packaged food — and your production lead time is 2-6 weeks depending on batch size, raw material availability, and lab testing or certification requirements.

• You’re working with a master distributor who holds the stock, and your reorder depends on their warehouse availability and willingness to allocate to your marketplace channels.

• You’re sourcing from China or regional factories, with 4-8 weeks production plus 2-4 weeks shipping, and you’re placing bets today on demand that won’t materialise for 2-3 months.

Each of these has a fundamentally different inventory rhythm. And the core question is always the same:

How fast can products reach my warehouse — and how fast can I expect to sell them?

Everything else — safety stock, reorder points, allocation splits — is downstream of those two variables. And the honest truth is that most brands I work with don’t have the luxury of fast reorder cycles. If your production lead time is 8 weeks and your shipping is another 3, you’re making bets 11 weeks out. That’s not a forecasting problem — that’s a capital allocation problem dressed up as operations.

If you work with a distributor: Your biggest lever isn’t forecasting — it’s the relationship. How much visibility does your distributor give you into their stock levels? Can you get priority allocation before a mega sale? Do they hold buffer stock for your marketplace channels, or do you compete with their offline retail for the same pool? I’ve seen brands lose 11.11 not because demand was wrong, but because their distributor allocated the stock to a department store chain instead.

If you manufacture yourself: Your lever is batch planning. Can you split production into smaller, more frequent batches instead of one large quarterly run? Smaller batches cost slightly more per unit but dramatically reduce overstock risk. The Shein principle — minimise the cost of being wrong — doesn’t require 300 factories. It just requires the willingness to order 500 units instead of 5,000 and reorder fast when something moves.

If you’re importing: Your lever is pipeline visibility. Map every stage — factory confirmation, production start, QC, shipping, customs, last-mile to warehouse — and build your reorder triggers around the total pipeline time, not just the production time. Most brands underestimate customs clearance by 1-2 weeks. That gap alone can turn a well-planned launch into an out-of-stock disaster.

But “minimising the cost of being wrong” looks different depending on your production reality. For a brand with a 2-week turnaround, it means small batches and fast pivots. For a brand with a 10-week pipeline, it means better forecasting inputs, larger safety buffers on hero SKUs, and the discipline to kill slow movers early enough to free up cash for winners.

What this means for Southeast Asia operators

Here’s my honest take as someone who has watched dozens of brands navigate this.

Most SEA ecommerce brands are not struggling because they don’t understand forecasting theory. They’re struggling because:

1. The org isn’t built for it. Demand planning sits awkwardly between marketing (who controls spend and campaigns), operations (who manages stock), and finance (who controls cash). Nobody owns the full picture.

2. Lead times are real. If you’re sourcing from China, your manufacturing lead time might be 4-8 weeks, plus 2-4 weeks shipping, plus customs clearance. That means you’re placing orders today for demand that won’t materialise for 2-3 months. In a market where TikTok can change everything in 48 hours, that’s a structural mismatch.

3. Cash is the constraint. For most SME brands, the inventory problem is really a cash flow problem. You can’t afford to hold 3 months of safety stock across 4 markets and 3 platforms. So you under-order, and then you stock out during the one campaign that could have made your quarter.

4. Mega sales distort everything. In SEA, 40-60% of annual revenue can come from Q4 double-digit sales (9.9, 10.10, 11.11, 12.12). That means your forecasting has to account for demand patterns that look nothing like the rest of the year.

As I wrote in the marketing calendar piece, the best operators plan across platform mechanics, cultural triggers, and business realities simultaneously. Inventory planning is where all three calendars collide.

Over the weekend, I was in One Utama, a large shopping mall in KL and was reminded by my wife to get a replacement bottle of body shampoo from Bath & Body Works Body Shampoo. However, I quickly replied her saying that we probably can get a good deal online and have it shipped over for free.

Later that night, I checked on Shopee and Lazada and TikTok Shop as well and realize this brand was not on any of the marketplaces. I was puzzled, I could guess some reasons for not being on the marketplaces but still….. it just feels odd isn’t it?

Why isn’t this brand going harder on Shopee?
Why does that brand only pursue their own proper DTC store?
Why is another brand everywhere on marketplaces, but nowhere memorable?

On the surface, these sound like channel questions.

They’re not.

Because this region is marketplace-first by default. Consumers are comfortable discovering on social, comparing on marketplaces, and buying wherever the friction is lowest. That makes Shopee, Lazada, and TikTok Shop impossible to ignore. But it does not mean every brand should treat marketplaces as their flagship home.

That’s the part too many brand teams still miss. Marketplaces are brilliant at aggregating traffic. They are not automatically good at helping brands build pricing power, customer memory, or long-term loyalty. Those are different jobs.

As I wrote in Building Beyond Marketplaces, the smartest brands in the region are not choosing “marketplace versus DTC” in some binary way. They’re deciding what to outsource, what to orchestrate, and what to own.

The real divide is not online versus offline

The lazy take is that “good brands go DTC” and “mass brands go marketplace.”

Nice theory but this probably falls apart in about five minutes.

A better way to look at Southeast Asia is that brands tend to fall into four operating camps.

The first camp is what I’d call the DTC fortress.

These are brands that treat owned channels as strategic infrastructure, not a side project. UNIQLO is the cleanest example. It has country-specific sites across Malaysia, Singapore, the Philippines, and Thailand, and its app strategy is not cosmetic. The app bundles member offers, click-and-collect, inventory visibility, reviews, styling content, and event hooks into one owned experience. That is not just ecommerce. That is consumer control by design.

The second camp is the selective participant.

These brands are not anti-marketplace. They’re just careful about where they show up, what they sell, and how much of the experience they hand over. Bath & Body Works is a good example. It runs owned sites across Malaysia, Singapore, the Philippines, and Thailand, but it also maintains official marketplace presence in places like Lazada Philippines. That’s not confusion. That’s selective distribution.

H&M sits in a similar bucket, but with an important caveat: its marketplace footprint looks less clean-cut than people assume. Its Shopee Malaysia store URL and Lazada Malaysia shop still surface, but at the time of checking, it looks like all products are already delisted and set as OOS. So the right read here is not “H&M has exited Shopee,” but rather “H&M’s marketplace presence appears selective and inconsistent.” That distinction matters if you care about being right.

The third camp is marketplace-friendly premium.

These are brands that have decided marketplaces can work even when the brand is more considered, more curated, or more premium than average category fare. Lush is a nice example because it disproves the common assumption that premium beauty always avoids mass channels. Lush is present on Shopee Singapore, and Lazada Singapore. So this is not a brand that has rejected marketplaces on principle. It has simply chosen where the model works for it.

You can put MUJI and Nespresso in that broader zone too. MUJI shows up through official marketplace storefronts in Singapore and the Philippines. Nespresso has official presence on Lazada Singapore and Shopee Malaysia. The point is simple: premium does not automatically mean marketplace-averse. Sometimes it just means marketplace-disciplined.

The fourth camp is the one nobody talks about enough: digitally present, but not digitally built.

This is where a lot of legacy FMCG brands live.

Take the Campbell, Prego, and Swanson setup on Lazada Singapore’s RedMart pages. Or Campbell’s Malaysia product page, which gives you product-level brand presence without telling a strong direct-commerce story. Or the various Campbell/Kimball-related pages surfacing on Lazada Malaysia. These brands are clearly online. But being online is not the same thing as running ecommerce as a core brand capability. Often what you’re seeing is retailer-led digital distribution, not a fully intentional owned-commerce system.

And that distinction matters more than most teams realise.

Some brands are not “avoiding DTC.” They are just not built for it yet

A lot of incumbent brands tell themselves a flattering story.

They say they are “channel agnostic.” or…
They say they are “letting the market evolve.” and let resellers handle the selling…
They say they are “focused on winning where the consumer shops.”

Sometimes that is true.

Sometimes it is just corporate poetry for: we do not have the org structure to run ecommerce properly.

This is where Bain is useful. Its diagnosis of incumbent consumer brands is brutally familiar: the biggest blockers are ownership, expertise, and agility. In plain English, nobody really owns the online P&L, the ecommerce team is underpowered, and decisions that should take a day take three weeks because they bounce between marketing, trade, sales, finance, supply chain, and distributors.

That sounds abstract until you’ve lived it.

Marketing wants better brand content.
Trade wants promo compliance.
Key accounts want retail harmony.
Finance wants margin protection.
Supply chain wants forecast stability.
Distributors want channel priority.
And ecommerce gets shoved in the middle and told to “make it work.”

That’s how you end up with half-built storefronts, inconsistent pricing, weak PDPs, stale assortment, and a team that spends more time negotiating internally than serving the customer.

The blocker, in other words, is often not consumer demand. It is org design.

DTC is not a “must pursue” sort of strategy. It only works when the company is clear on what direct commerce is actually supposed to do. Is it there to drive incremental revenue? Generate first-party insight? Create a better brand experience? Support omnichannel sales? If you cannot answer that clearly, building a direct-sales channel usually creates more confusion than advantage.

That is why I’d be careful with the usual chest-thumping advice that every brand should “own the customer.” Sounds great on LinkedIn. Less great when you’re selling low-involvement pantry staples through fragmented distributor structures in four Southeast Asian markets with different price ladders, tax structures, and retail politics.

The unit economics can turn ugly very quickly

Even if the org is ready, the numbers still have to make sense.

And too often, they don’t.

The cleanest mistake brands make is looking only at the headline commission rate. What matters is the full take-rate stack: commissions, payment fees, shipping subsidies, platform campaigns, vouchers, affiliate costs, returns, staffing, content production, and the periodic ritual humiliation known as mega-sale discounting.

That stack adds up fast.

HKTDC’s marketplace case work in ASEAN notes that commissions can eat 30% to 50% of profit margin in certain marketplace selling models, with major sales events often pushing sellers toward 30% to 50% discounts just to stay competitive. That is not a rounding error. That is a business model issue. Source

The platform fee guides tell the same story from the ground level. Shopee Malaysia charges platform fees plus transaction fees plus fixed order fees. Lazada Malaysia layers commission and payment fees. TikTok Shop in Singapore, Malaysia, the Philippines, and Thailand all take their own slice through commission and transaction charges, before you even start talking about paid media or creator spend. None of this means marketplaces are bad.
It just means “sales growth” and “profitable sales growth” are very different hobbies.

This is also why the argument from Building Beyond Marketplaces still holds up: the goal is not to abandon marketplaces. The goal is to know when renting marketplace demand is cheaper and smarter than trying to force every sale through owned infrastructure — and when it isn’t.

Control matters more in some categories than others

If you sell detergent, instant noodles, or canned soup, convenience wins a lot of the time. People know what they want. They want to buy it fast. They do not need a sonnet about the founder’s philosophy before adding to cart.

If you sell fashion, beauty, wellness, gifting, or anything with identity wrapped into it, the job is different.

Now the customer may need education.
Or reassurance.
Or routine-building.
Or social proof.
Or a reason to care before a reason to click.

That is why category matters so much. Premium fashion and beauty brands tend to be more selective because the brand is not just the product. It is the context around the product. And marketplaces are efficient, but they are not always generous environments for nuance.

That is why the cleanest strategic question is not “Should we be on marketplaces?”

It is: Which moments in the consumer journey are too important for us to outsource?

The smartest brands map channels to moments

This is the part I think more brands in Southeast Asia need to internalise.

You do not need to own every transaction.
You do not need to force every customer through your app.
You do not need to become a DTC purist to be strategically sane.

But you do need to know where your advantage lives.

As I argued in Thinking Full Funnel and Orchestrating Shopee, TikTok Shop, FB & Affiliates, different channels do different jobs. Social drives discovery. Marketplaces reduce friction at conversion. CRM, community, and owned touchpoints do the heavy lifting on retention and loyalty. The brands that win are not the ones choosing one channel to worship. They’re the ones building a system.

That system usually looks something like this.

Marketplaces are excellent when the consumer already knows the product and wants speed, trust, and price comparison. They are also useful when you need regional reach quickly, especially in markets where building local owned infrastructure would be expensive, slow, or strategically unnecessary.

Owned channels matter more when education, retention, bundling, subscriptions, loyalty, or first-party data make the economics work better over time. BCG makes the point well: transaction-led DTC is often expensive and slow to pay back, but engagement-led DTC can be far more powerful because it helps brands learn faster, deepen relationships, and build repeat behaviour without treating the site purely as a checkout machine.

That is why the best operator question is not “How do we get more people onto our website?”

It is “Where do we make money, where do we build memory, and where do we learn something competitors cannot?”

Those are not always the same place.

So how should brands decide?

Here’s the straight answer.

If you sell replenishment-heavy, low-consideration products, marketplace-first is often the sensible default. Win search. Win availability. Win promo moments. Do not build an expensive direct-sales theatre production if the customer just wants the thing and wants it now.

If you sell products where ritual, education, identity, or service matter, you should probably own more of the middle and lower funnel than you currently do. Not because websites are glamorous, but because memory, margin, and customer understanding are built there.

And if you are a legacy brand sitting somewhere in the middle — online, but not really ecommerce-native — your first job is not launching a prettier storefront. Your first job is fixing ownership. Who owns the P&L? Who owns content? Who owns assortment? Who owns pricing logic? Who owns repeat purchase? If the answer is “sort of everyone,” then the answer is actually no one.

That middle zone is where a lot of Southeast Asian brands will win or lose the next five years.

Because the easy years are over.

Marketplaces are more crowded. Traffic is more rented. Content is more commoditised. Price competition is nastier. And AI is making mediocre creative cheaper by the hour. As I wrote in The Great Middle-Market Hollow-Out, the brands that survive this squeeze will be the ones that stop confusing distribution with defensibility.

The bottom line

Southeast Asia is not a “pick one channel” market.

It is a pick your moments market.

Use marketplaces when they give you reach, trust, and operational efficiency.
Use owned channels when the relationship is worth more than the immediate conversion.
And if you are not ready to run direct commerce properly, be honest about that too. There is no shame in not being DTC-first. There is only danger in pretending you are building a strategic channel when all you have really built is another place for the customer to bounce.

The strongest brands in this region are not asking whether Shopee, Lazada, TikTok Shop, stores, apps, and CRM should compete with each other.

They’re asking how each one should do its job.

That is the real shift.

Not marketplace versus DTC.
Not online versus offline.
Not brand versus performance.

Just this:

Rent the reach. Own the relationship.

That’s not the advice I would have given three years ago. Back then, the mental model was simple: if you want to sell in Singapore, you put stock in Singapore. Want to expand into Philippines? Find a 3PL in Manila, forecast carefully, and cross your fingers.

But that’s changed. And it’s changed faster than most brand owners realise.

The platforms — Shopee, Lazada, TikTok Shop — have quietly built logistics infrastructure that makes a local warehouse optional. At least for the first phase.

You can list on Shopee Philippines from Kuala Lumpur today. A Korean skincare brand can ship to Vietnamese buyers from Seoul. A Malaysian footwear brand can reach Singapore consumers without registering a single overseas entity.

I’m not saying warehouses don’t matter. They do, eventually. What I’m saying is that the old sequence — warehouse first, then sell — is backwards.

Prove demand first. Then warehouse.

THE MARKET OPPORTUNITY

The short version: intra-SEA cross-border eCommerce is heading toward US$14.6 billion by 2028. That’s roughly 2.8× where it was in 2023. And buyers who shop cross-border tend to spend about 21% more per order than buyers shopping from local sellers on the same platform.

I know. Stats are easy. But these ones match what we’ve been seeing on the ground.

Shopee Korea grew order volume 63% year-on-year in 2024 — 1.8× the platform’s global growth rate. Vietnam-bound K-beauty orders grew 134%. Thailand, 182%. Singapore, 58%.

And it’s not just beauty. Lazada added 20 million Korean SKUs to LazMall through a Gmarket partnership. The Tmall-Lazada bridge is moving Chinese brands into SEA at scale. SEA’s digital economy is projected to cross US$330 billion in 2025.

The demand is there. What’s missing for most brands isn’t access — it’s probably figuring out which model fits them. Here’s 3 models we see for the brands we have managed or spoken to.

THREE MODELS. AND THEY’RE NOT ALL THE SAME.

Model 1 — Platform Cross-Border Programmes

This is where I’d tell most Malaysian brands to start.

Shopee SIP (International Platform) and the Shopee Direct Programme let you sell into Singapore, Philippines, Taiwan, and Thailand from your existing Malaysian shop — without setting up an overseas entity, managing foreign payment systems, or worrying about customs documentation. The platform handles all of it.

Early SIP adopters reported 43× increases in daily orders and 60× jumps in daily GMV after enrolling. I’ll be honest — those numbers sound inflated, but they reflect what happens when you go from zero cross-border visibility to full in-market listing overnight.

The requirements aren’t daunting: at least one completed order per month, a cancellation rate below 10%, and a Shopee logistics partner enabled. Most active sellers already qualify.

Lazada’s LazGlobal covers six markets — Malaysia, Singapore, Thailand, Philippines, Indonesia, Vietnam — with no upfront fees. Commission runs 2–4% on sales. You ship to a consolidation hub; Lazada handles the cross-border leg. There’s a Fulfillment by Lazada (FBL) option for faster delivery and a local-seller badge, which makes a real difference in conversion.

TikTok Shop’s Cross-Border Engine is worth paying attention to, especially if you’re in beauty or lifestyle. The live commerce piece isn’t optional in certain markets anymore. In Vietnam, 31% of K-beauty orders on Shopee came from livestream purchases. Thailand, 26%. Malaysia, 20%. If you’re not building a live strategy into your cross-border launch, you’re starting behind.

Model 2 — International Brands Selling Into SEA via Platform

This one’s more relevant if you’re a brand from China, Korea, or Japan looking at SEA without wanting to set up local operations.

A Korean beauty house I know well entered Malaysia, Singapore, and Vietnam through this model, using creator seeding as the primary awareness engine before committing a single ringgit to local inventory. Smaller emerging brands on Shopee Korea grew by 2,000%+ in a single year. Take that number with context, but the direction is real.

For Chinese brands, the Tmall-Lazada bridge and Shopee’s China cross-border programme offer the fastest path to SEA buyers. A global pharmacy chain I’ve worked with uses this model for Singapore and Malaysia — their China-side fulfilment centre does the dispatch; the platform handles customs; the buyer gets next-day or 2-day delivery.

The playbook for this model is fairly consistent:

1. Validate demand via platform cross-border. Zero local capex.

2. Build search ranking, review volume, and follower base on-platform.

3. Find your top-3 SKUs per market — the ones that actually resonate.

4. Only then think about local inventory or FBL.

Don’t skip step 3. Most brands that struggle with cross-border skipped step 3.

Model 3 — Own-Brand D2C with Logistics Partners

This one’s more work. But if you want full control — over the customer relationship, the data, the brand experience, and the margin — it’s worth understanding.

The setup: Shopify Markets as your storefront, with Delivered Duty Paid (DDP) enabled at checkout so buyers see the true landed cost upfront. Then you plug in a cross-border logistics partner: Teleport (run by AirAsia) for first-mile from Malaysia into SEA, Ninja Van for broader coverage across 46 countries, POSLaju International for lighter shipments, or J&T for cost efficiency on volume.

Teleport’s model is interesting. They’ve improved delivery times from 4 days to next-day on certain routes using hybrid passenger-cargo flights. End-to-end service — pickup, sorting, mid-mile, customs clearance, last-mile. Ninja Van has been expanding aggressively, adding 44 new international shipping lanes in 2025.

This model suits brands that already have Shopify infrastructure, or those in categories where the platform marketplace experience feels wrong — specialty food, artisan goods, anything where brand context matters more than search ranking.

The honest trade-off: you own the customer relationship, but you own the complexity too. Returns, customs disputes, cross-border customer service — that’s all on you which requires more planning in general.

FIVE THINGS WE’VE FIGURED OUT (THE HARD WAY)

We’ve helped a footwear brand, a Korean beauty house, a fashion accessories label, a global pharmacy chain, and a Thai cosmetics brand navigate cross-border in SEA. Not as consultants handing over a deck and walking away — actually in the accounts, running the operations. Here’s what we know.

1. Listing quality is everything.

This sounds obvious. It isn’t obvious in practice.

Cross-border buyers are comparing your product against local options they already trust. If your hero image looks like it came from a product catalogue and your description hasn’t been touched since your domestic launch, you’re invisible. Localise the creative. Native-speaker descriptions — not machine-translated ones. Imagery styled for local aesthetic sensibility. Use-case copy that reflects how people in that market actually live.

The footwear brand we worked with saw e-commerce sales grow 43.9% in a recent quarter. A big part of that was the investment in digital asset quality and localised storytelling for their regional channel expansion. Not a single new product. Just better presentation.

2. Your margin has to work before you start.

Cross-border adds costs that stack up fast: platform cross-border fees, FBS storage and handling, currency conversion spread, higher return rates (typically 5–8% cross-border vs 2–4% domestic). If your domestic gross margin is sitting below 40%, don’t go cross-border yet. Fix the margin first. Going cross-border with a thin margin model just makes the problem bigger, faster.

3. Platform fulfilment is the conversion unlock.

The Malaysian footwear brand I mentioned above — when they switched from self-ship to Fulfilled by Shopee for their Singapore channel, conversion improved significantly. The “local seller” badge. The 3–5 day delivery promise. These are visible trust signals. Buyers see them before they read your product description.

FBS adds roughly 8–15% to your cost structure. But in most categories, the conversion lift offsets that within 60–90 days. It’s usually worth it.

4. Marketing doesn’t cross borders automatically.

What converts in Malaysia won’t necessarily convert in Vietnam or Thailand. Creator ecosystems are different. Platform ad mechanics are different. Purchase triggers are different. The Thai cosmetics brand we worked with didn’t achieve a 62% ROI increase on TikTok Shop by translating their Thai campaigns — they rebuilt the entire Malaysia-specific creator strategy from scratch.

This is the thing brands underestimate most. The product can cross borders. The marketing has to be rebuilt for each market.

5. The first 90 days aren’t about sales.

They’re about foundation. Follower growth rate. Review volume and velocity. Search rank for 3–5 target keywords. Repeat purchase rate by day 30.

If those metrics are moving in the right direction, you’re building something durable. If you optimise purely for promotional GMV in month 1, you’re building a discount dependency that’s very hard to unwind.

THE MARKETING GROUNDWORK

Most cross-border entries fail because the brand shows up in a new market with no awareness, no social proof, and no platform authority — then wonders why conversions are low.

I’ve seen this enough times that it’s become a checklist item now. Build before you launch.

Here’s the three-phase approach we run before any new market goes live.

Phase 1 — 60 Days Before Launch

Creator seeding is an important piece of the puzzle. For small targeted awareness against your potential buyers.

Find 15–25 micro-creators and affiliates in the target market. Followers between 5,000 and 200,000. Genuine audience alignment with your category. Send product at cost — not a paid collaboration, just product. Ask for honest reviews.

This isn’t about going viral. It’s about having social proof the moment your shop goes live. It’s about building a creator network you can actually activate during launch month. It’s about doing the social listening in parallel to understand how the category conversation runs in that specific market.

Set up your brand page on the target-market platform at the same time. Follow local creators. Start building playlist content. Don’t wait until launch week.

Phase 2 — Launch Month

• Pricing: 20% – 25% below your intended steady-state price. Think of this as acquisition cost, not discounting. Stack platform vouchers with seller vouchers to maximise the perceived deal.

• First 50 buyers: put a QR-code review card in every parcel. First 50 reviews will determine your next 500 buyers. This is one of those things that costs almost nothing and has an outsized impact.

• Live commerce: 2–3 sessions per week with local creators. In beauty especially, livestream purchase share is not a nice-to-have.
  Vietnam: 31% of K-beauty orders via livestream.
  Thailand: 26%. If you’re not doing live, you’re giving up a third of your potential orders.

• Platform ads: budget 15–20% of your monthly GMV target as ad spend. Search ads for high-intent keywords. In-feed or TopView on TikTok for awareness in beauty and lifestyle.

Phase 3 — 30 to 90 Days Post-Launch

These are the metrics that tell you if it’s working:

• Conversion rate: aim for ≥2%

• Return rate: keep it under 8%

• Repeat purchase rate: target >15% by day 30

• Store rating: hold at ≥4.7

Weak conversion? Look at listing quality, pricing vs. local competitors, and review volume first.

High returns? Usually a sizing or description accuracy problem. Or packaging that isn’t surviving the cross-border transit.

Low repeat purchase? Build a messaging sequence: order confirmation, delivery update, review request, 14-day repurchase nudge. It’s boring. It works.

The scale trigger: When a single market generates USD 20,000+ GMV for three consecutive months, that’s when you seriously evaluate a local fulfilment pop-up, or a proper warehouse partnership. Not before.

ARE YOU ACTUALLY READY?

Not every brand should launch cross-border right now. Here’s how I’d think about it honestly.

Go Now — if you have:

• 500+ reviews on your home market

• Store rating at 4.7 or above

• Gross margin above 45%

• Non-perishable, non-regulated product with no IP complications

• Domestic GMV already above RM 100k/month

Phase It — meaning do the prep work first, then go:

• Health, beauty, and supplements (check regulatory status in destination market before listing — it’s different in every country)

• Fashion accessories before apparel (return complexity is real)

• Brands with 4.5–4.7 ratings (run a domestic review campaign first)

• Gross margin 35–45% (do the pricing architecture work first)

Wait / Partner — structural issues to solve before cross-border makes sense:

• Cold-chain dependent products

• IP-heavy categories requiring local trademark registration

• Domestic GMV below RM 100k/month (fix your home market first)

• Gross margin below 35% (cross-border will make this problem worse, not better)

I know that last category is hard to hear. But launching cross-border with a weak domestic foundation is one of those things that looks like expansion and feels like distraction.

YOUR 30-DAY QUICK-START CHECKLIST

Shopee SIP / Direct Programme

☐ Audit top 10 SKUs for cross-border margin viability

☐ Verify eligibility (1+ orders/month, cancellation rate <10%)

☐ Apply for SIP via Shopee MY Seller Centre

☐ Brief local creator in target market

☐ Prepare localised hero images and product descriptions

☐ Enable Shopee logistics cross-border option

☐ Set intro pricing: standard price × 1.15 (absorb fees) × 0.80 (launch discount)

☐ Set 30-day review target: 25 minimum

Lazada LazGlobal

☐ Register on LazGlobal seller portal (no upfront fee)

☐ Submit business registration documents

☐ Upload product catalogue — minimum 10 SKUs to start

☐ Activate Fulfilment by Lazada (FBL) for top 5 SKUs

☐ Brief local creator(s) aligned to your category

☐ Allocate 15% of monthly GMV target as Lazada Sponsored Discovery budget

Shopify + Logistics Partners

☐ Enable Shopify Markets for target countries

☐ Activate DDP (Delivered Duty Paid) at checkout

☐ Integrate Teleport or Ninja Van API for cross-border shipping rates

☐ Set minimum order value for logistics cost viability (typically RM 80+)

☐ Build country-specific landing pages with local currency, sizing, and FAQs

☐ Set up post-purchase flow: order confirmation → tracking → delivery → review request

☐ Run creator seeding (10 creators minimum) before launch

LAST THOUGHT

I keep coming back to the same thing.

The infrastructure is there. The data is encouraging. The platforms have made it genuinely feasible for a brand in Kuala Lumpur to sell in Manila and Ho Chi Minh City without an overseas entity, a local warehouse, or a team on the ground.

But most brands I talk to are still waiting for the perfect moment. The moment when everything’s ready. When margins are perfect, when the product range is complete, when the team has capacity.

That moment doesn’t come.

What comes instead is 60 days of quiet creator seeding. A controlled launch with proper pricing. 90 days of watching the metrics and adjusting. And then — if the numbers support it — the warehouse conversation.

The warehouse isn’t the start. It’s the reward for getting everything else right first.

Last week we diagnosed the problem.

A mid-size beauty brand. Repeat rate dropped from 38% to 24% in 18 months. Hero serum undercut by three Shopee sellers at half the price. Gross margin collapsed from 62% to 47%. Ad spend climbed. Returns didn’t.

This week, we flip it.

Missed Part I? We diagnosed the three forces squeezing the middle market out of existence: the China supply shock, platform fee creep, and the AI content flood. → The Great Middle-Market Hollow-Out

Five brands — two Malaysian, two Indonesian, one Singaporean — who faced the exact same pressures and built their way out. Not through a lucky viral moment. Not through a VC war-chest. Through deliberate strategy, uncomfortable decisions, and a level of brand clarity most mid-market operators defer until it’s too late.

The categories differ. The moats differ. The pattern beneath all five is the same.

Here’s what they did — and what you can take back to your own business. I hope this research will help you as much as they will help us and inspire our team.

Why These Five?

Selection criteria matters. We’re not profiling brands that simply survived. We’re looking for brands that faced the specific forces from Part I — commoditisation from Chinese suppliers, platform fee compression, margin erosion, content noise — and made moves that produced a defensible outcome.

Four filters:

1. The brand faced real middle-market pressure — not a premium-only or budget-only play from day one

2. Revenue growth is verifiable — real numbers, not PR estimates

3. The moat is structural, not temporary — something that a Shein supplier or Shopee algorithm cannot easily replicate

4. The decisions were deliberate — not lucky accidents

Against those criteria, here are the five:

Let’s go deep on each.

Case Study 01 / Malaysia 🇲🇾

The Raw — The Science-First Moat

Founded: 2019
Revenue 2024: RM 50M (43% YoY growth from RM 35M)
Capital Raised: Zero

The Starting Point

Adibah Mazlan was a pharmacist. Not a beauty influencer, not a serial entrepreneur with a pitch deck.

She was a board-certified pharmacist who couldn’t find local skincare that met her own standards.

“Many were trendy but lacked substance,” she told Vulcan Post. “I saw a huge gap: Malaysians wanted affordable, effective skincare — but there was a trust deficit. I built The Raw to fill that gap.”

She started The Raw in 2019 with RM 30,000 of personal savings. No co-founders. No VC. No white-label shortcut. Just product formulations developed with university labs, nine-to-twelve month R&D cycles, and a single founding principle: real science, real results.

The Escape Move: Own a Credential That Can’t Be Faked

The Raw’s escape from the middle market was built on one structural asset — the pharmacist’s credential as a brand position.

In a market flooded with white-label serums, MCO-era influencer drops, and products backed by little more than an Instagram filter, a board-certified pharmacist building products from pharmaceutical-grade research is genuinely differentiated. Not because Malaysian consumers are unusually science-literate — but because in a trust-deficit market, “pharmacist-formulated” is a signal that cuts through noise in a way “clean beauty” and “K-beauty inspired” simply cannot.

This is the authority quadrant of brand building: you earn loyalty not through aspiration or community, but through the verifiable credibility of your knowledge base.

Content Strategy: Educate, Don’t Sell

The Raw’s early content looked nothing like a typical skincare brand’s feed.

This is a critical distinction. Every other brand was running discount campaigns and influencer drops. The Raw ran skincare education. Ingredient explainers. Myth-busting. The pharmacist’s guide to reading a label.

The platform reward: organic reach. The business reward: customers who buy because they understand why the product works, not because it’s on promotion. That produces a fundamentally different repeat-purchase dynamic — and a customer who is much harder to poach with a 20%-off competitor voucher.

Distribution as a Trust Signal

Getting into Guardian and KPJ hospital pharmacies wasn’t just a channel decision. It was a credibility statement.

“Retailers took us seriously because our numbers backed us up,” Adibah said. “We showed demand, repeat purchase rate, and a clear brand identity.”

The retail credibility reinforces digital trust. When your product sits next to prescription skincare in a hospital pharmacy, the authority positioning becomes self-evident to the consumer. Online and offline aren’t separate channels — they’re two parts of the same trust story.

The Numbers

• 2024 Revenue: RM 50M — 43% YoY growth from RM 35M

• Distribution: 800+ retail outlets (Guardian, AEON Wellness, KPJ hospital pharmacies)

• TikTok Shop ROAS: 12.9×

• TikTok Shop GMV Uplift: +99.9%; Orders uplift: +104.6%

• Startup Capital: RM 30,000 personal savings — bootstrapped, zero external investment

• IPO Target: 2028

• Expansion Pipeline: Indonesia, Brunei, Singapore, Middle East

The Escape Pattern

Authority moat. Own a credential competitors cannot buy, and communicate it with enough specificity that consumers can tell it’s real.

The diagnostic question: What does your founding team know — genuinely know — that no Chinese white-label supplier can replicate?

Case Study 02 / Malaysia 🇲🇾

Khairul Aming — The Creator Who Became the Supply Chain

Brand: Khairulaming (Sambal Nyet, Dendeng Nyet, Rendang Nyet)
Revenue 2024: RM 42M (110% YoY growth from RM 20M in 2023)
Single Stream Record (Feb 2026): RM 2.3M in 12 hours

The Starting Point - “Hello, What’s up guys?”

This is not a story about an entrepreneur who built a social media following.

This is a story about a content creator who understood his audience’s trust so completely that he transformed it — methodically — into a production business, a logistics operation, and a sales machine. In that exact order.

Khairul Aming built his following through cooking content. Patient, accessible, step-by-step Malaysian home cooking tutorials. His Ramadan recipe series became appointment viewing. He wasn’t a brand. He was a person you trusted in the kitchen.

When he launched Sambal Nyet Berapi in 2021 (right smack in the middle of Malaysia’s MCO/ Circuit Breaker), that trust translated into immediate demand. The product sold out within minutes. Demand vastly outstripped supply — and he recognised this as a problem, not a celebration.

The Escape Move: Audience First. Product Second. Supply Chain Third.

KA’s business model is the inversion of the standard e-commerce playbook.

Most brands build a product, then struggle to build an audience. KA built the audience first, then built the product — and then built the supply chain to deliver it at scale.

The result is a business where 70% of revenue comes directly from Shopee and subsequently TikTok Live. No Shopee margin compression. No Lazada ranking politics. A direct line from creator to consumer, mediated entirely by trust.

This is the creator-commerce model in its purest form: the audience is the distribution channel.

The Social Commerce Numbers (2024–2026)

The data is remarkable by any regional standard:

• 2024 Sales: RM 42M — 110% YoY growth from RM 20M

• 2023 Sales: RM 20M (up from RM 14M by mid-2022 — first million bottles)

• Feb 2026: RM 2.3M in a single 12-hour TikTok Shop live — Malaysia’s all-time platform record

• 5.42M views on that single stream; 51,710 products sold

• Malaysia Book of Records, 2025: RM 1M in 19 seconds

• Previous record he held himself: RM 1M in under 4 minutes (2024)

• Team: 100 staff, 2 factories in Kelantan, 3rd factory under construction

• Live economics: “An hour of livestream can generate RM 50K, RM 100K, or RM 500K — depending on planning and execution.”

The Trust Architecture

What makes KA’s model genuinely defensible isn’t his follower count. Follower counts are rented. Trust is built.

KA’s trust architecture rests on three consistent behaviours that are rare in Malaysian brand-building:

Radical Transparency in Crisis. When the initial Sambal Nyet launch had logistics failures — leaking jars, damaged products — KA didn’t issue a statement. He reached out personally, explained what went wrong, and detailed what the team was fixing. The affected customers became more loyal, not less.

Founder Proximity. KA hosts all the livestreams himself. He answers questions in real-time. He brings customers inside the business — factory visits, packaging changes, supply constraints. Customers feel like insiders, not consumers.

Staff Generosity as a Public Brand Signal. When KA closed both factories for five days to take his entire team to Penang, it became a public story. When he handed out RM 100 envelopes for Hari Raya, it was news. This generosity is woven into the brand narrative — and it reinforces what customers believe about him: that he’s a good person running a business with genuine values. You can’t buy that narrative. You have to earn it.

The Escape Pattern

Community trust as distribution. Build the audience before you build the product — or if you already have the product, start building the audience now.

The diagnostic question: If your brand disappeared from Shopee and TikTok Shop tomorrow, would anyone notice? Would anyone come looking? If the honest answer is no, your brand does not yet have an audience. It has a traffic problem.

Case Study 03 / Indonesia (SEA) 🇮🇩

Skintific — The Localisation Sprint

Origin: Chinese brand (Guangzhou Feimei) engineered for SEA
SEA Sales: US$114M+ cumulative in 2 years
TikTok Shop: 2022 Global Sales Champion
Shopee/Lazada Rank: #3 in SEA skincare (behind SENKA and L’Oréal)

The Starting Point

Skintific entered Southeast Asia with a thesis that turned out to be right: local brands had a trust advantage that international brands lacked, but no one had actually engineered a Chinese product to solve specifically local problems.

This sounds like a small distinction. It isn’t.

The Escape Move: Solve the Problems Local Brands Ignored

SEA’s climate is genuinely different from China’s. Hot, humid, tropical — and the consumer problems that flow from that (acne caused by humidity, oil control, sunscreen compliance, skin barrier repair in high-UV environments) were systematically underserved by both Chinese imports and legacy local brands that had been formulating to older standards.

Skintific’s “Refreshing Acne Removal” line, formulated specifically for tropical climate acne and oiliness using natural plant ingredients suited for heat and humidity, wasn’t a marketing story. It was a product truth.

The four pillars of their localisation playbook:

1. Local pain points first. Formulate for what the market actually experiences — not what the home market assumes.

2. Platform-native commerce. TikTok Shop + Shopee + Lazada simultaneously, with a heavy creator network in each market. They became #1 on TikTok Shop before they entered retail — the reverse of how Western brands typically sequence their channel strategy.

3. Celebrity-anchored trust. Nicholas Saputra for the Indonesian “5× Barrier Repair Cream” campaign. One deal. One familiar face. National credibility for a brand that was, technically, foreign. The question they answered wasn’t “Is this a good product?” — it was “Can I trust this brand?”

4. Price architecture. RMB 40–60 (US$6–8.50). Not cheap. Not premium. Precisely positioned as “premium healthy skincare” — above the commodity tier, below the imported luxury tier. This price point avoided the race to the bottom while staying within reach of the mass middle market.

The Competitive Reality

Skintific’s US$114M in two years wasn’t won by outspending competitors on ads. It was won by out-thinking them on product-market fit.

The China brands that struggled in SEA tried to import Chinese consumer preferences. Skintific imported Chinese manufacturing capability and applied it to SEA consumer needs.

The local brands that lost ground to Skintific — including several with more heritage and larger distribution networks — had products that solved yesterday’s problems and marketing that assumed consumer loyalty was permanent.

Neither assumption held.

The Escape Pattern

Cultural specificity as competitive advantage. Localise the product before you localise the marketing.

The diagnostic question: If a well-funded Chinese manufacturer decided to target your exact consumer segment tomorrow, what specific product truth would they have to spend years learning to replicate?

Case Study 04 / Indonesia 🇮🇩

Somethinc — The Community Engine

Founded: 2019 SKUs by 2024: 200+ (from 3 serums)
TikTok Live: #1 beauty category Q1 2023, #2 overall platform Indonesia
Local Market Share: >50% of Indonesia local beauty market
Single campaign: IDR 10.75B in 15 days

The Starting Point

In 2019, Somethinc launched with three serums. By 2024, it had over 200 SKUs and more than half the Indonesian local beauty market.

That trajectory is not organic. It’s engineered.

The engineering was done in TikTok Live studios, with nano-influencer networks, through tiered pricing logic, and across every major platform in Indonesia simultaneously. Somethinc didn’t wait to become a brand before building distribution. It built distribution infrastructure as the brand was forming.

The Escape Move: Live Commerce as the Primary Revenue Engine

Most beauty brands treat TikTok Live as a promotional channel — a boost during campaigns, a supplement to Shopee sales.

Somethinc treated it as the primary revenue engine from the beginning.

The cadence — approximately four livestreams per day — is not a content strategy. It’s a sales operation. With scripts, products staged by profit margin, hosts trained for conversion, real-time inventory management, and an affiliate network that extends the reach of every session without scaling the fixed cost.

In TikTok Live commerce, the consumer isn’t asking “Is this product effective?” , they’re asking “Does this feel exciting?” and “Does this brand reflect who I am?”

Somethinc’s live content is engineered for both. Entertainment-first, identity-confirming, with the product embedded naturally — not pitched.

Four Strategic Bets

1. TikTok Live-first commerce. Built the live commerce muscle before retail. Audience-first sequencing, the same playbook as KA — but for a beauty category with hundreds of competitors.

2. Tiered pricing for market capture. Affordable entry SKUs that brought in first-time buyers, feeding into a premium line that protected margin. This allowed Somethinc to claim >50% local market share without sacrificing the ability to scale gross margin over time.

3. Two-tier influencer strategy. International K-pop partnerships (Han So Hee, NCT Dream) for aspirational brand positioning and PR reach. Nano-influencers (hundreds of micro-creators) for authentic, high-conversion local buzz. Neither tier alone would have built the brand they built.

4. Full omnichannel coverage. Shopee, TikTok Shop, Tokopedia, Lazada, Blibli, and offline retail simultaneously. Present everywhere. Not dependent on any single platform. If Shopee’s algorithm changes tomorrow, Somethinc’s business doesn’t collapse — because it never was a Shopee business. It’s a Somethinc business that happens to sell on Shopee.

The Broader Lesson

Somethinc’s moat is community flywheel combined with product breadth.

The community flywheel: every live session creates engagement, which creates user-generated content, which creates social proof, which drives new viewers to the next live session. The loop is self-reinforcing, and it’s extremely hard to replicate without both the infrastructure and the content credibility.

The product breadth: 200+ SKUs means that a consumer who discovers Somethinc through one product has 200 reasons to stay. Low churn is structural when the SKU range covers every stage of a skincare routine at every price point.

The Escape Pattern

Community flywheel × platform diversification. Build the live audience and the omnichannel footprint at the same time — neither alone is defensible.

The diagnostic question: What percentage of your revenue comes from a single platform? If it’s above 60%, you are not a brand — you are a tenant.

Your Escape Roadmap

Map yourself to the brand closest to your situation.

If you’re a local Malaysian or Singaporean brand building for local market: Study The Raw. Your local knowledge — of local skin concerns, local climate, local cultural context, local regulatory landscape — is a genuine advantage over Chinese brands. Be more specific, not more broad. Own a credential. Communicate it with enough specificity that consumers can verify it themselves.

If you’re building through social commerce: Study Khairul Aming. The creator-commerce model isn’t exclusive to food brands with charismatic founders. The architecture — audience first, product second, supply chain third — applies to any category where trust drives purchase. You don’t need KA’s following. You need his approach: show your face, be transparent about your process, treat your customers like insiders.

If you’re a foreign brand entering SEA: Study Skintific. Don’t localise your marketing and leave your product unchanged. The brands that win in SEA solve SEA problems with SEA-appropriate formulations (or products), then communicate that specificity with local-market authenticity. Ask: what does this market experience that our home market doesn’t, and have we genuinely engineered a response to it?

If you’re a category leader defending market share: Study Somethinc. Build the live commerce flywheel. Diversify across platforms. Create the omnichannel footprint that makes you indifferent to any single platform’s algorithm change. And build a two-tier influencer strategy — international credibility at the top, authentic local conversion at the bottom.

If you’re an established brand that has lost its focus: Study Love, Bonito. The discipline of subtraction. Before you spend on acquiring new customers, audit your SKU range and brand positioning with radical honesty. What would you keep if you could only keep 40%? Cut the rest. Then scale what’s left with clarity of purpose.

Bonus Case Study / Singapore 🇸🇬

Love, Bonito — The Art of Subtraction

Shorter-form, because the lesson is more concentrated.

Founded: 2010 (as BonitoChico 2005)
Revenue 2023: S$88M (+37% YoY)
Target: S$100M + EBITDA profitability in 2025

The Problem: By 2021, Love, Bonito had raised US$36.5M, expanded into Hong Kong and the United States, faced a pandemic, generated 200+ customer complaints, and reported a US$8.1M loss. A cautionary tale about premature scaling in the making.

The Move: New CEO Dione Song made three decisions that were uncomfortable in the short term and correct over a three-year horizon:

1. Cut 60% of SKUs. Using ten years of customer data + machine learning, reduced overall styles by 60%, restructuring the range into three lines: Signatures, Staples, and limited capsule collections. Less confusion. More focus. Higher sell-through rates.

2. Fixed operations before re-scaling. Rebuilt logistics and warehousing before adding markets. 7% HQ staff reduction. Closed loss-making expansions.

3. Sharpened the brand’s singular proposition: The go-to destination for the everyday Asian woman. Not all women. Not all fashion. Asian women. On her terms. The 2025 rebrand — new logo (LB heart monogram), new palette (Bold Brown, Cool Pink, Heart Red), new tagline “Ready-to-live, not just ready-to-wear” — made this legible in every touchpoint.

The Result: 37% revenue growth in 2023 while cutting losses 27%. Partnership channel delivering 60% of revenue with 253% QoQ growth. S$100M and first EBITDA profitability in sight for 2025. Double-digit growth every year since 2020.

The Escape Pattern:
Brand clarity before scale. The middle market is often a consequence of trying to be everything to everyone. Cut what dilutes you before you scale what defines you.

The diagnostic question: If you could only keep 40% of your current SKU range, which 40% would you keep? Are you currently investing your marketing budget in that 40%, or spreading it across everything?

What All Five Have in Common

Different categories. Different countries. Different strategies. Same five moves beneath the surface.

Pattern 01: They Own Something Irreplaceable

The Raw owns a pharmacist’s credential. Khairul Aming owns two decades of audience trust. Skintific owns tropical-formulation expertise. Somethinc owns a live-commerce flywheel. Love, Bonito owns cultural positioning of Asian femininity on Asian terms.

None of these can be replicated by a Chinese manufacturer with lower COGS. None can be commoditised by a platform algorithm. They are genuine defensible moats.

Pattern 02: They Built Trust Before They Built Scale

The Raw educated before selling. KA built content for years before launching a product. Skintific solved a real local problem before running campaigns. Somethinc built the live audience before expanding SKUs. Love, Bonito fixed the customer experience before re-entering new markets.

In the middle market, brands typically do the reverse: scale fast, retrofit trust later. The brands that escape do it the other way around.

Pattern 03: Authenticity as Operating Infrastructure

This isn’t a marketing tactic. In each of these brands, authenticity is structural — it’s in the product decisions, the content strategy, the customer service response, and the founder’s public behaviour.

When authenticity is genuine, it compounds over time in a way ad spend cannot. You cannot buy KA’s Hari Raya generosity. You cannot replicate Adibah’s pharmaceutical credential. You cannot manufacture the trust Somethinc’s live community has built with its audience over thousands of hours of content.

Pattern 04: Uncomfortable Decisions Made Early

The Raw stayed bootstrapped in a market that celebrates venture funding. KA kept his product range to two SKUs for three years. Skintific priced above the commodity tier when it would have been easier to race to the bottom. Somethinc invested in live infrastructure before it was standard practice. Love, Bonito cut 60% of its product range when the business was bleeding.

All five brands made decisions that felt wrong in the short term and right over a three-year horizon.

Pattern 05: Platform Presence Without Platform Dependency

All five brands are active on TikTok, Shopee, and Lazada. None of them would collapse if one platform changed its algorithm or increased its take-rate.

The Raw’s authority comes from pharmacy distribution and direct customer education. KA’s power comes from his personal brand and his live audience. Skintific’s strength comes from its product formulation and creator network. Somethinc’s moat is its omnichannel reach and flywheel. Love, Bonito’s foundation is its physical stores and community.

The Bottom Line

The middle market is not disappearing because the economy is weak. Southeast Asia’s economies are growing at 4.5–5% annually. Private consumption is heading to US$5 trillion by 2035.

The middle market is disappearing because undifferentiated is not a viable position when Chinese suppliers can match your product at half the cost, platform algorithms commoditise your visibility the moment you stop spending, and AI-generated content has made it nearly impossible to cut through with generic brand storytelling.

The brands that escape don’t do it by being better at the same game. They escape by changing the game — owning an asset that can’t be replicated, building trust that can’t be bought, and making the kind of long-term decisions that most brands defer until the margin pressure becomes a crisis.

The Raw didn’t win because she had better ingredients. She won because she was a pharmacist who cared enough to explain why the ingredients matter.

KA didn’t win because he had a better sambal. He won because three million Malaysians trusted him before they ever tasted it.

Skintific didn’t win because it had better distribution. It won because it solved a problem no one else had solved in a language the market could verify.

Somethinc didn’t win because it had better content. It won because it built infrastructure — live commerce, omnichannel, creator flywheel — while its competitors were still treating TikTok as a promotional tool.

Love, Bonito didn’t win because it had better clothes. It won because it knew exactly who it was for, cut everything that confused that answer, and said something true about Asian women that no one else was saying.

That’s the escape. Not a tactic. A position.

Choose harder. Build deeper.

If this resonated, forward it to someone fighting the same battle. Reply to this email if you want to discuss where your brand sits in this framework.

Past newsletters:

• The Great Middle-Market Hollow-Out (Part I) — March 2026

• The Tightening Squeeze: Shopee Fee Analysis — January 2026

• Thinking Full Funnel — January 2026

• Building Beyond Marketplaces — October 2025

Something feels off.

I recently spoke with the founder of a mid-sized beauty brand — smart, passionate, a decade of genuine work behind her. Real products. Honest formulations. A loyal Shopee following. And she was losing.

Not catastrophically. Just slowly. Steadily. Like a slow drain you can’t quite locate.

Her repeat rate dropped from 38% to 24% over eighteen months. Her hero vitamin C serum was being undercut by three Shopee sellers at half the price. Her ad spend was climbing to hold the same GMV. Her gross margin had gone from 62% to 47%. And on TikTok, her content was getting buried under an avalanche of scroll-stopping videos she was increasingly certain weren’t made by humans.

She is not alone. I hear this often.

When I asked her what she thought was causing her declining repeat rate, her first answer was “the competition.” When I pushed harder, she said “maybe our ads aren’t targeting the right people.” When I pushed even harder, she said “honestly, I’m not sure.”

And the reason is not that these founders are bad at business — it is that the market has fundamentally restructured itself around them while they were busy running the day-to-day. The rules changed. The game board changed. Nobody sent a memo.

This is that memo.

I want to take you on a proper journey through what is actually happening in Southeast Asian e-commerce right now — not a summary, not a bullet-point list of trends, but a real exploration of the forces reshaping who wins and who gets squeezed out. Because I think if you understand the mechanics of what is happening, you will make very different decisions. Better decisions. And you will stop mistaking symptoms for causes.

We will cover the macro forces

1. The strong economy that somehow makes selling harder.

2. The China supply shock that rewrote the price floor for entire categories.

3. The rise of AI-generated content that is flooding platforms and making authentic storytelling simultaneously harder to do and more valuable than ever.

We will look at

1. What the middle market trap actually looks like from the inside, with a diagnostic you can use on your own business this afternoon. And then

2. We will get granular — deeply, practically granular — about what to actually do about it, with real case studies, real numbers, and real playbooks.

Let’s build that clarity together. Grab a coffee or a glass of your favorite whiskey on the rocks 🍺— this one is going to take a while. It is worth it. 🍻

The Paradox — The Economy is Strong. The Market is Brutal. Both Are True.

Southeast Asia is doing well — genuinely well. The SEA-6 economies are growing GDP at 4.5–5% annually. Private consumption is projected to reach US$5 trillion by 2035. Vietnam and the Philippines are near 6% GDP growth. By any macroeconomic measure, the region is a growth story.

So why are brands struggling?

Because GDP growth and brand market conditions are now two different things. From the Bain & Company / NielsenIQ November 2025 report — the most comprehensive read on SEA consumers right now:

• 42% of SEA consumers are actively seeking the same products at lower prices

• 25% have already switched to what they call “better value” brands

• 43% are cutting non-essential household spending

• 52% of brand executives anticipate slower growth in 2026

What this creates is not austerity. It is selective frugality paired with selective premiumisation. Consumers are cutting indifferent categories and paying up for things they genuinely care about. Health, beauty, baby, pets — growing. Generic mid-range — getting slaughtered. This is not a collapse. Just... a grinding squeeze.

Deloitte calls this the “More Value for Price” (MVP) shift. Consumers aren’t hunting for cheap things. They’re hunting for the right ratio — quality to price — and they now have real-time tools to evaluate that ratio across dozens of options in seconds.

Premium consumers still spend confidently. Budget consumers are served by an unprecedented range of affordable options. And the middle — brands positioned between “genuinely aspirational” and “genuinely affordable” — is being hollowed out.

RBC Capital Markets called it “economic bifurcation.” Modern Retail called 2025 the year “bifurcation defined brands.” This is not a SEA-specific blip — it is a global structural shift, amplified here by two forces unique to this region: the China supply shock and the AI content flood.

The China Supply Shock

Let me give you the scale before the mechanics.

China’s trade surplus with ASEAN grew from $90 billion (2021) to $190 billion (2024) — up 111% in three years. ASEAN became China’s largest export market in 2023, overtaking the US and EU. Chinese exports to ASEAN for Jan–Aug 2025 hit $434 billion (+14.7% YoY).

Chinese e-commerce companies — Alibaba, ByteDance’s TikTok Shop, Shein, PDD’s Temu — now account for approximately 50% of the online shopping market in Indonesia, Thailand, and the Philippines (Bain, 2024). That is not a projection. That is half your addressable e-commerce market already under Chinese platform control.

Why it accelerated: When US tariffs on Chinese goods hit 145% in 2025, Chinese factories needed a new market. SEA was the natural pivot — and the infrastructure (Temu, TikTok Shop cross-border, Shopee cross-border) was already built.

What it does to your pricing: Chinese sellers can operate on margins local brands cannot match for three structural reasons:

• Cost base: A Shaoxing garment factory running at 5% margin (down from 20%) still sustains a business. Your Malaysian equivalent at 5% is losing money.

• Platform support: Chinese sellers on TikTok Shop and Shopee often receive subsidised ad credits, preferential ranking during campaigns, and managed-service access that effectively shares their operating costs.

• Iteration speed: A Chinese manufacturer can develop, produce, and test a new SKU in weeks. Local brands run on 6–12 month product development cycles.

The factory-direct TikTok layer: In April 2025, the NYT covered a trend that’s arrived in SEA: Chinese factory owners posting TikTok videos showing the exact production process for branded goods, posting BOM costs, and inviting consumers to buy direct. “You’re paying RM200 for a product that costs RM30 to make.” Whether accurate or not, the narrative that brand premium is a marketing construct is extraordinarily persuasive to value-primed consumers.

The platform take-rate reality (which we covered in depth in our January 2026 issue on Shopee’s fee changes):

• Shopee Mall Malaysia fashion: 19.5% take-rate (Feb 2025)

• Shopee Mall beauty: ~15%

• TikTok Shop: started at ~5%, migrating upward as the platform matures

• Add VAT (8–12% by market), shipping subsidy, transaction fees — the effective cost of platform selling is substantially higher than the headline rate

The collateral damage is visible:

• Indonesia: 250,000 garment/textile workers laid off (2022–2024); Sritex — one of SEA’s largest textile manufacturers — entered bankruptcy in 2025

• Malaysia: Muar furniture hub containers shipped fell from 7–8K/year (2021) to 5–6K/year (2024)

• Thailand: Lampang ceramics industry contracted to 89 operating units

• The top 35 non-APAC FMCG multinationals collectively lost SEA market share over five years (Bain)

A Shaoxing garment factory owner quoted in the Straits Times put it plainly: “For any given product, there is no lowest price — only lower prices.” That philosophy is now operating inside your market.

The AI Slop Accelerant

Just as the pricing floor was collapsing, the content landscape had its own identity crisis.

Meltwater data: mentions of “AI slop” grew 9× between 2024 and 2025 — from 461,000 to 2.4 million conversations. Negative sentiment hit 54% in October 2025. The Guardian (December 2025) found that >20% of videos shown to new YouTube users qualify as AI-generated slop. Instagram CEO Adam Mosseri issued an internal memo stressing the need to label AI content, surface credibility signals, and reward originality.

Why this matters for your brand specifically:

On TikTok Malaysia in a typical week, tens of thousands of skincare-tagged videos are published. A growing proportion uses AI voiceover, AI-generated visuals, or AI-scripted content. They follow the same format conventions as creator content. They hit the same algorithmic triggers. But they have no genuine relationship behind them, no real product experience, no authentic perspective.

The algorithm doesn’t discriminate based on authenticity — at least not yet. Your authentic 50K-view video competes for the same real estate as an AI content farm’s 200K-view video. You can lose the algorithm race and still win the trust race. The question is whether you are building the latter.

For SEA beauty and lifestyle brands, the application is direct. A founder who appears on camera, formulates publicly, responds to real customer comments, and shows the actual manufacturing process is creating content that cannot be replicated by a Chinese cross-border seller with an AI content machine. It is inconvenient. It is time-consuming. It is also, in March 2026, the most defensible content strategy available.

AI slop creates a scarcity of authentic content at the exact moment when authentic content is most valued. That is an opportunity — if you build for it deliberately, not accidentally.

What to measure instead of raw view counts:

• Save rate (intent signal)

• Comment quality vs. comment volume

• Account follows generated per video

• % of revenue from organic (non-paid) traffic

These are brand-building metrics. View count alone is a content farm metric.

We explored how to build an authentic creator community at scale in our January 2026 Affiliate Arbitrage issue — including how leading brands are building 30–40% of GMV through creator networks without running a traditional influencer program.

The Diagnostic: Are You Already In the Trap?

The Middle Market Trap is not a single event. It is a gradual migration to a position where your brand is too expensive to compete on price — and not compelling enough to justify the premium. Here are seven signs. Score yourself honestly.

🔴 Sign 1: Your promotions are doing your brand’s job. New customers found you through a flash deal, not through your story. You are training deal-seekers, not building loyalists.

🔴 Sign 2: Your repeat purchase rate is below 30%. Healthy benchmark for beauty/personal care/health: 35–50% on a 90-day cycle. Below 30% is a leaky bucket. Rising CAC makes this terminal over 24–36 months.

🔴 Sign 3: Your cheapest SKU is your bestseller — and it’s over 40% of revenue. Customers are using you as a value option, not a brand they believe in. They are not ascending your range.

🔴 Sign 4: You can’t name three brand values your customers repeat back to you. Not what you believe your brand stands for — what your last 50 customers would say unprompted. Vague answers (”good quality,” “affordable”) = differentiation gap.

🔴 Sign 5: Your marketing and e-commerce teams run separate playbooks. Aspirational TikTok creative → Shopee product page full of discount badges. Brand contradiction, experienced in real time by your customer.

🔴 Sign 6: Your contribution margin is below 15% and falling.

For a full breakdown of how platform fee changes affect your unit economics, see our January 2026 issue: The Tightening Squeeze.

🔴 Sign 7: Your brand story hasn’t changed in three years. Read your About Us and your last week of social content side by side. Do they sound like the same brand at the same moment in its evolution? If the answer is no, you have a coherence problem.

Three or more? You’re in the trap. The exit requires building things your competitors can’t easily copy.

2 Brands That Found the escaped this trap

Sidenote: I took help from AI research with this

Case Study: Mayora Indah — The Distribution Moat

Mayora is one of the most instructive brand-building stories in SEA business history — and most people underestimate it because the products are unremarkable. Biscuits. Candy. Instant coffee. Founded in Jakarta, 1977.

The insight was simple: distribution is brand. Not logos, not advertising — the depth, breadth, and consistency of your presence in markets where competitors cannot or will not follow.

Mayora built proprietary logistics infrastructure across Indonesia and into 90+ countries, investing in store-level visibility and replenishment in markets too small or too thin for competitors to serve profitably. The result: ~6% annual revenue growth, net income doubled from 2016 to 2023, even as FMCG multinationals were losing share in the same categories.

Then in 2021, Vincenzo happened. The Korean Netflix drama became a global hit, and the lead character was shown consuming Kopiko coffee candy in multiple scenes. Organic, unpaid, completely unplanned. Hundreds of millions of impressions. A generation of K-drama viewers discovered the brand overnight.

The lesson is not “get your product in a K-drama.” The lesson is: Mayora’s decades of distribution investment meant that when the cultural moment arrived, the product was already there — on the shelf, at the right price, in consistent supply. The viral moment would have been worthless without the operational readiness.

For e-commerce brands: your Shopee store, your TikTok shop, your product pages — this is your distribution. Are you showing up in search for every relevant keyword? Does your product page convert every time someone lands on it? Do you ever go out of stock on a hero SKU during a campaign? Distribution in e-commerce is presence quality, and it is a moat you can build systematically.

Case Study: MAOGEPING 毛戈平 — The Authority Moat

Mao Geping is barely known outside China and the beauty industry. It should be known by every brand operator thinking about premiumisation.

Mao Geping is a makeup artist first, a brand second. He spent two decades becoming the defining authority on Chinese aesthetic beauty — inspired by traditional art, opera, and portraiture, explicitly positioned against the Western beauty canon. He launched his brand in 2000 when Chinese prestige beauty was completely dominated by Lancôme, Estée Lauder, Shiseido.

The model he built:

• 10 professional makeup artistry training institutes (3,800+ students in H1 2025 alone)

• 405 self-operated retail counters staffed by trained brand professionals

• Revenue: RMB 1.57B (2021) → RMB 3.88B (2024) — 35% CAGR

• Gross margin: 84.8% (vs Estée Lauder’s 74%, industry average ~58%)

• IPO: HKEX, 2024

• L Catterton (LVMH-backed) strategic partnership, 2026

That 84.8% margin is not magic — it is the quantified proof of a brand that has built genuine authority. The training institutes are the key. They train professionals, create deeply invested brand ambassadors, provide a credibility signal impossible to replicate, and feed into 405 self-operated counters where the brand controls every element of the customer experience — from product selection to application technique to the story told at the point of sale.

The pricing architecture reflects this: core products at $50–60 versus the typical domestic range of $15–30. The brand does not compete on price. It competes on authority. And authority, once established, is extraordinarily difficult for a competitor to dismantle — because it cannot be bought with a marketing budget, only built with time and consistency.

One more note on sequencing that is directly relevant for SEA brands with regional ambitions: international sales were just 0.05% of Maogeping’s revenue as of H1 2025. The L Catterton partnership is the beginning of a long journey. The lesson: domestic authority first, global expansion second. A brand globalised before it is truly authoritative is just another foreign brand on a crowded shelf.

The SEA lessons:

• Authority is a business model. A dermatologist founder explaining clinical evidence is an authority. An influencer licensing a white-label formula is not. The distinction determines your pricing moat.

• Cultural specificity is global currency. “Glow of the East” is specific and confident — not a compromise between Eastern and Western aesthetics. Your local ingredients, your local cultural context, your founder’s local knowledge — these are differentiation that no international brand can replicate.

• Vertical integration is protection. Control your formulation, your creator network, your CRM data, your community. Each link you own is a link no competitor can cut.

The Playbook: Five Pillars to Escape the Trap

Everything above is context — important context, but context. This section is what to do if you are feeling trapped. Five pillars, applied simultaneously. You can assess where you are on each one and prioritise accordingly.

1. Unify Marketing and E-Commerce Into One Revenue Function

The most common structural pathology I see across SEA brand businesses: marketing optimises for awareness, e-commerce optimises for GMV, and the customer experiences the collision. Beautiful aspirational TikTok creative leading to a Shopee store full of discount badges and lowest-price messaging. That brand contradiction costs you real money.

What integrated looks like:

• One brief drives creative and store optimisation

• One scorecard: CAC, LTV, LTV:CAC (target 3:1+), repeat rate (target 35–50%), brand search volume

• One weekly review: content performance and commercial performance in the same room

• Below RM5M revenue: one person owns both functions. Above: a Head of Revenue role accountable for fully-loaded unit economics, not just channel GMV

We covered the full-funnel integration model in our January 2026 issue: Thinking Full Funnel.

2. Own Your Customer Relationship Through CRM

Shopee, TikTok, and Lazada do not give you customers. They rent you access to theirs. Every customer who discovers you through Shopee search is Shopee’s customer — they can and will show that customer a competitor’s ad tomorrow.

WhatsApp is the most powerful CRM channel in SEA. Open rates: 70–90% vs. email’s 20–25%. It is not a marketing channel — it is the most personal digital space your customer inhabits. Brands that use it with discipline build relationships that survive competitor price promotions.

Email for longer-form relationship content — ingredient deep dives, founder letters, educational series. Think of WhatsApp as your conversation channel (high frequency, transactional when needed) and email as your relationship channel (lower frequency, brand-building over time). Build both.

For DTC channel architecture including WhatsApp, see our October 2025 issue: Building Beyond Marketplaces.

3. Build a Affiliate Creator Community, Not an Influencer Programme

An influencer program is a transaction: you pay someone to promote your product. A creator community is a relationship: people who genuinely use, believe in, and communicate your brand as part of their life.

The difference shows up in the content — and audiences in 2026 can tell.

The architecture:

• Target: 200–500 active affiliates per market (10K–50K followers), producing 5–10 videos/day in aggregate

• Tier 1 : Your inner circle. Early access, product co-development input, feel genuine ownership

• Tier 2 : Active advocates. Commission-based via TikTok Shop affiliate, consistent content cadence

• Tier 3 : Community layer. Customers who create organically, invited to graduate upward

TikTok Shop affiliate commission for beauty/personal care: 10% – 25%. A video generating RM10,000 in sales costs you RM1,500 in commission on average — and the content keeps driving sales for weeks. Compare that to a paid ad generating the same sales volume: you pay upfront, regardless of conversion, and the asset has a short shelf life.

The brief that works: Start with the customer problem, not the product feature. Tell the creator: “Our customers aged 28–38 often struggle with [specific problem]. We want you to show your audience how real this problem is in your own life, and share what has genuinely worked for you — which might include our product if it genuinely has.” Leave the creative execution entirely to the creator. Your compliance layer should be minimal: disclose the partnership, avoid specific medical claims. That is it. Creative freedom produces authentic content. Authentic content converts.

Platform strategy — be intentional, not omnipresent:

• TikTok and Shopee: discovery engine, top-of-funnel, trial driving. Your highest reach potential but lowest trust signal per view.

• Instagram: brand equity, lifestyle storytelling, Tier 1–2 partnership content. Skews older and more aspirational in Malaysia, Singapore, Philippines.

• WhatsApp: conversion and retention layer. Not a broadcast channel — a relationship channel. Treat it accordingly.

Full creator program mechanics, including how leading SEA brands are building 30–40% of GMV through affiliate communities, in our January 2026 Affiliate Arbitrage issue.

4. Know Your SKU Economics — Really Know Them

Most brand operators manage on top-line GMV. That is the accounting equivalent of driving while looking in the rearview mirror.

Your portfolio management decisions once you have SKU-level numbers:

• Hero SKUs (high margin, high volume): Protect obsessively. Never deep-discount. Never go out of stock.

• Builder SKUs (low volume, acceptable margin): Invest in or investigate why they’re not converting

• Strategic SKUs (breakeven, clear role): Only justified if you can measure the LTV of customers acquired through them. “We assume they trade up” is not measurement.

• Pruning candidates (negative contribution, no strategic rationale): Discontinue. They waste inventory capital, platform ranking bandwidth, and your team’s management attention. Every hour spent managing a loss-making SKU is an hour not spent on a hero SKU that actually builds the business.

Also track what most brands don’t: Campaign management fees per unit of GMV. Pick-pack-ship cost per order. Content production cost per piece and per unit of sales generated. Customer service cost per query. Each of these is a lever. You cannot pull levers you cannot see.

5. Create Value That Isn’t a Discount

If price reduction is your primary competitive response, you have entered a race you cannot win. The alternative is creating more value, not less margin.

The tools:

• Gift with Purchase (GWP): A $30 travel-size gift with purchases over $100 costs $8–12 in product and packaging. If it converts 20% more customers who would have waited for a sale and introduces them to a product they then repurchase, the LTV return is multiples of the cost.

• Bundles: A $129 “Morning Glow Bundle” (serum + SPF + travel pouch) changes the comparison frame entirely. Consumers are no longer comparing your $89 serum to a $45 Chinese competitor — they are comparing a bundle value proposition to a standalone product. The bundle wins that comparison almost every time.

• Memberships and Loyalty: Not a points scheme.
  An access-and-identity programme.
  Tier 1: early launches, exclusive content.
  Tier 2 (3+ purchases): free shipping always, VIP samples, product feedback panels.
  Tier 3 (highest LTV): founder access events, co-creation, birthday discount.
  The goal: make switching brands feel like leaving a community, not just changing a product.

• Educational content as value: A video teaching customers how to use your product correctly is customer service that increases perceived value and drives repeat purchase. In beauty, a dermatologist or pharmacist explaining the why behind your formulation is an authority signal that no AI-generated content can replicate. Education is the most underinvested, highest-returning content investment available.

For a full playbook on building DTC from scratch — including CRM flows, community strategy, and WhatsApp Commerce — see our two-part series: Cold Start Part 1 and Cold Start Part 2.

Which Brand Are You?

The playbook above applies differently depending on your starting position. Four archetypes:

🏠 The Local Hero (Malaysian, Indonesian, Thai, Filipino, or Vietnamese brand feeling the squeeze) Your asset is authenticity — something a Chinese cross-border seller cannot replicate. Double down on your founder story, ingredient sourcing, and cultural specificity. Priority: build CRM aggressively before the platform shows your customers a cheaper alternative. Invest in micro-creators who share your cultural context.

🌏 The Foreign Premium Entrant (international brand justifying premium in a more crowded market) Your challenge is credibility without years of local presence. Build it through clinical validation, professional-credential creator partnerships, and premium retail signalling. Don’t rush distribution. Maogeping’s lesson: be genuinely number one somewhere before being mediocre everywhere.

🚀 The SEA-Born Regional Brand (expanding from one SEA market into others) Your local brand story in Market A reads as foreign premium in Market B — the same halo Korean brands carry when entering Malaysia. Use that positioning consciously. Sequence your expansion: establish top-5 category presence in Market 2 before entering Market 3. Horizontal scale with shallow presence produces the same result as mid-market positioning: too thin to be loved anywhere.

💡 The Affordable Disruptor (genuinely targeting the value tier, but with a real brand) Your opportunity is enormous — Deloitte’s data shows a 2% shift in SEA consumer spending toward MVP brands, worth billions at scale. But accessible pricing must be a door to a relationship, not a destination. Skintific and O.TWO.O captured significant Indonesia market share in under three years by combining competitive pricing with massive TikTok creator communities and real-time product feedback loops. Price gets them in. Community keeps them.

The Bottom Line

The middle market is being hollowed out by four convergent forces:

1. China supply shock — destroyed the price floor in most consumer categories

2. Platform consolidation — extracting more brand margin in exchange for consumer access

3. AI content flood — made authentic storytelling simultaneously harder and more valuable

4. Consumer bifurcation — rewarding the genuinely premium and genuinely affordable, punishing everything in between

If you are in the middle, staying there is not a neutral choice. It is a slow retreat. The forces are moving against you, and they are accelerating.

The escape requires three things: clarity about which side you are on; investment in the five pillars (integrated revenue function, owned CRM, creator community, SKU economics, non-discount value creation); and intellectual honesty about what the diagnostic actually tells you — not what you hope it says.

Here is what I genuinely believe: this moment favours brands with invested depth and lots of authenticity. Brands built by people who know their ingredients, tell true stories, show up consistently, and build real relationships. .

A rather lengthy piece this week to help me clear my thoughts as we work through navigating the middle market trap for some of our clients. Hope you enjoyed it too!

Thanks for reading.

Took a week’s break from writing after diving deep into the Ecommerce Cold Start problem written in 2 parts. This week, I would like to focus more on a specific category that is also one of the largest and most competitive one in the Ecommerce space, Beauty.

If you’ve been reading my previous deep dives on Fashion and Health & Wellness, I like to dig into the stuff that actually matters when you’re staring at a P&L wondering why your beauty brand isn’t scaling the way you thought it would.

This week, I’m knee-deep in beauty and skincare—not by choice initially, but because three different conversations last month forced me to really understand this category at a level I hadn’t before. A premium skincare founder in Middle East asking whether Singapore and Malaysia was worth the regulatory headache. A Korean beauty distributor wondering if shipping direct from Seoul made sense versus warehousing in each market. And a local Malaysian indie brand trying to figure out if they should stay local or go after the broader SEA opportunity.

What struck me across all three conversations was how different the beauty playbook is depending on who you are. The premium Western brand entering Singapore has completely different economics and growth levers than the affordable K-beauty label testing Vietnam, which is nothing like the local hero trying to break out of their home market.

So instead of giving you one-size-fits-all advice (which is usually one-size-fits-nobody), I’m going to break down this category through the lens of four different beauty brand archetypes—because your strategy, platform mix, and unit economics should look radically different based on where you’re starting from.

The State of Beauty E-commerce in Southeast Asia

Before we get into the archetypes, let’s talk about what’s actually happening in this market right now.