Cornett received about 45 percent against Kevin Stitt. Drummond received 49.7 percent and came within 2,045 votes of winning the nomination. Drummond’s team deserves a great deal of credit for exceeding expectations and keeping the AG competitive through the final vote.

That result brought back memories of 2018 because I worked for Stitt in that campaign. We were able to define the runoff clearly: Stitt was the conservative outsider, while Cornett was the moderate OKC mayor who had supported tax increases. Cornett had plenty of strengths and a strong geographic base, but the choice voters carried into the booth favored us.

The Drummond team kept this year’s race from becoming that kind of ideological contest. I had worked earlier in the primary for another candidate with a proven conservative record, but we failed to make that record the basis on which voters evaluated the field. We had research showing openings and messages that moved the voters we needed. Sadly we never turned those findings into the campaign’s central theme.

I’m probably too close to that loss, and it is still fresh. Still, it illustrates something I have come to believe after a pretty long time in campaigns: possessing the right information and building a campaign around it are very different matters.

Today, campaigns have more data, more sophisticated models and more ways to communicate than ever. Sadly, many still operate through the same basic structure we used decades ago. The pollster presents numbers, the consultants discuss them, each vendor returns to a separate lane, the media team writes scripts, and then the team waits for the next poll to learn whether it worked.

The best campaigns I have worked on operated differently, even before we had language for it.

During the 2016 Iowa caucuses (and even into Super Tuesday), the Cruz campaign effectively connected polling, modeling, field intelligence, digital behavior and message development. Information moved across the campaign and affected real decisions about resources, audiences and communication.

On Election Day, we saw that Google search behavior in many states gave us a better indication of where interest was moving than my tracking polls. That did not make polling obsolete. It showed me how many signals voters were producing outside traditional research and we adapted to that.

I saw another version of this in Missouri in 2022, when I worked for Eric Schmitt in the Republican Senate primary. Former governor Eric Greitens entered with high name recognition and a devoted following. Schmitt’s record as AG gave us a stronger and more current case to make: he was the conservative who had been leading the fight with the Biden administration. The campaign established that contrast and repeated it. Schmitt won the crowded primary with nearly 46%; Greitens finished third with about 19. Even President Trump’s wonderfully ambiguous endorsement of “ERIC” could not disrupt the choice voters had already formed.

Glenn Youngkin’s 2021 victory in Virginia followed the same pattern. Parents were concerned about school closures, curriculum, standards and the feeling that institutions had stopped listening to them. Our campaign recognized the common thread and organized those concerns around the idea that parents should have a voice. That gave a very diverse group of voters a shared reason to support Youngkin in a state Republicans had not won for governor in nearly a decade.

Companies have wrestled with this same challenge. Toyota’s production system became famous because information from the factory floor could change the work while it was underway. Problems did not have to travel through layers of management and wait for the next quarterly review. The system created a habit of listening, deciding, acting and learning.

That is what I mean by a new campaign operating system. Polling, narrative intelligence (like EyesOver), analytics and AI will supply more signals than most campaign teams can absorb on their own. The advantage now goes to campaigns (and consultants) that can quickly ingest those signals, establish a clear strategic choice, give a trusted team authority to act and learn while time still exists to change the outcome.

The victories taught me how powerful that can be. The losses, especially the recent one here at home, have taught me how expensive it is when a campaign never quite gets there.

But there was a problem: the polling company wasn’t real.

Median Strategies appeared to be a legitimate research firm. It published methodology, field dates, sample sizes and the other things those of us who have spent our careers in polling instinctively look for. One of its surveys showed Los Angeles Mayor Karen Bass leading challenger Nithya Raman by about 12 points. The Bass campaign shared it. Median also published polling in Wisconsin and Nevada.

Then the company disappeared.

Median eventually acknowledged that the polls were fabricated and described the entire operation as a “short-term social experiment” designed to see how unverified polling information could move through the political ecosystem. According to the Associated Press, established aggregators had declined to include the surveys because Median could not provide sufficient information about its methods, but that didn’t prevent the numbers from traveling.

There is a temptation to treat this as a funny story about political Twitter getting fooled. I think that misses the point.

Median Strategies demonstrated how easily a number can acquire authority simply because it looks like real data.

This has happened before

Polling has always depended on trust. Most people who see a poll will never inspect the questionnaire, evaluate the sample design, examine weighting procedures or ask whether the interviews underlying the toplines actually occurred. They rely on institutional credibility.

We learned how badly that system can fail with Research 2000.

In 2010, Daily Kos terminated its relationship with the polling firm after statisticians examining its weekly surveys found patterns they concluded could not plausibly have come from random polling. Daily Kos subsequently said it could no longer trust the data and accused Research 2000 of providing results that were fabricated or manipulated beyond recognition.

Go back another decade and the 2000 South Carolina presidential primary produced a different corruption of the same basic idea. So-called “push polls” were used to deliver political attacks while masquerading as research. As CBS polling director Kathy Frankovic explained at the time, a push poll isn’t a poll at all; it is political telemarketing dressed up to look like research.

Twenty-six years later, we have almost the inverse problem. Instead of persuasion pretending to be research, we now have fabricated data pretending to be evidence.

What makes 2026 different is how inexpensive it has become to manufacture the appearance of credibility. AI can help create the firm, the website, the methodology, the analysis, the charts and the social content around it. None of that makes the underlying numbers real.

It is, in a sense, quantitative slopaganda.

But real polls create a different problem

The more interesting lesson from this summer may actually come from Wisconsin.

Marquette Law School is a real polling operation run by people I respect. Its final Democrat gubernatorial primary survey had Francesca Hong at 38 percent and David Crowley at 7 percent, with 34 percent undecided. Crowley ultimately won with 39.8 percent; Hong finished at 39.3 percent.

At first glance, that looks like an enormous polling miss. Look closer and something more interesting happened.

Marquette measured Hong almost perfectly. She polled at 38 percent and received 39.3 percent. What changed was everyone else.

Three candidates left the race during a chaotic final month. Crowley re-entered, consolidated support and captured an extraordinary share of the voters who made their decisions after the poll was completed. Marquette’s Charles Franklin has since pointed out that the overwhelming majority of those late-deciding voters appear to have gone to Crowley.

That isn’t fake polling. It isn’t even necessarily bad polling. It’s a reminder of what a poll is: a measurement taken at a particular moment in time.

And that’s why I think Median and Wisconsin belong in the same conversation, even though they represent completely different problems.

Median gave us a signal that wasn’t real while Marquette gave us a legitimate signal from an electorate that wasn’t stable. Crowley had the momentum. But polling doesn’t measure momentum.

The answer is triangulation

I’ve spent most of my professional life around polling, so I’m not particularly interested in declaring its death. Good polling is still one of the most valuable instruments available to a campaign.

But I increasingly believe the mistake is allowing any single instrument to become the truth.

If a poll suddenly tells me a candidate has moved 15 points, I want to know what else changed. Did sentiment move? Did an issue suddenly become more salient? Did a message begin resonating? Did the coalition become more volatile? Did the narrative environment change? Are voters behaving differently?

If the answer to all of those questions is no, that doesn’t prove the poll is wrong. It does tell me I should start asking questions.

That’s one reason I’ve become such an advocate for looking at campaigns through multiple lenses. Polling tells us what voters say. Predictive analytics helps tell us which voters matter. Continuous behavioral measurement (like EyesOver) lets us observe whether the environment around those voters is actually changing.

The power isn’t in choosing one over the others. It’s in understanding what it means when they agree, and paying particular attention when they don’t.

A race like Wisconsin, for example, might have looked very different through that lens. Hong’s 38 percent wasn’t necessarily the warning sign. The warning sign was an electorate with a third of its voters still undecided, candidates entering and leaving the race, a rapidly changing endorsement environment and enormous uncertainty about where displaced voters would land.

I wouldn’t claim that any data platform could have looked at those conditions and predicted Crowley at precisely 39.8 percent. That’s not how this works.

But a good campaign intelligence system should have been screaming something much more useful: This race is not stable.

That is information a strategist can act on.

The whole dashboard

The Median Strategies story arrives at an interesting moment. We are entering an information environment in which producing something that looks authoritative is becoming almost frictionless. Polls are one example, but the same problem applies to news, video, research, social conversation and increasingly almost every other source campaigns use to understand the world.

That means verification can no longer mean checking whether something has a professional-looking methodology page. It means looking for corroborating signals.

For most of my career, the argument inside campaigns was about which poll to believe. I think that question is becoming less useful.

The better question is whether the poll agrees with everything else we can observe about the electorate.

Sometimes the number will be legitimate and the electorate will move immediately afterward. Sometimes the methodology will be flawed. And, apparently, sometimes somebody will simply make the whole thing up.

The response to all three situations is essentially the same.

Never trust one instrument when you can see the whole dashboard.

The article calls it atomization. I think that is a useful way to look at what campaigns are confronting, although perhaps not for the reason people might assume.

We still talk about the electorate as if it consists of a few large and relatively stable groups: Republicans, Democrats and independents; suburban women and working-class men; Black, Hispanic and young voters. Those categories are useful. I have relied on them for a pretty long time, and I’m not suggesting we throw them away.

But they increasingly describe only one part of a voter’s identity. A person can be economically populist, culturally conservative, distrustful of large institutions, concerned about the environment and unaffiliated with either political party.

The voter may listen to a podcast that reaches more people than most CNN or MSNBC programs while also participating in an online community that most campaign consultants have ever heard of. Which part of that voter’s identity matters politically can change from week to week.

The electorate is not simply divided. It is constantly reorganizing.

President Trump’s 2024 coalition is a useful example. It included traditional Republicans, working-class voters, young men, technology entrepreneurs, podcast audiences, parents concerned about health and voters motivated primarily by inflation or immigration. Those people did not necessarily share the same ideology, media habits or reasons for voting. For a period of time, however, their frustrations pointed in the same direction.

The corporate world has been dealing with a similar change for years. There was once a mass audience that could be reached through a small number of television networks, magazines and newspapers. That audience splintered across streaming platforms, podcasts, social networks and thousands of online communities. Companies could no longer assume that their customers watched the same programs, trusted the same institutions or even used the same language to describe what they wanted.

Campaigns have adapted tactically to this world, but I’m not sure they have adapted strategically.

We now buy audiences instead of programs. We microtarget digital advertising. We divide voter files into increasingly precise segments. We test more messages against more groups and produce more content for more platforms.

All of that has made campaigns more efficient, but the underlying model has not changed very much. We are still placing voters into predefined categories and looking for the message most likely to move each one.

That can conceal as much as it reveals. Two voters of the same age, race, education level and party registration may inhabit completely different information environments. One may reach a political judgment through local economic concerns, while the other is reacting to a national cultural argument. On a spreadsheet, they look nearly identical. In the real world, they may have very little in common.

The Clinton campaign discovered the limits of demographic confidence in 2016. Its coalition appeared sufficient on paper, and the traditional models suggested that states such as Michigan, Pennsylvania and Wisconsin would hold. What those models did not fully capture was the connection forming among voters who felt ignored economically, culturally and politically.

Brad Parscale understood this. And targetd voters in a way that help Donald Trump perform better across demographic categories. He assembled dissatisfaction across several of them.

Kodak made a comparable mistake in a very different setting. It continued to understand photography through the business of film, even as consumers were changing what photography meant. People no longer wanted only to take and print pictures. They wanted to create, share and organize them instantly. Kodak knew its customers, but it was looking at them through categories formed by an earlier market.

That is the risk for campaigns now. We know a great deal about the voters as they have been classified. We know less about the communities and concerns bringing them together at a particular moment.

Glenn Youngkin’s 2021 campaign in Virginia offers a better model. Concerns about schools, parental authority and institutional responsiveness began connecting suburban parents, traditional Republicans and even minority voters dissatisfied with Democratic governance. Those were not identical constituencies, and not every voter entered the issue from the same direction. The campaign recognized the connection and gave it a simple organizing idea: parents should have a voice.

That was more than targeting. It was coalition construction.

Nike has done something similar in the consumer world by building a community around the idea of being an athlete rather than around a narrow demographic definition of who buys athletic shoes. Its customers include elite competitors, weekend runners and people who may never enter a race. They have different backgrounds and motivations, but the brand gives them a shared identity large enough to hold them together.

This is where newer technology, including artificial intelligence, can help campaigns. It can identify emerging conversations earlier, detect connections among audiences that do not fit the usual categories and follow the movement of narratives in something closer to real time. That is a significant improvement over depending entirely on periodic polling and the instincts of people sitting in a campaign headquarters.

But seeing a coalition is not the same as building one.

A dashboard can show that several groups are expressing versions of the same concern. It cannot decide what a candidate should stand for, which political risks are worth taking or what larger argument can turn those concerns into a governing majority. Those are questions of strategy and judgment.

Barack Obama’s campaign understood this in 2008. It used data and technology more effectively than any presidential campaign before it, but the technology was not the reason millions of people with different backgrounds saw themselves as part of the same movement. “Hope and Change” gave them a common story. The tools helped find, organize and mobilize the coalition. They did not supply its meaning.

Netflix provides the corporate version of the same lesson. It has extraordinary behavioral data and can divide viewers into remarkably specific taste communities. But data can tell Netflix what people watched; it cannot reliably manufacture the next cultural phenomenon. Eventually, someone still has to exercise creative judgment about which story is worth telling.

The dangerous response to atomization would be to conclude that every voter should receive a different message. AI will make that increasingly easy. A campaign could generate thousands of variations, each optimized for a narrow audience or even an individual person.

That might increase engagement, but it could also leave the campaign with no central identity. If every voter is told a different story, what does the candidate actually believe? What would unite the coalition after the election? A campaign cannot assemble a governing majority by becoming as fragmented as the electorate it is trying to reach.

We have seen a version of this in social media. Personalized feeds were remarkably effective at giving individuals more of what held their attention. They were less successful at preserving a shared understanding of events. What worked for engagement did not necessarily work for society.

The best political campaigns have always done something larger. Ronald Reagan’s 1980 coalition included economic conservatives, national-security voters, evangelical Christians, working-class Democrats and Americans who were simply exhausted by a sense of national drift. Those groups did not agree on every issue. Reagan gave them a broader argument about restoring American confidence and strength.

Apple accomplished something comparable with “Think Different.” It did not create separate identities for designers, entrepreneurs, students and technology enthusiasts. It placed all of them inside one larger story about creativity and challenging convention. The audiences were different. The organizing idea was coherent.

That is the opportunity for campaigns in an atomized America. Better technology can help us listen more closely and understand how concerns are moving across communities. It can show us connections that our traditional categories miss. But the purpose cannot be to perfect a separate appeal for every fragment.

The purpose is to find the connective tissue among them.

America may be atomizing, but politics cannot stop at understanding the fragments. The real work of a campaign is assembling them into something capable of winning—and, eventually, governing.

That’s a difficult sentence to write. It was even more difficult to watch unfold in real time because this wasn’t a campaign flying blind. Quite the opposite. It had experienced consultants, quality polling, sophisticated data, and real-time behavioral measurement showing how voters were responding to its advertising as the campaign progressed.

What struck any observer wasn’t the lack of information. It was the abundance of it.

Three completely different systems kept pointing in roughly the same direction. The polling identified the issues and messages voters care about/preferred. Predictive analytics identified the voters most likely to be persuaded. Behavioral measurement showed which advertisements were generating positive movement and which were creating negative movement.

None of it impacted the strategy.

When one ad underperformed, the response wasn’t to stop and ask why. It was to produce another one. More money was spent. More media was purchased. The feedback loop was there for everyone to see, but they didn’t want to believe it.

Watching that happen, I found myself thinking less about one campaign and more about the profession itself.

What exactly is the role of the campaign consultant in 2026?

For most of modern political history, the answer would have been obvious.

Campaign consultants existed because they possessed something extraordinarily valuable: information.

They knew how to buy television, understood polling, managed direct mail, coordinated fundraising, and-perhaps most importantly-pulled all of those pieces together into a coherent campaign. If you were running for governor in 1998, or even 2018, there simply weren’t many other places to get that kind of expertise.

The general consultant became the campaign’s operating system.

Looking back, it was a perfectly rational model.

I don’t think it is anymore.

One of the biggest structural changes in politics over the last decade has received surprisingly little attention. The campaign itself is no longer the center of gravity. Increasingly, that role is shared-or sometimes dominated-by party committees, congressional leadership funds, governors associations, outside organizations, and highly specialized firms.

Leadership Funds have media capabilities that rival many statewide campaigns.

The RNC has invested heavily in sophisticated data infrastructure.

Behavioral measurement platforms barely existed a few years ago.

Conversational AI is beginning to reshape voter engagement.

None of those developments are criticisms of consultants. They’re simply evidence that the information monopoly consultants once held is disappearing.

Interestingly, the same transformation is happening far outside politics.

Management consulting firms like McKinsey, BCG, Deloitte, and Accenture are all rethinking what clients actually hire them to do. Recent research on AI-enabled consulting suggests that the consultant’s value is shifting away from producing analysis and toward helping organizations make better decisions in increasingly complex environments.

That distinction matters because it mirrors exactly what’s happening in politics.

Analysis is becoming abundant.

Between polling, predictive analytics, behavioral measurement, and AI, campaigns now have access to more information than at any point in history.

Judgment, on the other hand-the ability to synthesize conflicting evidence, recognize meaningful patterns, and change course when the facts demand it-is becoming the scarce resource.

This is why I don’t believe the campaign consultant disappears.

I think the campaign consultant who refuses to evolve disappears.

For decades, experience was enough. If you’d run enough races, bought enough television, and won enough campaigns, your instincts carried enormous weight.

They still should. But experience now has company.

That’s an important distinction. I’m not arguing that data replaces experience. The best consultants I’ve worked with combine both. They use experience to ask better questions, and they use evidence to challenge their assumptions.

The problem begins when experience stops being a guide and becomes a substitute for evidence.

Polling tells us what voters say.

Predictive analytics tells us who matters.

Behavioral measurement tells us what’s happening.

Conversational AI is beginning to help us understand why what is happening is happening.

Ignoring any one of those sources of information is increasingly difficult to justify.

One lesson I’ve learned over nearly four decades in politics is that confidence is often gets mistaken for competence.

Some of the best consultants I ever work with are also the most intellectually curious. They treat every campaign as another opportunity to learn something they don’t already know.

The worst are often the opposite.

They believe experience exempts them from learning and are uninterested in “bleeding edge” strategies/technologies.

That’s a dangerous mindset in any profession.

It’s especially dangerous in one changing as quickly as political campaigns.

I suspect the campaign consultant of 2030 will spend far less time producing things and far more time interpreting them. They’ll still help shape strategy, but increasingly they’ll be integrating polling, predictive analytics, behavioral measurement, conversational AI, and dozens of other signals into a coherent point of view.

Their competitive advantage won’t come from having access to information no one else has.

It will come from knowing which information matters.

Which brings me back to the campaign I watched earlier this year.

Perhaps the campaign consultant isn’t ending after all.

Maybe what’s ending is the idea that one person’s instincts should outweigh multiple independent sources of evidence.

That’s a healthy evolution.

Politics has always been part art and part science. The balance is simply changing. The consultants who embrace that change will become more valuable than ever. The ones who don’t will slowly find themselves solving tomorrow’s problems with yesterday’s operating system.

Because the future won’t belong to the consultant with the strongest opinions.

It will belong to the one who learns the fastest.

For the first time in quite a while, I missed two weeks of publishing. My wife, Stephanie, and I took our family on a cruise through Alaska’s Inside Passage. Between watching humpback whales breach, hiking near glaciers that have existed for thousands of years, and spending uninterrupted time with our kids and parents, I did something campaign professionals rarely allow themselves to do.

I disconnected.

Well…mostly. Aside from a few data science projects that needed attention, it was just family and a landscape that has been changing, almost imperceptibly, for centuries.

The funny thing about stepping away is that you often come back seeing things more clearly.

Somewhere between Juneau and Ketchikan, I found myself asking a question I hadn’t considered before:

If I walked into the headquarters of the best campaign in America in 2030, what would actually be on the wall?

For years, our industry has debated tactics. Will AI write better fundraising emails? Will it generate more persuasive television ads? Will it produce social media content in seconds instead of hours?

Those are interesting questions, but I don’t think they’re the important ones.

The real transformation isn’t about how campaigns create content. It’s about what campaigns choose to measure.

When I got my start in politics, campaign headquarters were fairly simple. The walls were covered with fundraising goals, county maps, polling crosstabs, media plans, and turnout targets. Over time, the dashboards became more sophisticated. Polling became more frequent. Voter files became richer. Predictive analytics helped campaigns decide who to target. Digital advertising introduced entirely new performance metrics.

Each generation added another lens through which campaigns understood the electorate.

Yet if you walked into most statewide campaigns today, the dashboard would still look surprisingly familiar. Fundraising totals. Polling averages. Field contacts. Media impressions. Advertising performance. They’re all valuable, but I don’t believe they’ll define the next generation of winning campaigns.

Over the past year, I’ve written about persuadability, voter volatility, issue salience, behavioral movement, and narrative intelligence. Those aren’t isolated ideas. They’re pieces of a larger shift that’s taking place in political strategy.

Instead of asking, “Are we winning?” campaigns will increasingly ask questions like: How stable is our coalition? Which issues are voters actually talking about today? Which narratives are reshaping the environment? Which messages are genuinely changing attitudes rather than simply generating engagement?

Traditional polling was never designed to answer many of those questions. Polls are extraordinarily good at measuring opinion during the period they’re in the field. But political narratives often emerge, evolve, and disappear far faster than even the best tracking survey can detect.

That’s why I believe continuous behavioral measurement will become just as important as traditional polling. They’re not competitors. They answer different questions.

Artificial intelligence will accelerate that transition, but probably not in the way most people expect. The biggest impact won’t be AI-generated fundraising emails or campaign ads. It will be helping campaigns understand extraordinarily complex environments in real time.

Take conversational AI as one example. Platforms like SparkFire make it possible to engage voters through individualized, two-way conversations at a scale that was unimaginable only a few years ago. Every interaction becomes both persuasion and research, helping campaigns understand not only what voters believe, but why they believe it.

Now combine those conversations with continuously updated intelligence from platforms like EyesOver and a conversational interface powered through Model Context Protocol (MCP). Instead of opening half a dozen dashboards, exporting spreadsheets, and waiting for an analyst to build a presentation, a strategist can simply ask:

“Why has suburban support softened over the past ten days?”

The answer isn’t another chart. It’s a synthesized explanation connecting polling, online conversation, earned media, advertising performance, issue salience, SparkFire interactions, and historical context.

That’s a fundamentally different way of making decisions.

For decades, campaign technology has focused on collecting more information. I think the next generation will focus on helping campaigns understand it. The strategist’s job shifts from assembling data to asking better questions.

And I suspect that’s what will separate successful campaigns from everyone else.

One lesson I’ve learned over nearly four decades in politics is that confidence is often mistaken for competence. The most dangerous campaigns aren’t the ones that lack information. They’re the ones that stop believing they have anything left to learn.

As our ship moved quietly through Glacier Bay, I couldn’t help thinking about how enormous change often looks almost invisible while it’s happening. The glaciers don’t appear to move until you compare today’s photograph with one taken twenty years earlier.

Campaigns are going through a similar transformation.

We’re still debating television versus digital. We’re still organizing campaigns around demographic coalitions. We’re still celebrating bigger datasets. Yet quietly, an entirely new way of understanding elections is emerging—one built around continuous learning instead of periodic measurement.

I don’t believe polling is disappearing. I don’t believe predictive analytics are disappearing. And I certainly don’t believe experienced strategists are disappearing.

I think they’re all becoming part of something much larger: a campaign dashboard that doesn’t simply describe where an election stands, but helps explain where it’s going.

• The average suburban woman.

• The average independent.

• The average evangelical.

• The average Hispanic voter.

• The average union household.

Every election seemed to produce another label that consultants quickly converted into a targeting universe. Entire media plans, messaging strategies, and turnout operations were built around these broad demographic buckets. And for a long time, that made reasonable sense. Demographics were among the best predictors available.

The problem is that demographics describe people, but they do not reliably explain behavior.

That distinction is becoming one of the most important developments in modern politics.

Political science has always tried to answer a deceptively simple question: Why do people vote the way they do?

The first generation of answers focused on demographics. Age, income, education, religion, geography, and race explained a remarkable amount of variation in voting behavior. Later came party identification, ideology, and consumer data. Each iteration added another layer of understanding, and campaigns became more sophisticated as a result.

But over the last decade, something changed: the electorate became less predictable because people who look similar on paper increasingly behave differently in practice.

Two suburban women of the same age, income, education, and party registration can now respond completely very differently to the same political message. Meanwhile, two voters who appear to have almost nothing in common demographically may react to exactly the same issue for exactly the same reasons.

The average voter is becoming less useful as a planning concept.

I was reminded of this several years ago during a statewide research project.

At the outset, the campaign’s assumptions were perfectly reasonable. Like most campaigns, it believed suburban women would primarily be responding to the issues dominating national political coverage. We tested those issues, of course. But we also went looking for something else. We wanted to identify issues that were uniquely salient; not to broad demographic groups, but to smaller behavioral segments that traditional political analysis often overlooks.

One issue kept coming up in focus groups: Red-light traffic cameras.

On its face, it sounded absurd. No serious political consultant would have walked into a strategy meeting and declared that automated traffic enforcement was one of the defining persuasion issues for a critical slice of suburban voters.

But the findings kept telling the same story.

The issue wasn’t really about traffic cameras. It was about fairness, local control, government overreach, and the feeling that ordinary citizens were being treated as revenue sources rather than constituents. The cameras had become a proxy for something much larger.

The lesson had very little to do with transportation policy and had everything to do with behavior.

The most interesting political questions increasingly aren’t “Who is this voter?”

They’re “What motivates this voter?” and “Under what conditions does this voter move?”

This is where data science begins to change the discipline. Traditional political analysis groups voters into categories. Data science identifies patterns.

That sounds like a subtle distinction, but it changes the entire operating model of a campaign.

Modern clustering algorithms don’t begin by asking whether someone is male or female, suburban or rural, Republican or Democrat. They begin by looking for relationships that humans are unlikely to discover on their own. They identify people who behave similarly, consume information similarly, respond to similar narratives, or exhibit similar patterns of movement; even if they share very little demographically.

The result is that campaigns think less in terms of demographics and more in terms of behavioral fingerprints.

That shift is already happening outside politics. Recommendation engines don’t organize people by age. Fraud detection systems don’t organize customers by geography. Modern medicine increasingly groups patients by biomarkers rather than symptoms alone.

Politics has moved in the same direction.

Recent academic work points in this direction as well.

Research on AI-driven political persuasion has increasingly shown that persuasion effects are highly heterogeneous. The same message produces dramatically different outcomes depending on the individual receiving it, and behavioral outcomes often diverge from attitudinal ones.

In other words, changing someone’s opinion and changing someone’s behavior are not the same thing.

The implication is profound: broad averages increasingly conceal the very dynamics campaigns most need to understand.

Behavioral economics has been pointing toward this conclusion for years. Context matters. Identity matters. Timing matters. Salience matters. Human beings do not process information as interchangeable members of demographic groups. They respond through a complex interaction of values, incentives, emotions, habits, and social identity.

The models have caught up. Some campaigns have also caught up, but far too many have not.

Over the past year, one of the most interesting parts of my work has been helping develop a new generation of voter indexes designed to measure characteristics campaigns have historically treated as intuition. Not demographics, but:

• Behavior.

• Persuadability.

• Issue salience.

• Coalition volatility.

• Narrative pressure.

• Behavioral activation.

These aren’t labels. They’re probabilities. They describe how people are likely to behave under changing political conditions, not simply who they are.

I suspect this is where the next decade of campaign strategy is headed.

The implications extend well beyond politics.

Businesses increasingly segment customers by predicted behavior rather than demographics. Financial institutions assess risk through behavioral models. Healthcare systems identify patients based on likelihood of intervention rather than broad population averages.

Politics has been slower to make that transition, but the direction is becoming difficult to ignore.

Campaigns built around averages are slowly giving way to campaigns built around movement. But too slowly, as this “giving way” also gives way to profit margins.

The next generation of well run campaigns won’t simply ask whether a voter is a suburban woman or an independent voter.

They’ll ask whether she is persuadable.

Whether her support is stable or volatile.

Which issues are most salient to her today; not six months, six weeks or six days ago.

Whether she’s becoming more or less likely to act.

Those are fundamentally different questions.

And they require fundamentally different tools.

For most of modern political history, campaigns asked:

Who is this voter?

Increasingly, the more important question is:

What kind of movement is this voter capable of?

The campaigns that answer that question first won’t just have better data.

They’ll have a better map of the electorate itself.

I’ve spent a pretty long time measuring public opinion, working on campaigns, and helping organizations solve difficult problems. One lesson keeps showing up.

The biggest mistakes usually happen when we look through only one lens.

Pollsters become prisoners of the latest survey. Consultants become obsessed with the latest narrative. CEOs become so focused on quarterly earnings that they miss long-term opportunity.

Each lens tells part of the story but none tell the whole story.

As America approaches her 250th birthday, I’ve been thinking about that lesson.

Recent polling has generated plenty of headlines. Gallup found that American pride has reached its lowest level since they began asking the question. The partisan differences are striking. Republicans remain overwhelmingly proud of America. Democrats have become significantly less so. Other research paints a similar picture, while also showing that most Americans remain proud of our history but are less confident about our present and future.

Those numbers matter but they’re only one lens.

Lens One: Perspective

Polling tells us what people are feeling.

That’s incredibly valuable. Feelings shape behavior. They influence elections, markets, hiring decisions, and even whether someone decides to start a business or move across the country.

But polling is also a snapshot.

It captures a moment, not a destiny.

I was reminded of that while working on Glenn Youngkin’s campaign in Virginia.

Republicans hadn’t won a statewide race there in over a decade. If you looked only at the political environment, there were plenty of reasons to conclude the trend was irreversible.

But underneath the frustration was something different. Voters weren’t looking for someone to validate their pessimism. They were looking for someone who could convince them the future could be better than the present. Glenn Youngkin did that in a way I hadn’t witnessed since Reagan.

Perspective mattered.

It just wasn’t the whole story.

Lens Two: Principles

The second lens asks a different question. Not, “How do we feel?” but, “What do we believe?”

Every generation has debated America’s direction. That’s nothing new. What strikes me today isn’t that Americans disagree. We’ve always disagreed.

It’s that we’re increasingly disagreeing about first principles.

Some believe America has drifted away from the ideals that made her exceptional.

Others question whether those ideals were ever exceptional in the first place.

That’s a very different conversation.

History suggests that nations are remarkably resilient when they debate policies.

They’re much less resilient when they stop agreeing on the principles that hold them together.

Companies experience something similar. Healthy organizations constantly improve. They adapt. They innovate. They challenge assumptions. The strongest ones rarely abandon the mission that made them successful.

Lens Three: Purpose

The final lens may be the most important: what are people doing?

This is where I often find myself looking beyond the headlines.

Some people say America is declining yet millions still want to come here. Entrepreneurs continue risking everything to start businesses here. Researchers continue making breakthroughs here. Families continue sacrificing for their children here. Communities continue rebuilding after disasters here.

None of those facts erase the challenges we face.

But behavior often reveals something polling alone cannot. People invest in what they believe has a future. That’s why I remain optimistic.

Not because I think America is perfect. And not because the polling and DSA primary wins don’t concern me. But because I still see millions of Americans quietly building, serving, creating, teaching, volunteering, inventing, and raising families every single day.

Those stories rarely trend on social media. They’re also the stories that have sustained this country for nearly 250 years.

I think that’s the lesson.

Perspective helps us understand the moment. Principles remind us who we are. Purpose determines where we’re going.

A single lens rarely tells the whole story. Neither does a single election. Or a single poll. Or a single generation.

As we celebrate America’s 250th birthday, perhaps the best thing we can do is look through all three.

Happy Independence Day.

He glances briefly at the patient, nods confidently, and proceeds based entirely on experience. No heart rate. No blood pressure. No imaging. No lab work. Just instinct.

The scenario sounds absurd because medicine has evolved. We expect doctors to combine experience with evidence. The most respected surgeons aren’t the ones who ignore data. They’re the ones who know how to interpret it.

Politics, however, remains full of people who proudly do the opposite.

Recently, I watched a statewide campaign spend millions of dollars communicating messages that consistently underperformed. That’s not unusual. What made this situation different was that the campaign wasn’t flying blind. In fact, it had access to more information than most in campaign history.

Polling had identified which messages voters preferred.

Analytics had identified which voters mattered most.

Behavioral measurement was showing, in real time, which communications were helping and which were hurting.

Three different systems. Three different methodologies. Three different lenses.

All pointing in roughly the same direction.

And yet the campaign continued to spend money amplifying messages that were failing.

The problem wasn’t a lack of information. The problem was a refusal to learn.

For most of modern campaign history, consultants had a legitimate excuse. The tools simply didn’t exist. You could run a poll, build an ad, place the buy, and wait. If the campaign won, the ad was considered brilliant. If the campaign lost, it wasn’t. There were few ways to know what was actually happening between those two points.

As a result, successful consultants developed instincts. Some of those instincts were earned over decades. They learned which messages tended to work, which issues tended to matter, and how voters typically behaved. In an environment with limited information, experience often was the best available guide.

The problem is that the environment changed.

Campaigns today don’t operate with one dashboard. They operate with three.

Polling remains the first lens. It tells us what voters say. It helps identify message preference, issue importance, candidate image, and coalition structure. Polling is still valuable, but it only answers certain questions.

The second lens is predictive analytics. Analytics helps identify who matters. Not every voter has the same probability of voting, persuading others, donating, volunteering, or deciding an election. Analytics turns a mass audience into a strategic audience. It tells campaigns where to focus their attention.

The third lens is behavioral measurement. This is the newest and, increasingly, the most important. It doesn’t measure what voters say. It measures what they do. Which messages are gaining traction? Which narratives are resonating? Which issues are becoming more salient? Is support becoming stronger or weaker? Are voters moving toward you or away from you?

Taken together, these three lenses provide a remarkably complete picture of a campaign.

At least they do if you’re willing to look through them.

What’s fascinating is that the biggest resistance to this new environment rarely comes from candidates. Candidates, in my experience, are often surprisingly open to information. They want to win. They want to know what’s working.

The resistance usually comes from people whose authority was built in a different era.

The most dangerous consultant in politics is not the inexperienced one. It’s the consultant who believes experience exempts him from learning.

Everyone in politics has met this person.

He’s the consultant who knows which ad works before it’s tested.

He knows which issue matters before measuring salience.

He knows which voters matter before looking at the analytics.

He knows which message resonates before examining the evidence.

Sometimes he’ll even tell you that regression analysis makes his head hurt (yes, that actually happened).

And occasionally, he’s right. That’s what makes him dangerous.

The consultant who is wrong all the time doesn’t survive very long. The consultant who was right often enough in the past develops something far more powerful than confidence. He develops certainty.

Certainty is difficult to update.

One of the recurring themes in my writings has been the growing importance of volatility, persuadability, salience, and behavioral movement. Those concepts aren’t replacing traditional political strategy. They’re adding another layer to it.

The future campaign advantage won’t belong to the campaign with the best creative. And it won’t belong to the campaign with the biggest budget. And increasingly, it may not even belong to the campaign with the most experienced consultants.

It will belong to the campaign that learns fastest. The campaign that can identify when:

• a message is underperforming,

• an issue is losing relevance,

• a coalition is becoming unstable,

• or a narrative is moving in the wrong direction,

and adjust before competitors do.

That sounds obvious when written on paper. Yet it remains remarkably uncommon in practice.

Why?

Because information creates accountability.

If the data says an ad is failing, someone owns that failure.

If the analytics say you’re targeting the wrong voters, someone owns that decision.

If behavioral measurement shows your message is creating negative movement, someone has to answer for it.

Ignoring the data doesn’t eliminate the problem. It simply delays the reckoning.

Twenty years ago, campaigns had an excuse. They couldn’t see. Today they can.

Polling can tell us what voters say.

Analytics can tell us who matters.

Behavioral measurement can tell us what is actually happening.

The challenge is no longer visibility. The challenge is willingness.

Because the most dangerous person in politics isn’t the consultant who lacks information.

It’s the consultant who already knows the answer.

Then I came home.

Within minutes, I found myself looking at AI-generated political ads, synthetic images being shared as news, campaign videos that had never been filmed, and political content being produced at a volume that would have required an entire communications department just a few years ago.

That was the moment I realized something.

The biggest political story of 2026 may not be AI itself. It may be what artificial intelligence is doing to our ability to recognize what matters.

Today, Oklahoma voters (where I live) are heading to the polls. Across the country, campaigns are already preparing for the next round of primaries, runoffs, and general election fights. Most are focused on how AI can help them create more content, reach more voters, and communicate more efficiently.

I suspect they’re asking the wrong question.

The more important question is whether anyone will still be able to identify genuine political movement once the information environment becomes saturated with synthetic content.

Researchers have recently started using a term for this phenomenon: Slopaganda.

It’s an inelegant word, but it describes an increasingly important reality.

The age of information scarcity is ending. The age of signal scarcity is beginning.

When Content Becomes Free

For most of modern political history, information was expensive.

Television advertising required production budgets (and as anyone doing a campaign this year knows, many still do—and big ones).

Direct mail required graphic artists.

Radio required unique voice over talent or custom music.

Even publishing an opinion generally required access to institutions that controlled distribution.

The challenge was reaching voters.

Campaigns fought relentlessly for attention because attention was difficult to acquire.

Today, the economics have changed entirely.

A campaign intern with access to modern AI tools can produce more content in a single afternoon than many statewide campaigns generated during an entire election cycle twenty years ago.

• Images.

• Videos.

• Memes.

• Articles.

• Social posts.

• Voiceovers.

• Targeted messages.

All created at virtually no marginal cost.

This shift is so profound that we still haven’t fully absorbed what it means.

Most discussions about AI focus on whether the content is accurate.

• Will voters be fooled by deepfakes?

• Will synthetic images spread misinformation?

• Will AI-generated propaganda distort public understanding?

Those are legitimate concerns.

But they are also familiar concerns.

Every major communications technology, from newspapers to radio to television to social media, created new opportunities for misinformation.

What makes AI different is not simply its ability to create false information.

It is its ability to create virtually unlimited information.

That distinction matters and I haven’t seen many campaign professionals talking about it.

The Researchers Saw It Coming

A growing body of academic research is beginning to explore this phenomenon.

Researchers Mark Alfano, Michał Klincewicz, and Amir Ebrahimi Fard recently introduced the term “Slopaganda” to describe AI-generated political content designed to influence beliefs, attitudes, and behavior at scale.

Their argument is straightforward: Propaganda has always existed. What is new is the cost structure.

Historically, creating persuasive political content required resources, expertise, and distribution.

Today, much of that cost has disappeared.

The result is a political environment capable of generating content at a volume that would have been unimaginable only a few years ago.

Meanwhile, researchers at Oxford, Stanford, the UK AI Security Institute, and the London School of Economics have demonstrated that conversational AI can influence real-world political behavior; not merely opinions. In controlled studies, AI systems significantly increased actions such as petition signing and political participation.

That finding should concern political professionals for a different reason than it concerns academics.

The challenge is no longer simply understanding persuasion.

The challenge is understanding persuasion in an environment where the volume of political content is increasing exponentially.

The New Arms Race

Over the past year alone, we’ve seen examples emerge from nearly every corner of the political world.

Governments have used AI-generated imagery as political communication.

Campaigns have experimented with AI-generated advertising.

Researchers and technology companies have identified foreign influence operations using AI-generated content to amplify narratives and shape political conversations.

OpenAI recently disclosed multiple influence campaigns using generative AI to create and distribute political content across digital platforms.

The common thread in all of these examples is not deception; it is scale.

The ability to generate political communication faster than human systems can evaluate it.

And that creates a problem most campaigns are not prepared for.

The Signal Problem

Imagine you’re managing a campaign in 2012:

• A message starts spreading online.

• An issue begins gaining traction.

• A narrative starts appearing in earned media.

The reasonable assumption is that something meaningful is happening beneath the surface:

• Voters care.

• Activists are engaged.

• Opinion is shifting.

Today, those same signals are becoming harder to interpret. A narrative may be gaining traction because:

• voters genuinely care about it;

• activists (or paid influencers) are amplifying it;

• journalists are covering it;

• algorithms (or paid bots) are promoting it;

• or AI-generated content networks are flooding the information environment.

Increasingly, all five may be happening simultaneously.

That changes the nature of political analysis. The challenge is no longer gathering information. The challenge is identifying which information actually matters.

In other words, the scarce resource is no longer content. It is signal.

This Is Where Campaigns Get Into Trouble

One of the recurring themes in my writings has been the distinction between movement and measurement.

Campaigns routinely confuse:

• attention with persuasion,

• persuasion with activation,

• and activation with voter movement.

The Slopaganda Election makes that problem even harder.

Because AI-generated content can create enormous amounts of attention without necessarily creating persuasion. Likewise, genuine voter movement can sometimes appear small or insignificant inside a flood of synthetic content.

The danger isn’t simply that voters will believe things that aren’t true. The danger is that campaigns will increasingly struggle to identify what is actually driving behavior.

That’s a very different problem. And arguably a more dangerous one.

What Comes Next

Since joining EyesOver after the 2024 election cycle, one of the most interesting developments I’ve watched has been the growing importance of distinguishing movement from noise.

Not simply measuring:

• sentiment,

• awareness,

• volume,

• or engagement.

But identifying:

• salience,

• volatility,

• behavioral shifts,

• and genuine changes in public opinion.

Because as AI continues flooding the information environment, those distinctions become increasingly valuable.

The future advantage will not belong to the campaign that creates the most content. It may not even belong to the campaign with the best AI. It will belong to the campaign that most accurately identifies genuine signal while everyone else is drowning in noise.

The Bottom Line

Most people think the danger of AI in politics is misinformation. I increasingly think the larger challenge is signal collapse.

For two centuries, campaigns operated in a world where information was scarce and attention was abundant enough to be captured.

Today, information is becoming effectively unlimited.

What is becoming scarce is our ability to determine what actually matters. In the past, campaigns fought information scarcity. In the future, they will fight signal scarcity.

And that may prove to be the defining political challenge of the AI era.

• The average Republican.

• The average suburban woman.

• The average independent voter.

• The average Hispanic voter.

• The average evangelical.

• The average swing voter.

Entire campaign strategies, media plans, mail programs, polling plans, digital targeting, and messaging operations have been built around the assumption that these groups exist and can be understood through broad demographic categories.

There’s a big problem with that: The average voter doesn’t exist.

Or at least not in the way most campaigns think.

And as artificial intelligence, predictive analytics, and behavioral modeling (combined into a powerful tool) continue to reshape politics, campaigns that still rely primarily on demographic averages are finding themselves at a growing disadvantage.

Because the future belongs to campaigns that understand behavioral clusters, not demographic categories.

A Lesson From Texas

One of the most valuable lessons I learned about voter behavior came from a statewide research project several years ago.

At the time, conventional wisdom suggested suburban women were being driven primarily by the same issues dominating national political coverage.

• Healthcare.

• Education.

• Taxes.

• Partisan politics.

Those issues mattered. But when we dug deeper into the data (and first learned this in a focus group in The Woodlands), another issue kept appearing among a particular segment of suburban voters: Red-light traffic cameras.

To many political observers, that would have sounded ridiculous. How could traffic cameras possibly matter in a major statewide election? But the data suggested something deeper was happening. These voters weren’t reacting to the cameras themselves. They were reacting to what the cameras represented:

• Government overreach.

• Fairness.

• Local accountability.

The feeling that citizens were being treated as revenue sources rather than constituents. The issue carried far more emotional intensity than most observers realized. More importantly, it revealed something larger: voters who looked identical demographically often behaved completely differently politically.

Two suburban women of the same age, income, education level, and party registration could have dramatically different political motivations.

The average voter model could not explain that (but the data could).

The issue eventually gained enough traction that Texas outlawed red-light traffic cameras statewide in 2019.

But the larger lesson wasn’t about traffic cameras; it was about discovery.

The data identified a meaningful behavioral signal long before most political observers considered it politically important.

The Problem With Demographics

Traditional political analysis groups voters into broad buckets.

• Age.

• Race.

• Gender.

• Geography.

• Party registration.

Those variables still matter, but increasingly they are insufficient. Because people are not simply collections of demographic traits.

They are collections of motivations:

• Some voters are motivated by economic anxiety.

• Others by cultural identity.

• Others by institutional trust.

• Others by personal freedom.

• Others by fairness.

• Others by status.

• Others by belonging.

Two voters who look identical on paper can respond to entirely different messages. And two voters who appear completely different demographically may respond to exactly the same message.

That reality is becoming impossible to ignore.

The Rise of Behavioral Clusters

When I first entered politics (I won’t list the year), segmentation was relatively simple: Republican versus Democrat; urban versus rural; male versus female.

Today, the most sophisticated campaigns increasingly think differently.

They focus on:

• turnout probability

• issue sensitivity

• persuasion likelihood

• coalition alignment

• emotional activation

• behavioral volatility

The question is no longer: Who is this voter?

The question is: What kind of voter is this?

That distinction is subtle, but it changes everything.

What Data Science Sees That Humans Miss

One of the most important developments in modern politics is the ability to identify patterns that are nearly impossible to detect intuitively.

Using large-scale data systems, machine learning models, and behavioral datasets, analysts can now identify relationships that would have been invisible a decade ago.

Increasingly, we are building entirely new voter indexes designed to measure things that traditional demographics struggle to capture:

• persuadability

• issue salience

• sentiment volatility

• narrative alignment

• activation potential

These measures often tell us more about future behavior than age, gender, or party registration ever could.

That doesn’t mean demographics are dead, but it means they are no longer enough.

Why This Matters In 2026

This is especially important in the current midterm environment.

Most campaigns will continue targeting demographic categories.

• Suburban women.

• Independent voters.

• Young voters.

• College-educated voters.

But the campaigns that gain an advantage will increasingly identify the behavioral clusters hidden inside those categories.

• Because not all suburban women think alike.

• Not all independents behave alike.

• Not all Republicans respond to the same messages.

The campaigns that recognize those differences first will possess an enormous strategic advantage.

The Future Of Campaigns

Artificial intelligence is accelerating this shift. AI doesn’t think in demographic averages.

• It identifies patterns.

• Correlations.

• Clusters.

• Relationships.

In many ways, AI is forcing campaigns to abandon one of the oldest assumptions in politics: The idea that voters can be understood primarily through broad demographic labels.

Instead, campaigns are increasingly moving toward a world where understanding behavior matters more than understanding demographics.

That is a profound change.

And it may ultimately prove more important than any individual technological breakthrough.

The Bottom Line

For most of modern political history, campaigns asked: Who is this voter?

Increasingly, the better question is: What kind of voter is this?

Because the average voter is disappearing.

And campaigns that recognize that first will possess one of the most powerful advantages in politics.

Most analysis of races like this still focuses on:

• polling margins,

• spending decisions (campaign and outside),

• endorsements,

• advertising,

• consulting teams,

• and media coverage.

Those things matter (some a lot more than others).

But increasingly, they are not the most important variables.

The more important question is: How durable is the coalition underneath the numbers?

Because two candidates can poll similarly and possess completely different levels of support stability.

One coalition may be hardening. The other may already be quietly fracturing underneath the surface.

Traditional polling often treats those situations as identical. They are not even remotely the same.

The Biggest Mistake Campaigns Make

Most campaigns assume support is uniform. It isn’t.

A candidate sitting at:

• 46%,

• solid favorables,

• strong name ID,

• and acceptable job approval

can still be sitting on an extraordinarily fragile coalition.

Another candidate with similar toplines may possess:

• emotionally durable support,

• activist intensity,

• stronger identity alignment,

• and lower volatility.

Those are very different political positions.

But topline polling often struggles to distinguish between them. That is where many so-called “surprise” outcomes come from. Not polling failure.

Volatility failure.

Polling Measures Support. Volatility Measures Fragility.

This distinction matters enormously. Polling generally measures:

• expressed preference,

• current position,

• and declared support.

Volatility measures something different: How easily that support can break, shift, or accelerate under pressure.

And in modern politics, pressure arrives faster than ever. Because today’s campaigns operate inside:

• compressed news cycles,

• fragmented media ecosystems,

• algorithmic amplification,

• and emotionally charged narrative environments.

That changes how coalitions behave. Support is no longer static. It is increasingly conditional.

Some Coalitions Harden Under Pressure. Others Fracture.

This is one of the most misunderstood dynamics in politics.

Some political coalitions become stronger when attacked. Criticism reinforces:

• identity,

• tribal attachment,

• emotional commitment,

• and turnout intensity.

Other coalitions do the opposite. Under pressure, they soften:

• enthusiasm drops,

• volatility increases,

• turnout weakens,

• and persuadable voters begin drifting.

The problem is that both coalitions can initially look very similar in polling. That is why campaigns that appear stable sometimes collapse rapidly late in races. The warning signs often existed long before the toplines moved.

We’ve Seen This Repeatedly in Recent Cycles

In the 2024 Texas Republican primaries, several incumbents entered the cycle with:

• structural advantages,

• strong fundraising,

• institutional support,

• and seemingly stable positions.

Many still lost. Why? Because underneath the toplines:

• ideological intensity was shifting,

• anti-establishment pressure was building,

• issue salience was changing,

• and coalition volatility was rising.

The campaigns that recognized those conditions early adapted. The campaigns relying solely on traditional assumptions often reacted too late.

That pattern has repeated itself across modern politics:

• late movement,

• compressed persuasion windows,

• rapid coalition shifts,

• and turnout volatility among voters traditional polling often underestimates.

Those outcomes are rarely random. Most are signs of movement building underneath the surface long before Election Day.

AI Is About to Accelerate Volatility

This dynamic becomes even more important in the AI era.

A recent study from researchers at Oxford, Stanford, the UK AI Security Institute, and the London School of Economics found that conversational AI significantly increased real-world political behavior and activism. (If I have any “regular readers” I recognize this is the third time I’ve written about this study, but it is stunning to me it’s not getting more attention as it is truly revolutionary to how campaigns should be operating).

But the most important implication may not be persuasion itself. It is speed.

AI accelerates:

• engagement,

• emotional activation,

• issue amplification,

• identity reinforcement,

• and behavioral movement.

That means coalition volatility may now develop faster than traditional systems can fully detect.

In other words: Political coalitions are becoming more dynamic precisely as campaigns become more dependent on static measurement systems.

That is a dangerous combination.

This Matters To Me Personally

One of the reasons I joined EyesOver as CEO in January 2025 was because I saw firsthand during the 2024 cycle how useful real-time sentiment and movement analysis had become in understanding races traditional systems struggled to fully explain.

Since taking over, one of our major focuses has been evolving the platform beyond simple sentiment tracking and building tools designed to better identify:

• volatility,

• salience shifts,

• movement conditions,

• and coalition instability in real time.

Because increasingly, the most important strategic question in campaigns is no longer simply: “Where is the race?”

It is: “How stable is the race underneath the surface?”

That distinction matters more every cycle.

The Real Battlefield

The most important number in modern politics may no longer be: “Who is ahead?”

It may be: “Whose support is more likely to move?”

That is a fundamentally different way of thinking about campaigns.

Because the future divide will not simply be between:

• Republicans and Democrats,

• digital and television,

• or AI and traditional campaigning.

It will increasingly be between campaigns that understand volatility and campaigns that do not.

The winning campaigns will be the ones that can distinguish:

• stable support from fragile support,

• attention from activation,

• and polling position from movement probability.

That is the real battlefield now.

The Bottom Line

As Republicans watch the Paxton versus Cornyn race unfold tonight, the broader lesson extends far beyond this single runoff.

Modern campaigns are no longer operating in stable political environments.

They are operating inside rapidly shifting emotional, behavioral, and informational systems.

And in that world: Polling measures support. Volatility measures durability.

The campaigns that win will increasingly be the campaigns that understand the difference before everyone else does.

They assume the battlefield is filled with:

• undecided voters,

• loosely attached partisans,

• and people waiting to hear the right message.

That assumption drives almost every modern campaign:

• media plans,

• persuasion universes,

• message testing,

• digital targeting,

• direct mail,

• and turnout strategy.

The problem is that much of it is built on a flawed premise. Most undecided voters are not persuadable. And many persuadable voters do not look undecided at all.

That is the persuadability illusion.

The Real Battlefield Is Tiny

A recent report from Cross Screen Media estimated that only about 8% of adults are persuadable, and only around 1.7% of voters may ultimately decide competitive races. I wrote about it here:

That is an astonishingly small slice of the electorate.

And yet campaigns still routinely build persuasion programs as though large portions of voters are fluid. They aren’t. Most voters are:

• structurally partisan,

• emotionally locked in,

• habitually disengaged,

• or simply unreachable.

The real opportunity exists in a much narrower segment:

• voters who are lower propensity,

• high-affinity,

• and movable under the right conditions.

That distinction matters enormously. Because campaigns waste extraordinary amounts of money trying to persuade people who were never realistically going to move in the first place.

Undecided Does Not Mean Persuadable

This is one of the biggest mistakes in modern politics. Campaigns see:

• low information,

• weak engagement,

• or undecided responses

and assume persuasion opportunity.

But many of those voters:

• do not follow politics closely,

• are unlikely to participate,

• or are psychologically disconnected from the race altogether.

Meanwhile, some of the most persuadable voters in an election already appear “decided” in polling. Their support looks stable on the surface. But underneath, it is volatile.

That is where late breaks happen.

And it is why campaigns so often act shocked by outcomes that were visible long before Election Day. Polling measures position. Winning campaigns focus on movement probability.

Those are very different things.

This Is Where Predictive Analytics Changes Everything

Of course, all of this assumes your campaign is already using predictive analytics sophisticated enough to identify:

• low-turnout voters,

• high-affinity voters,

• turnout likelihood,

• support probability,

• and activation potential.

That should not be viewed as exotic anymore. It should be table stakes (not to go all Nate Silver, as I don’t play poker, but did enjoy his last book). Because the most valuable voter in modern politics is often not:

• the undecided voter,

• or the high-information swing voter.

It is the voter:

• unlikely to participate,

• but disproportionately likely to support your candidate if properly activated.

That is where elections are increasingly won. Especially in primaries.

We’ve Seen This Repeatedly in Recent Cycles

In several recent Republican primaries, campaigns with massive spending advantages and stable polling leads still lost.

Why?

Because they misunderstood the actual persuasion universe.

In the 2024 Texas Republican primaries (and to a lesser extent in 2026), multiple incumbents appeared structurally secure until late movement developed underneath the surface.

The campaigns that adapted quickly:

• identified persuadable clusters earlier,

• recognized issue salience shifts faster,

• and reallocated communication resources before the broader political environment fully recognized what was happening.

The same dynamic has appeared repeatedly in modern campaigns:

• late-cycle movement,

• compressed persuasion windows,

• and turnout activation among lower-propensity voters who traditional polling often underestimates.

That is not random volatility. It is usually a targeting and movement problem.

AI Is About to Magnify This Gap

This becomes even more important in the AI era.

A recent study from researchers at Oxford, Stanford, the UK AI Security Institute, and the London School of Economics found that conversational AI significantly increased real-world political behavior and activism. (I wrote about it here):

But the key insight was deeper than the technology itself. The researchers found that: changes in political attitudes and changes in political behavior were often uncorrelated.

That finding reinforces something campaigns are beginning to confront: Persuasion is not broad. It is conditional. Behavioral. Identity-driven.

And highly uneven across the electorate.

Which means AI will not magically solve persuasion.

It will simply make sophisticated campaigns more efficient at identifying and activating the right voters. And it will make unsophisticated campaigns more efficient at wasting money on the wrong ones.

The New Campaign Divide

The future divide in campaigns will not simply be:

• Democrat vs Republican,

• digital vs television,

• or AI vs traditional media.

It will increasingly be between campaigns that understand behavioral movement and campaigns that do not.

The winning campaigns will not necessarily persuade more voters overall. They will:

• identify the right persuadable voters earlier,

• understand what conditions drive movement,

• and recognize activation opportunities before their opponents do.

That is the real battlefield now. And it is much smaller than most people think.

The Bottom Line

Most campaigns still think persuasion is a mass-market exercise. It isn’t.

The future belongs to campaigns that understand:

• who can move,

• who will actually vote,

• and what conditions trigger action.

Because elections are not won by persuading everyone. They are won by identifying the small number of voters who actually matter—and moving first.

• Did favorability move?

• Did ballot numbers shift?

• Did the message test improve?

But a new AI persuasion study suggests campaigns may be measuring the wrong thing entirely. Because changing someone’s mind and changing someone’s behavior are not the same thing.

And in politics, behavior is what actually matters.

This new study from researchers at Oxford, Stanford, the UK AI Security Institute, and the London School of Economics tested whether conversational AI could persuade people not just to change opinions, but to take real political actions.

The results should get every campaign strategist’s attention.

AI Produced Real Political Action

The researchers ran two large experiments involving nearly 15,000 participants and measured real-world behavioral outcomes, including:

• signing political petitions,

• donating money,

• supporting organizations,

• and participating in activism-related tasks.

The AI systems weren’t just persuasive (creating ads is cute an all). They were behaviorally effective.

In one study, AI-driven conversations increased petition signing by nearly 20 percentage points.

That is meaningful, and unusual political movement.

And importantly, the conversations were short. On average, participants interacted with the AI for roughly seven minutes. The implications here are enormous.

The Industry’s Blind Spot

Most political persuasion research—and frankly, most campaigns—still optimize around attitudinal change.

Can we move:

• favorability,

• issue agreement,

• or message perception?

But the most important finding in the paper may not have been the size of the effects.

It was this: AI persuasion effects on attitudes and behavior were largely uncorrelated.

Read that again.

The messages that changed opinions were often not the messages that changed behavior.

That distinction matters far more than most general consultants or pollsters realize.

Because elections are not decided by attitudes. They are decided by:

• turnout,

• persuasion,

• donations,

• volunteering,

• activation,

• and ultimately behavior.

Polling measures position. Campaigns are won and lost on movement. This is the third time I’ve written about this (the prior two based on other academic research).

At some point campaigns will wake up and pay attention. I pray it’s the GOP first as we have a tool we can use: Sparkfire.

We Saw This Years Ago

During the 2016 Iowa Republican caucuses, our team was not simply monitoring topline movement. We were tracking how specific messages affected specific voter segments in real time.

As Sasha Issenberg detailed in his Bloomberg piece How Ted Cruz Engineered His Iowa Triumph, the campaign ultimately built more than 150 distinct voter segments in Iowa alone.

Those segments weren’t just ideological. They included:

• personality types,

• behavioral profiles,

• turnout likelihood,

• issue-specific motivation,

• and hyper-local concerns.

One example stood out: Our analytics and targeting work identified a tiny segment of roughly sixty Iowa Republicans who statistical models suggested could be moved by the issue of legalizing fireworks sales in Iowa.

At the time, that sounded ridiculous to many people. It became national news.

But the logic was simple: Campaigns are not won in averages. They are won in pockets of movement.

As Issenberg reported, we tested dozens of local issues through Facebook experiments, measuring behavioral engagement and click-through response before integrating the strongest signals into microtargeting models.

The campaign then adapted:

• phone scripts,

• digital creative,

• volunteer outreach,

• and turnout operations

based on those behavioral insights.

This was precision activation (not really persuasion).

And importantly, it reinforced something most campaigns still struggle to understand: Different voters are moved by different things at different times.

That observation is even more important in the AI era than it was in 2016.

Facts Changed Attitudes. Emotion Changed Behavior.

One of the most fascinating parts of the new AI persuasion study was what actually drove behavioral change.

The researchers tested multiple persuasion strategies:

• emotional activation,

• identity labeling,

• commitment escalation,

• implementation intentions,

• informational persuasion,

• and combined “mega” persuasion frameworks.

The informational approaches performed well at changing attitudes.

But they were weaker at changing behavior.

The strongest behavioral effects came from:

• emotional activation,

• commitment structures,

• identity reinforcement,

• and implementation planning.

In other words: Facts changed opinions. Identity and emotion changed action.

That observation aligns with what many campaigns quietly discover the hard way. Most voters do not move because they received new information. They move because:

• an issue became salient,

• a message resonated emotionally,

• a social identity activated,

• or a condition shifted beneath the surface.

That is why sentiment analysis matters so much more than most campaigns understand.

AI Is About to Scale Behavioral Activation

Most people think AI in campaigns means:

• faster content,

• better targeting,

• automated responses,

• or synthetic media.

That is the shallow version of what is happening.

The deeper reality is this: AI doesn’t just scale communication. It scales behavioral activation. That is a fundamentally different thing.

The paper repeatedly suggests that exposure and engagement matter more than factual persuasion alone.

That means future campaigns will increasingly compete on:

• interaction,

• personalization,

• emotional sequencing,

• and real-time activation dynamics.

Not simply broadcasting messages.

And in a world where the true persuadable universe may only be 1–2% of the electorate, as the recent Cross Screen Media report suggested, the stakes become even higher.

Because if only a tiny portion of voters actually matter…

understanding behavioral movement becomes the entire campaign.

The New Campaign Divide

The campaigns of the next cycle will split into two categories.

The first group will still operate traditionally:

• polling,

• media buys,

• static targeting,

• message testing,

• broad persuasion assumptions.

The second group will operate differently. They will focus on:

• behavioral movement,

• real-time persuasion dynamics,

• sentiment volatility,

• issue salience,

• and activation sequencing.

The difference between those two models will define modern campaigning.

Because the future campaign advantage will not belong to the campaign with the best messaging. It will belong to the campaign that best understands:

• who can move,

• what will move them,

• and when that movement is beginning to happen.

The Bottom Line

Most campaigns are still trying to win the argument.

The next generation of campaigns will focus on something much more important: Triggering action.

Because elections are not ultimately decided by what voters think.

They are decided by what voters do.

• more reach = more persuasion “we need more GRPs!”

• more content = more influence “we need more ads!”

That’s not how voters move.

The Biggest Mistake in Modern Campaigns

The data show most campaigns are measuring the wrong thing. They measure:

• support

• favorability

• ballot position

Those are outcomes, not causes.

Voters don’t move because they’re reached. They move because the conditions are right.

Movement Is Uneven; And That’s the Whole Game

A recent report from Cross Screen Media made this clear in a different way.

• Only ~8% of adults are persuadable

• Only ~1.7% are persuadable in competitive races

That means most voters aren’t moving at all.

And the ones who do move:

• don’t move evenly

• don’t move at the same time

• and don’t respond to the same inputs

Political science backs this up. Large-scale campaign experiment research (e.g., American Political Science Review) consistently finds that persuasion effects are heterogeneous and context-dependent.

In plain terms:

Some voters move, BUT most don’t.
Some messages land, BUT most don’t.

We’ve Seen This Up Close

During the 2016 Iowa Republican caucuses, I ran data and polling for Ted Cruz. We weren’t just tracking topline support. We were monitoring how specific messages moved specific segments of voters in real time.

One example stood out.

A message tied to the legalization of fireworks—a niche issue by any standard—produced measurable movement within a distinct segment of caucus voters.

Not broadly. Not across the electorate. But within a very specific group where that issue mattered.

That kind of movement would never show up clearly in a traditional poll.

But it was real and it mattered.

As Sasha Issenberg has documented in his reporting our data-driven campaign, modern campaigns that win are increasingly those that understand voter behavior at a granular level, mapping how specific messages activate specific audiences rather than assuming uniform effects.

AI Is Accelerating the Wrong Layer

The industry is already moving to the next phase.

As highlighted in this recent New York Times piece, campaigns are rapidly adopting artificial intelligence to:

• generate content

• personalize outreach

• and engage voters at scale

That shift is real and it’s happening fast. But it doesn’t solve the core problem. AI improves how campaigns act. It does not improve what they know. If anything, it makes the gap more dangerous. Because now campaigns can:

• produce more content

• target more voters

• engage more frequently

without improving their understanding of:

• who is persuadable (a real GOP media consultant recently said on a call that regression analysis made his “brain hurt”)

• what issues are driving movement

• when sentiment is shifting

AI doesn’t fix bad targeting. It scales it.

The Four Drivers of Voter Movement

Voter movement isn’t random. It is driven by conditions that can be observed—and measured.

1. Persuadability

Who can still move. Not who supports you or who agrees with you already. Who is open to changing their position. Most campaigns assume this is broad, but it isn’t.

2. Issue Salience

What voters care about right now. Not what tested well. Not what worked last month. What is actively shaping attention in the current moment.

3. Message Resonance

What sticks and spreads. Most messages don’t fail because they’re wrong. They fail because they don’t propagate.

4. Sentiment Volatility

How stable support really is. Some voters are locked in. Obviously others are not. That difference is where “unexpected” outcomes come from.

Most Campaigns Miss This

Because their systems aren’t built to detect it. They have:

• polling

• media plans

• messaging

But those systems are:

• static

• siloed

• backward-looking

They tell you where the race is. They don’t tell you where it’s going.

From Measurement to Advantage

The campaigns that win don’t just:

• produce more

• spend more

• or poll more

They detect movement earlier. They understand:

• which voters can move

• what will move them

• and when that movement is starting

Then they act.

And This Matters More in 2026

Everything is amplifying this dynamic:

• Smaller persuadable universes

• Fragmented media environments

• Faster narrative cycles

• More late decision-making

Cross Screen’s data shows just how small the true battlefield is.

Which means missing movement isn’t a small mistake. It’s the entire race.

The Bottom Line

Campaigns don’t lose because they fail to reach voters. They lose because they don’t understand what actually causes voters to move.

They:

• measure position instead of conditions

• optimize for efficiency instead of change

• react after the movement has already happened

In 2026, that won’t be enough. The campaigns that win won’t just reach voters.

They’ll understand when voters are ready to move, and act first.

Final Line

Voters don’t move because they’re reached. They move because the conditions are right.

Linear. CTV. Digital. Mail. Texting. Radio.

Then everyone argues about percentages.

That is accounting with political vocabulary and not strategy.

The real question is not: “How should we allocate the budget?”

The real question is: “Where is voter movement possible, and what channel gives us the best chance to create it?”

That is a very different question. And most campaigns are simply failing to answer it.

The Old Model Is Breaking

For years, campaign media planning followed a familiar rhythm:

1. Poll the race

2. Build the target universe

3. Allocate media

4. Launch the buy

5. Measure later

That model made more sense when voter attention was less fragmented.

That world is gone. A recent report from Cross Screen Media (which I also wrote about last week) makes the shift clear:

• Only about 8% of U.S. adults are both likely voters and persuadable

• Only ~1.7% are persuadable in competitive races

That means elections are not being decided by “the electorate.”

They are being decided by a tiny, fragmented slice of voters.

And those voters are harder to reach than ever:

• 22% are only reachable via CTV

• 14% are unreachable via both linear TV and CTV

This is not just fragmentation. It is structural fragmentation.

Reach Is Not the Same as Persuasion

Political science has been pointing to this gap for years.

A large-scale 2024 analysis of campaign experiments published in the American Political Science Review found that persuasive effects are real but highly variable and context-dependent, and that campaigns are generally poor at predicting in advance which messages will work.

That should be a warning.

Because most media plans assume persuasion is uniform. It is not.

Some voters move. Most don’t.

Some messages land. Most don’t.

Some moments matter. Most don’t.

The Channel Is Not the Strategy

This is where campaigns get lost.

They debate:

• Linear vs CTV

• Broadcast vs digital

• Paid vs organic

But none of those decisions answer the core strategic question.

Channels do not create persuasion. They only deliver it.

We are already seeing the next phase of this play out.

In Wisconsin, more than 1 million voters were contacted by an AI chatbot ahead of the 2026 Supreme Court race, with thousands engaging in direct conversation.

That is the future: One-to-one persuasion at scale.

I’ve written about the version we have on the right that does the same thing (and if you’re not using this method over voter persuasion, your opponent probably is).

But it does not solve the core problem. It amplifies it.

Because if you don’t know:

• who is persuadable,

• what they care about,

• and when they are moving,

then AI just helps you talk to the wrong voters faster.

The Better Model

The old model is: Plan → Buy → Execute → Measure

The new model has to be: Detect → Decide → Deploy → Reallocate

This is not theoretical. It is how high-performing systems operate everywhere else.

In digital marketing, companies do not lock budgets and walk away.

They optimize continuously based on:

• conversion probability

• behavioral signals

• real-time feedback

Campaigns are just starting to catch up thanks to companies like Cross Screen.

Why This Matters More in 2026 Primaries

Primaries amplify every weakness in the system.

• Smaller electorates

• Higher volatility

• Faster issue shifts

• More late decision-making

Political science research continues to show that late deciders play an outsized role in close elections, and that many voters remain open to influence much later than campaigns assume.

That makes static media plans especially dangerous.

Because they are built on assumptions that may already be outdated.

From Efficiency to Advantage

Most media plans optimize for efficiency: “How do we reach our target voters at the lowest cost?”

Winning campaigns optimize for something else: “Where can we create movement, and how quickly can we act on it?”

That requires connecting:

• Reach (where voters are)

• Persuadability (who can move)

• Salience (what matters now)

• Resonance (what sticks)

• Volatility (how stable support is)

That is a decision system and not a media plan.

Conclusion

Most media plans are structurally wrong because they are built around channels instead of voter movement.

They:

• allocate before understanding

• execute before detecting

• measure delivery instead of persuasion

And in a cycle where ~2% of voters decide everything, that is not a small mistake. It could be the entire game.

Campaigns in 2026 need to operate like full-time production studios—creating content constantly, across every platform, every day.

That’s where the world has gone.

But that’s not what will decide elections.

Because Elections Aren’t Decided by Everyone

They’re decided by almost no one.

A new report from Cross Screen Media lays this out more clearly than anything I’ve seen this cycle:

Roughly 1.7% of adults will decide control of Congress in 2026.

Let that sink in.

Campaigns are scaling content to reach millions…

…when the outcome is determined by a tiny, fragmented slice of the electorate.

The Industry Is Solving the Wrong Problem

Plouffe’s model is about production and distribution:

• More content

• More platforms

• Faster execution

Cross Screen’s data is about reachability:

• Where persuadable voters can be found

• How media consumption is shifting

• Which channels matter in which states

Both are right.

But they’re solving different halves of the same problem.

And the missing half is the one that actually decides outcomes.

The Real Constraint Isn’t Reach. It’s Precision.

Cross Screen shows that:

• Only about 8% of adults are persuadable

• Only about 2% are persuadable in competitive races

• 22% of persuadable voters are only reachable via CTV

• And 14% can’t be reached via TV or CTV at all

That’s a targeting and timing problem, not a reach problem. All the media consultants screaming for more GRPs are missing the target.

Campaigns aren’t losing because they aren’t loud enough. They’re losing because they don’t know:

• Which voters in that 2% actually matter

• Which ones are still movable

• And what will move them right now

Most Voters Aren’t Persuadable

This is the part campaigns still struggle with: a large share of the electorate is already locked in (even in primaries)

But not all of it.

And the difference between those two groups is every election is decided.

The challenge is that most campaigns don’t know which is which.

So they do what campaigns have always done:

They talk to everyone; and in doing so, they waste time (and money) on voters who will never move, and miss the ones who will.

From Content → to Movement

This is where the model needs to change.

The future isn’t just:

Produce more content
Distribute it everywhere

It’s:

Identify movement. Then act on it faster than your opponent.

That requires a different set of capabilities:

• Persuadability — Who can still move

• Issue salience — What matters right now

• Message resonance — What actually spreads

• Sentiment volatility — How stable support really is

Cross Screen helps answer:

Where can I reach them?

The next question is:

Who should I reach—and when does it matter most?

This Matters Even More in 2026 Primaries

General elections, even in non-prez years, are about scale.

Primaries are about precision.

• Smaller electorates

• Faster narrative shifts

• Higher volatility

• Late movement that actually matters

In that environment, reaching the wrong voters efficiently is still losing.

The New Operating Model

The campaigns that win in 2026 will combine both sides of this equation:

1. Reach (Cross Screen’s world)

• Channel mix

• Media consumption

• Platform strategy

2. Movement (what most campaigns are missing)

• Real-time sentiment

• Persuadability modeling

• Narrative detection

• Dynamic reallocation

Put together, that becomes:

Reach × Precision × Timing

That’s the game now.

The Bottom Line

Campaigns are becoming media companies; that’s true.

But media companies don’t win elections.

Campaigns do.

And campaigns win by understanding something much more fundamental:

• Not just where voters are

• But which ones matter

• And when they’re about to move

One Final Thought

Cross Screen’s report should be required reading for anyone building a media plan in 2026.

It’s one of the clearest looks at how fragmented, and how high-stakes, the reach environment has become (and it’s broken down by state).

The next step is connecting that reach to real-time voter movement.

Because in a world where 2% decides everything…

The advantage doesn’t go to the campaign that reaches the most people.

It goes to the one that reaches the right people—
with the right message—
at the right moment.

He’s right. But he’s also incomplete.

Because becoming a content machine doesn’t solve the core problem campaigns face.

It amplifies it.

The Real Shift Isn’t Content; It’s Control of Narrative

Plouffe’s core argument is that campaigns must:

• Produce more content

• Distribute across more platforms

• Move faster in execution

• Embed messages into the information ecosystem (including AI responses)

While that’s all true, here’s what’s missing:

What determines whether that content actually moves voters?

Because most content doesn’t. In fact, most content dies on contact.

The Hidden Constraint: Most Voters Aren’t Persuadable

Campaigns talk about persuasion constantly.

But political science, and recent campaign experience, suggest something different:

• Many (if not most) voters are already locked in

• Movement is uneven and concentrated

• Late shifts are driven by specific segments, not the whole electorate

Which means the real problem isn’t:

“How do we create more content?”

It’s:

“Which voters can still move, and what will move them?”

Why the “Production Studio” Model Falls Short

Plouffe argues campaigns should “harmonize” messaging across every channel and produce content continuously. I agree. But, without intelligence, that becomes:

• More content

• More spend

• More noise

Not more persuasion.

This is the same mistake campaigns have always made, just faster.

From Content Volume → to Sentiment Intelligence

The campaigns that win in 2026 won’t just produce more.

They will operate on a different model:

1. Measure Persuadability (Not Just Support)

Not who supports you. But who can still move.

2. Track Issue Salience in Real Time

Not what voters think. But what they care about right now.

3. Identify Message Resonance

Not what you say. But what spreads and compounds.

4. Monitor Sentiment Volatility

Not how strong your support looks. But how stable it is.

Polling shows where the race is (scoreboard).

Sentiment analysis shows where it’s going (momentum on the field).

What Plouffe Gets Exactly Right

There’s one part of the essay that is especially important:

Campaigns need to “flatten and speed up the road from content idea to execution.”

That’s critical.

Because once you identify movement…

Speed becomes the advantage.

But Speed Without Direction Is Useless

A campaign that produces content faster than its competitors, but doesn’t know:

• Which voters are persuadable

• Which issues are breaking

• Which messages are landing

…is just accelerating inefficiency.

What This Means for 2026 Campaigns

Primaries in particular expose this gap.

Why?

Because:

• Electorates are smaller

• Movement is faster

• Late shifts matter more

• A small persuadable segment often decides the race

In that environment:

The campaign that identifies movement first, and reallocates fastest, wins.

Not the one that produces the most content.

The New Operating Model

Plouffe is describing the future of execution, but execution is only half the system.

The winning model looks like this:

1. Detect movement (sentiment + salience + persuadability)

2. Identify opportunity (who can move and why)

3. Deploy content (targeted, not generic)

4. Reallocate resources in real time

That’s how campaigns stop reacting to the race…

…and start shaping it.

The Bottom Line

Campaigns are becoming media companies. That’s true.

But the campaigns that win won’t just be better at producing content.

They’ll be better at knowing:

• What matters

• Who it matters to

• And when it matters most

Because in modern campaigns:

The advantage won’t go to the loudest campaign or candidate.

It will go to the campaign that sees movement first, and acts on it.

The problem is campaigns don’t know which ones are.

So they do what campaigns have always done: talk to everyone.

And in doing so, they waste time on voters who will never move, and miss the ones who will.

That distinction matters in every election. As someone who seems to work for conservatives (who are often underfunded), it matters even more.

Primary electorates are smaller, more volatile, and more sensitive to late shifts in salience, candidate contrast, and turnout composition. In that environment, the most important strategic question is not, “Where is the race today?” It is, “Which voters are still probabilistically movable, and under what conditions?” That is a predictive analytics problem, not just a polling problem.

Political science has been moving in this direction, even if most campaigns have not.

A 2024 PNAS paper by David Broockman and coauthors found that political practitioners are not especially good at predicting which messages will persuade voters. In plain English: consultant intuition is much less reliable than the industry likes to believe.

A 2024 American Political Science Review article examining 146 campaign ad experiments reached a related conclusion. It found small but meaningful variation in persuasive effects across ads, and it found that common theories about what makes an ad persuasive have limited and context-dependent power to predict what will work. In other words, the effect is real, but you usually do not know it in advance unless you test and model it.

That should be a warning to anyone running a 2026 primary.

Most campaigns still talk about persuasion as if it were a broad creative exercise. It is not. Persuasion is a probabilistic exercise. The real task is identifying which voters remain open to movement, how open they are, and what kind of message environment is most likely to shift them.

That is why I keep coming back to the Persuadability Quotient (here it is without my attempt at getting AI to draw the text backwards—in Leonardo style).

The Persuadability Quotient is not about who says they support you today. It is about who can still move tomorrow.

That distinction is bigger than it sounds. A voter can look “decided” in a traditional survey and still be highly movable if the underlying conditions are right. Another voter can look “soft” and yet be functionally locked in. Campaigns that fail to distinguish between those two types of voters waste enormous amounts of money talking to people who are never going to move while ignoring the voters who actually can.

Recent academic work on forecasting points in the same direction. A 2025 article in PS: Political Science & Politics argues that many existing forecasting approaches either rely on aggregation procedures that lack strong theoretical foundations or focus too heavily on macro variables while overlooking voter-level decision dynamics. The paper explicitly argues that voter-based regression models can improve predictions and notes that this kind of voter-level modeling is still used only rarely in forecasting exercises.

That is exactly the gap.

Campaigns often have a polling memo. They almost always have a media plan. They may even have some targeting. What they usually do not have is a serious probability model for voter movement.

They are often measuring support without modeling elasticity. That is a dangerous mistake in a primary.

In a general election, a lot of strategy is about coalition management at scale. In a primary, especially a competitive one, the game is often decided at the margins: which segment is still movable, which issue just became more salient, which message is actually resonating, and whether the campaign recognizes that quickly enough to reallocate money and attention.

That is where predictive analytics becomes decisive.

Not because models are magic. They clearly are not. Andrew Gelman and his coauthors make that point clearly in a 2024 review of election forecasting: uncertainty, nonsampling error, and model assumptions all matter, and they matter a lot.

But that is not an argument against modeling. It is an argument for better modeling.

A good predictive model does not eliminate uncertainty; it helps you reason inside it.

That matters because the alternative is usually much worse: gut feel, stale polling, and a consultant class that is often overconfident about what voters find persuasive. That is not strategy—it’s theater with a memo attached.

The right way to think about 2026 is this: Polling is a snapshot. Predictive analytics is a map of possible movement.

The snapshot still matters. But in a volatile primary, the map matters more.

We have seen versions of this repeatedly in the last three cycles. Stable-looking races break late. Narrative shifts arrive faster than traditional tracking can absorb. Campaigns that recognize the shift and reallocate quickly outperform expectations. The public lesson is usually framed as “momentum.” The underlying reality is almost always about asymmetric movement inside the electorate.

That is why campaigns need to think less about average voters and more about probability distributions.

Not every voter has the same chance of moving.

Not every issue has the same chance of becoming decisive.

Not every message has the same chance of landing.

And not every dollar has the same probability of changing the outcome.

The best 2026 campaigns are the ones that build strategy around those probabilities.

A 2024 article in the Journal of the European Economic Association models campaigns as dynamic contests and shows how resource allocation should change over time depending on how popularity leads decay and on the timing of voting. It is a more formal way of stating something practitioners should already understand: when conditions are dynamic, static spending plans are strategically inferior.

That should be especially relevant this cycle.

If you are in a 2026 primary, you are almost certainly operating in an environment where:

• attention is fragmented,

• issue salience can shift fast,

• turnout composition is uncertain,

• and late movement can matter disproportionately.

In that world, the old question—“What is our number?”—is not enough.

The better questions are:

• Which voters are still persuadable?

• How persuadable are they?

• What issue environment is most likely to move them?

• Which messages have the highest modeled probability of changing their behavior?

• And where should we spend next to exploit that information before the race moves again?

That is what the Persuadability Quotient is trying to get at. It is not a vanity metric. It is an attempt to measure the strategic elasticity of the electorate.

And it becomes much more powerful when paired with predictive analytics.

Because the point is not merely to label voters (even individually). The point is to estimate movement before it happens.

That is the hidden edge in a primary.

Not just knowing who is with you. Knowing who can still move.

Not just measuring the race. Measuring the conditions that decide it.

A lot of consultants find this difficult to accept because it forces a humbling conclusion: experience alone is not enough. Gut doesn’t cut it.

The academic literature is increasingly clear that persuasion is context-dependent, message effects are hard to predict intuitively, and voter-level probability models remain underused relative to their strategic value.

That leaves an opening. And in 2026 primaries, openings matter.

The campaigns that win will not necessarily be the ones with the best topline poll at a random point in time.

They will be the ones that understand, earlier than their opponents: which voters can move, what will move them, and how likely that movement is to occur.

That is not just polling. That is campaign science.

Most people assume there was a polling error.

It was not; It was because polling measures where voters are, while the outcome is determined by what’s about to move them.

The Blind Spot in Modern Campaigns

Campaigns today have more data than ever.

But most of it answers the same question:

“Who do you support right now?”

That’s useful, but not decisive.

Because elections aren’t decided by static opinion.

They’re decided by movement.

And movement is driven by a different set of variables entirely.

Sentiment Isn’t One Number; It’s a System

What actually determines whether a race breaks isn’t topline support, and I get that this is hard for most consultants to understand.

The truth is, what causes the break is what’s happening underneath those toplines.

Over the past several cycles, we’ve seen the same pattern emerge across multiple races:

• Incumbents with stable polling leads lose suddenly (Cantor 2014, Crowley 2018, multiple Texas GOP state reps 2024)

• Late-breaking shifts that traditional tracking missed (DeSantis 2018, Youngkin 2021, Cruz 2024)

• Campaigns that reallocated quickly outperform expectations (Abbott 2014, Kemp 2018, Pillen 2022, Stitt 2022)

In each case, the difference wasn’t better messaging.

It was better measurement of the conditions driving change.

The Metrics That Matter

Here are a few of the variables that consistently separate winners from losers:

1. Persuadability Quotient

This isn’t about who supports you. It’s about who can still be moved.

In multiple recent primaries, campaigns assumed their universe was fixed.

It wasn’t.

There were large segments of voters who appeared “decided” in polling, but were still highly persuadable under the right conditions.

The campaigns that identified and engaged those voters early saw outsized gains late.

2. Issue Salience Tracker

Not what voters think.

What they care about right now.

In one high-profile race, the dominant narrative shifted rapidly around an issue that wasn’t leading in earlier polling.

Campaigns that aligned early with that shift gained traction quickly.

Those that stayed anchored to prior assumptions were left reacting.

3. Message Resonance / Narrative Alignment

Most campaign messaging doesn’t fail because it’s wrong.

It fails because it doesn’t stick or spread.

In several recent races, we observed moments where a single message broke through and began compounding across channels; earned, digital, and direct voter contact.

Campaigns that recognized and amplified that message early created momentum.

Others missed it entirely.

4. Sentiment Volatility

Not all support is equal.

Some is stable. Some is fragile.

We’ve seen races where a candidate held a lead in polling, but underlying sentiment was highly volatile.

That volatility translated into late movement that flipped the race.

To a campaign relying on toplines, it looked like a surprise.

It wasn’t.

What We’ve Seen in Recent Cycles

In multiple recent contests:

• Incumbents with structural advantages lost primaries unexpectedly

• A late-cycle surge tied to a major cultural moment reshaped the trajectory of a race

• A challenger cleared the field decisively, without a runoff, by consolidating persuadable voters early

In each case, the signals were there. They just weren’t visible in traditional polling.

Why This Matters Now

Most campaigns are still optimizing for:

• Name ID

• Favorability

• Ballot position

Those are outcomes.

The real drivers are:

• Persuadability

• Salience

• Resonance

• Volatility

Polling tells you where the race is.

These metrics tell you where it’s going.

From Measurement to Advantage

This is the shift underway in campaigns right now: Not just collecting data, but measuring the conditions that cause movement.

Because the campaigns that win aren’t just the ones with support.

They’re the ones that understand:

• Which voters can move

• What will move them

• When that movement is starting to happen

And act before anyone else does.

The Bottom Line

Most campaigns are measuring the race.

A few are measuring what actually decides it-those are the ones increasing their chance to win.

Each had recent polling. Each felt confident in their positioning. Each believed they understood where the race stood.

And each was wrong.

Not because their pollster was bad. Not because the sample was flawed.

But because the ground had already shifted—and they didn’t see it.

The Problem Isn’t Polling. It’s Timing.

Polling still has value. It tells you where things were.

But campaigns today aren’t static environments. They are dynamic systems, where:

• Issues rise and fall in days

• Narratives spike and collapse in hours

• Voter sentiment shifts continuously—not in 2-week intervals

Yet most campaigns are still operating like this:

Poll → message → wait → poll again

That model worked when the information cycle moved slower.

It doesn’t anymore.

The Pattern We Keep Seeing

This isn’t new. It’s just more obvious now.

• In 2012, Obama didn’t just collect data; his team operationalized it, constantly adjusting field, digital, and messaging based on updated models.

• In 2016, Trump proved that narrative velocity can outpace polling reality; by the time numbers moved, the race had already changed.

• In 2021, Youngkin capitalized on a late-breaking education surge that reshaped the electorate faster than traditional tools could capture.

Different campaigns. Different contexts.

Same underlying truth:

The campaigns that win are the ones that adjust faster than the environment changes.

What’s Happening Right Now in Primaries

Across multiple races this cycle, we’re seeing a consistent pattern:

• Large blocks of voters remain undecided or soft

• Top issues, like cost of living (“affordability”) and education, are driving movement

• Narrative shifts are happening before polling reflects them

In one recent primary environment, nearly 40% of voters were still movable.

Most campaigns respond to that with more spending. But spending without precision is just noise.

From Signal to Spend: A New Operating Model

The campaigns pulling ahead right now aren’t just measuring sentiment.

They’re reallocating resources in real time based on it.

Here’s what that looks like in practice:

1. Detect the Shift (EyesOver)

We identify:

• Which issues are gaining traction

• Where sentiment is moving

• Which audiences are actually persuadable

A surge in cost-of-living concerns among soft GOP primary voters isn’t just interesting; it’s actionable.

2. Build the Decision Layer

Not all voters are equal:

• Loyalists → reinforce

• Leaners → convert

• Persuadables → prioritize

And not all messages perform equally across channels.

This is where most campaigns fall short; they see the data, but don’t translate it into decisions.

3. Act Immediately (Sparkfire)

Seeing the shift isn’t enough. The advantage comes from acting on it before your opponent does.

In multiple campaigns this cycle, we’ve paired real-time sentiment tracking with immediate voter engagement; targeting persuadable voters with messages aligned to what they’re already responding to.

The result:

Trackable double-digit movement among undecided and soft voters, in days, not weeks.

4. Reallocate the Budget (Where This Really Matters)

This is where campaigns either win, or waste money. Instead of defaulting to static media plans, we shift based on live signal:

• Linear TV

  • Maintain for scale and credibility

  • Reduce broad persuasion spend

• CTV

  • Increase targeted issue messaging

  • Align creative to real-time salience

• Digital (Meta/YouTube)

  • Concentrate on persuadables

  • Rapid creative iteration based on feedback

• AI Text (Sparkfire)

  • Direct engagement with undecided voters

  • Reinforce messages and address objections in real time

The result isn’t just better messaging, it’s more efficient spending.

This Is the Shift No One Is Talking About

Most campaigns still allocate budgets based on:

• Last poll

• Prior cycles

• Consultant instinct/incentives (15% commissions)

But the environment they’re operating in is changing daily.

The gap between what voters are feeling and what campaigns think they’re feeling is widening.

And that gap is where elections are won or lost.

Why We’re Opening This Up

For GOP campaigns, we’re making the EyesOver core platform available for free to kick the tires. Use it for 30 days to see the impact.

Because the biggest advantage isn’t having better data.

It’s seeing what others don’t, while there’s still time to act on it.

If You’re in a Primary Right Now

Ask yourself one question:

Are you making decisions based on where the race was…

or where it’s going?

If you’re not tracking narrative shifts and voter sentiment in real time, you’re not just behind;

you’re blind to what’s happening.

The campaigns that adopt this early will have an advantage.

The ones that don’t will eventually wonder why their polling looked right; but the result didn’t.