Synthetic personas speed up research, but human judgment remains the compass 

AI is reshaping the way marketers run research. MarketingTribune sat down with Paul McCormick, PhD, Head of Digital Business at Empathy Lab by EPAM, to explore how synthetic personas can change the game. “It’s about scale, speed, and validating sooner,” he says. 

With the launch of Synthetic Audiences, the Netherlands now has access to a new layer of speed, scalability, and data-driven experimentation. To show what the system can do, McCormick demonstrates how quickly synthetic personas can be generated. Feed in a few traits – age, place of residence, occupation – and within seconds you get a full profile, from attitudes and buying behaviors to specific preferences. “The more precise your input, the sharper the persona,” he explains. “But even a very minimal description delivers surprisingly robust profiles.” 

McCormick then puts the freshly created personas to work, testing three fictional candy ideas: two realistic fruity chewy concepts and a deliberately bad one: a salty anchovy chew bar. The system evaluates each idea, scoring them on innovation, credibility, purchase intent, and more, and then explains the reasoning as if real consumers were weighing in. “You don’t only get the numbers,” McCormick says. “You also get the rationale behind them. And you can keep the conversation going, ask the personas follow-up questions, like how you might improve a concept.” 

The demo shows that synthetic personas aren’t deterministic, but they are remarkably stable. When he generates multiple personas using the exact same input, the results only differ slightly. “That mirrors what you see with people when you ask them the same question on a scale over and over again,” he notes. “The variation is small, but never zero, and that’s how it should be.” 

“We saw a 70 to 80 percent overlap in the big trends. Not identical, but all of the major patterns showed up again.” 

Paul McCormick, PhD, Head of Digital Business at Empathy Lab by EPAM 

Patterns

So how reliable is synthetic research?  

McCormick shares how Empathy Lab validated outcomes against traditional studies. In two client engagements, healthcare professionals and staff in commercial kitchens, the team first spent weeks on classic research with interviews and deep dives. Then they ran the same research question synthetically. “We saw a 70 to 80 percent overlap in the main trends,” he says. “Not identical, but the important patterns all came back.” 

The approach also guided product innovation at Mars Wrigley, where Empathy Lab compared synthetic personas with an existing human test panel. “Even without additional data we hit 75 percent agreement; when we layered in historical concept data we climbed above 80 percent,” McCormick explains. One product stood out: synthetic respondents rated it higher than the human panel, while Mars’ internal experts strongly believed that flavor was on the verge of becoming a trend. “So the synthetic outcome diverged from the panel, but it lined up with market expertise. That raises a question: what’s really the right measure of ‘accuracy’?” 

Consistency 

That observation taps into a bigger truth: human research isn’t flawless either. McCormick references a dataset where respondents first claimed they would “never” buy a particular product, only to eagerly confirm they would “definitely” purchase the new version. “People can be inconsistent, distracted, or tell you what they think you want to hear,” he says. “That’s why it’s impossible to hold 100 percent accuracy up as the benchmark, either for AI or for humans.” 

Does that mean synthetic research is only for large enterprises? McCormick doesn’t think so. “Any organization can start experimenting with public models,” he says. “But if you want a scalable, repeatable research solution, you need to invest in configuration, data frameworks, and governance. That’s the work companies have to do if they want to make this a structural capability.” 

“People are inconsistent, distracted, or socially desirable.” 

Paul McCormick, PhD, Head of Digital Business at Empathy Lab by EPAM 

Safe by design

Governance is crucial across the board. Synthetic research can deliver such speed and volume that there’s a real risk of overdoing it. Companies still have to decide when synthetic insights can lead, when they need to be blended with human research, and when only real consumers should have the final say. “We’re absolutely not saying synthetic research replaces human research,” McCormick stresses. “For physical products like food, you still need a human taste test. What you can do is evaluate many more concepts up front so you’re far better prepared.” 

Synthetic research also offers privacy advantages. Because the personas are fully artificial, there’s no PII in the data, which makes it safer to share across borders. McCormick does caution companies that enrich personas with their own data. “Once you start feeding historical research data into the system, you need to pre-process it to make sure no personally identifiable information remains. But after that, because the personas themselves contain no PII, they’re actually easier to share across multiple markets.” 

The human compass

Asked how AI-driven insight generation will evolve, McCormick points to a familiar pattern across tech waves. “We’ll go through phases where companies lean too heavily on AI, and then rediscover the value of human input. Eventually you get a balance.” What’s inevitable, he says, is that the scale and speed will keep increasing. At the same time, a solid feedback loop is essential. “You run the synthetic tests, you launch a campaign or product, and then you have to feed the real-world results back into the system. That’s how you keep it sharp. Otherwise you risk the AI recycling its own output and becoming an echo chamber.” 

For teams that want to move faster without losing their instincts or human focus, synthetic research is becoming a critical new layer. “It helps teams learn faster, test more often, and make better-informed decisions,” McCormick concludes. “But people will always be the anchor.” 

Visit MarketingTribune to read the original article.