Why Focus Groups Lie: The Gap Between What People Say and What They Do

In 1985, Coca-Cola ran the largest consumer research programme in the history of the beverage industry. Nearly 200,000 blind taste tests. The results were unambiguous: people preferred the new formula. Coca-Cola launched New Coke. Within seventy nine days, they reversed the decision after a consumer revolt that cost the company hundreds of millions of dollars.

The taste tests were not wrong. People genuinely preferred the sweeter formula in a sip test. But a sip test does not measure what drives a purchase decision. Brand loyalty, emotional attachment, identity signalling, habit: none of these show up when someone is asked to compare two unmarked cups. The gap between what people said they preferred and what they actually bought was not a research failure. It was a structural limitation of the method.

That gap has a name. Behavioural scientists call it the stated preference vs revealed preference problem. It is one of the most replicated findings in the social sciences, and it shows up everywhere. Voters tell pollsters one thing and vote differently. Consumers say they prioritise sustainability and then buy the cheapest option. Job candidates claim to value work life balance and then accept the role that pays 15% more.

Focus groups sit squarely in the stated preference column. They are designed to capture what people say. Not what they do.

The room changes the answer

Put eight strangers in a room with a moderator and a one way mirror. Within five minutes, a social hierarchy emerges. One participant is louder, more articulate, more confident. That person's opinion begins to anchor the group. The other seven either agree, modify their position to align, or stay quiet.

This is not a moderation failure. It is a well documented phenomenon called informational social influence: when people are uncertain about their own preferences, they look to others for guidance. In a focus group, the loudest voice in the room becomes the reference point for everyone else. A skilled moderator can mitigate this, but they cannot eliminate it. The room itself is the problem.

Research from Solomon Asch's conformity experiments through to modern replications consistently shows that people will change their stated opinion to match a group consensus, even when the group is obviously wrong. In a focus group setting, where the questions are genuinely ambiguous ("Which packaging design do you prefer?"), the pull toward conformity is even stronger.

Social desirability bias: the quiet distortion

Beyond groupthink, there is a subtler problem. People perform for an audience. They give answers that make them look thoughtful, responsible, and virtuous.

Ask a focus group participant whether they read product labels before buying. Most will say yes. Observational data from eye tracking studies in supermarkets tells a different story: the average shopper spends less than two seconds looking at a product before it goes in the trolley. The participant is not lying. They genuinely believe they read labels. They have constructed a narrative of themselves as a careful shopper, and that narrative is what you hear in the focus group room.

This extends to every category. "I choose quality over price." "I prefer brands that align with my values." "I would pay more for sustainable packaging." These are not descriptions of behaviour. They are aspirational self portraits. The participant is telling you who they want to be, not who they are at 7pm on a Wednesday when they are tired and the cheapest option is right there.

Post-hoc rationalisation: the narrative trap

Even when people genuinely try to explain their decisions, they cannot do it accurately. Daniel Kahneman and Amos Tversky spent decades demonstrating that human decision making operates on two tracks: a fast, intuitive system that makes the actual choice, and a slow, deliberative system that constructs a rational story about it afterwards.

When a focus group participant says "I chose Concept B because the typography felt more premium," they are not reporting a decision process. They are narrating one. The actual decision happened in milliseconds, driven by pattern recognition, emotional association, and personality traits that operate below conscious awareness. The explanation came after, and it sounds reasonable, but it may have nothing to do with the real drivers.

This is the Pepsi Challenge problem scaled to every category. People preferred Pepsi in the sip test. They bought Coca-Cola in the shop. When asked why, they said things like "I just prefer the taste" or "It is the real thing." Those are post-hoc narratives, not causal explanations. The actual purchase driver was a complex interaction of brand memory, visual identity, social context, and habitual behaviour that no participant could accurately introspect on.

How synthetic panels sidestep the room

A synthetic persona responding to a stimulus has no room to perform for. There is no group to conform to. There is no moderator whose approval matters. There is no one way mirror creating performance anxiety.

Each persona in a Panel Studio panel responds independently. A panel of 400 personas produces 400 independent responses, each driven by a structured personality profile built on the DYNAMICS-8 framework. The persona does not know what the other 399 said. There is no anchoring effect, no conformity pressure, no social desirability filter.

More importantly, the persona cannot post-hoc rationalise. Its response is generated from its personality dimensions, its demographic context, and its simulated life experience. A high Impulsivity persona does not claim to be a careful comparison shopper. It reaches for the product that triggers an immediate emotional response, because that is what its personality profile predicts. A high Discipline persona does not pretend to be spontaneous. It evaluates the options methodically, because that is what Discipline does.

The reasoning trace that accompanies every response shows exactly which personality dimensions were active, how they interacted, and what contextual factors modulated the decision. You do not get a narrative. You get a mechanism.

The honest caveat

Synthetic panels model idealised personality driven behaviour. They show you what a person with a given personality profile, demographic background, and life context would most likely do, given that their response is driven entirely by those factors.

Real people are messier. They are tired. They are distracted. They bought the wrong thing because it was on the end of the aisle. They picked a brand because their friend mentioned it yesterday. These situational factors are genuine drivers of real world behaviour, and a personality simulation does not capture all of them.

The value of synthetic panels is not perfect prediction. It is the systematic removal of known biases. Focus groups introduce groupthink, social desirability bias, and post-hoc rationalisation. These are not minor distortions. They are structural features of the method. A synthetic panel eliminates all three by design.

If your research question is "What do people say they prefer?", a focus group is the right tool. If your research question is "What will people actually do, and why?", you need a method that does not let them perform for the room.

The gap between stated preference and revealed preference is not going away. It is human nature. The question is whether your research method accounts for it or pretends it does not exist.