Favorite Brand Is Not Refrigerator Brand

Ask someone their favourite juice brand and you learn who they want to be. Ask what is actually in their fridge and you learn who they are. They are rarely the same brand — and that gap is one of the oldest problems in consumer research. Stated preference is aspirational; observed behaviour is mundane.

So we ran a small synthetic study with one question: are AI-generated respondents sensitive to that gap? Not "do they pick the right brand" — that is the wrong test. The real test is whether the way you frame the question changes the answer, the way it does with humans.

This is a demo of the engine, not a validated study. We will be honest about that throughout.

The setup

Three parallel studies — Spain, the US, the UK. Fifty-two synthetic respondents each (156 total), four demographic segments per country. Each respondent answered the same five questions in one interview — the trick is that they are five different framings of the same thing:

• Favourite — "What is your favourite juice brand, and why?"
• Usual — "What do you usually buy for home?"
• Fridge — "If I opened your fridge on a Tuesday morning, what would I find?"
• Receipts — "On your supermarket receipts from the last three months, which shows up most?"
• No-price — "If price did not matter, what would you buy?"

Because every respondent answered all five, we can compare each person to themselves. That is what makes the effect traceable.

The finding

The same respondent gives a different kind of brand depending on the frame. We coded each answer as mainstream (a mass or private-label brand) or not. Here is the share of mainstream answers, by frame, in each country:

The pattern is identical in all three cultures. "Favourite" is the least mainstream of the realistic frames — it pulls toward identity (organic, fresh-squeezed, premium). "Receipts" is the most mainstream — it pulls toward the supermarket own-brand they actually buy. And "if price did not matter" collapses the mainstream share to almost nothing: identity, with the friction of cost removed.

The brand changes by country, but the move is the same. In Spain the receipts brand is Hacendado; in the UK it is the Tesco own-brand; in the US it is Great Value or Minute Maid. Remove the price and they all reach for something fresh-squeezed, organic, or Innocent.

Receipts: "Probably the Tesco own-brand, the one that is ninety-nine pence. I get that most weeks, without really thinking about it." — No-price: "Maybe I would try that Innocent Smoothies stuff more often? They do look good." (Same UK respondent.)

Why this matters for research

The shape of the question decides what you measure. "Favourite" measures aspiration. "Receipts" measures habit.

If you design a study that asks people their favourite and then report it as market behaviour, you have measured a daydream. The frame is not a wording detail — it selects between two different psychological registers. Knowing which one you are in is the difference between an insight and a mirage. That synthetic respondents reproduce this distinction — consistently, across three cultures — is the useful result, far more than any single brand number.

The honest part

These are synthetic respondents only — no humans were compared. Fifty-two per country is a demonstration of a pattern, not a market estimate. The brand shares above are not real-world shares and should never be read as such; the defensible claim is the direction of the frame effect, not the numbers themselves. The coding is heuristic, so the mainstream share is the robust signal and the "aspiration" score is directional. The next step is a human benchmark — the same questions to real shoppers, with blind coding — to test whether the direction holds.

Try it yourself

This came out of describing four segments in plain text and asking five framings of one question. You can run the same kind of frame test on your own category in minutes.

A full working paper — method, multi-category extension (juice, yoghurt, coffee) and the human benchmark — is in preparation. Powered by the StrataSynth engine.