What stated and revealed preference are
A stated preference is any answer to a question: a survey rating, an interview remark that a feature would be useful, a figure someone names when asked what they would pay, a ranking of options on a screen. A revealed preference is inferred from a choice: which tier new customers pick, which plan people leave, whether a prospect pays a deposit, whether a visitor who was offered two delivery options chose the slower, cheaper one.
The distinction matters because marketing teams constantly have to predict behaviour that has not happened yet, and asking is the cheapest way to get an answer. The cheapness is the trap. A question about the future costs the respondent nothing to answer generously, so a room full of people who say they would buy can turn into a launch nobody pays for. The guide to customer interviews builds its whole method around this: it asks about the last real episode rather than about intentions, and keeps hypothetical questions to the end, if they are asked at all.
Neither kind of evidence is simply better. They answer different questions, and the useful skill is knowing which question you are asking. The comparison of qualitative and quantitative research sorts methods by the kind of data they produce; the stated and revealed split cuts across it, because a survey can ask about past purchases and an interview can watch someone use a product.
Why they differ
The gap between what people say and what they do has a handful of ordinary causes. None of them requires anyone to lie.
| Source of the gap | How it shows in marketing | How to reduce it |
|---|---|---|
| Nothing is at stake when answering | Many say they would pay for a feature; few buy it when it appears | Ask for a commitment that costs something, or observe a past purchase |
| People are poor at predicting their own future | Intentions to switch supplier "next quarter" that never happen | Ask about the last time they actually switched, and what triggered it |
| Wanting to be helpful or agreeable | Warm reactions to a concept the interviewer plainly likes | Hide which option is yours; show the product late or not at all |
| Answering an easier question | "Would you like faster delivery?" answered without weighing the price | Present the trade-off as the buyer would meet it, with the price attached |
| Self-image | Respondents claim to choose on quality and buy on price | Compare what they say with what they bought, record by record |
| Memory | Confident answers about how they first heard of a brand | Ask close to the event and compare with the tracked source |
| The setting of the choice | Behaviour shaped by the prices, defaults and availability the business set | Vary one of those deliberately before reading behaviour as a preference |
The last row is the one that catches experienced teams. Observed behaviour is a response to the choices that were offered, so a business that only ever showed one price, one default plan and one delivery option will see buyers "prefer" exactly those. The guide to self-reported attribution shows the opposite problem on the stated side: it asks buyers where they heard of the business, then reads the answers against the tracked source on the same record because memory and the wish to give a tidy answer both shape what people say.
Combining the two in research and decisions
Start from the decision, not the method. The guide to research design and the brief turns a decision into research questions and warns that the obvious question is usually a hypothetical one. Once the questions are written, label each by the evidence it needs. "Why did customers choose us over the alternative?" needs stated answers about real past choices. "Will customers pay more for the premium tier?" needs revealed behaviour, or the closest thing the business can arrange.
Use stated answers to generate, and behaviour to decide. Interviews and open survey questions are good at finding reasons, objections, the words customers use and options nobody had thought of. They feed a HypothesisA testable statement of the change to make, the effect expected on a named metric and the evidence-based reason for expecting it. about what people will do. The hypothesis is then checked against what people do when a real choice is put in front of them. A Usability TestA research session in which a person attempts a realistic task on a site, app or prototype while an observer notes where they struggle. sits between the two: the task is set up by the researcher, but what the participant does with it is observed rather than reported.
Ask for commitment, not interest. The guide to testing demand before launch counts commitments such as deposits and pre-orders rather than clicks, because a click costs almost nothing. The same reasoning ranks evidence inside revealed preference: a paid order outranks a free trial, which outranks a waitlist place, which outranks a click.
Read price evidence from both sides. The pricing and packaging guide asks for stated and observed evidence together, and treats the gap between what buyers say they would pay and what they do pay as informative in itself. Stated price answers help set the range to test and explain why a price feels wrong; win and loss records, discounting and tier choice show what actually happens at the price.
Record a BaselineAn estimate of what a metric would have done without a given change, campaign or promotion, against which the effect is measured. before a change. Behaviour only reveals a preference when it can be compared with what happened before, or with a group that did not get the change. Without that comparison, a rise in bookings may be the season, not the offer.
Keep segments honest. A buyer personaA research-based profile of a typical customer type: goals, problems and buying behaviour. built only from what people said in interviews describes how buyers see themselves. Checking each persona against what that group actually bought, when and at what price keeps it tied to behaviour.
Worked example (illustrative)
A coworking operator prices a quiet zone
Jorpendace Coworking, an invented operator with 3 sites and about 420 members, surveys its members about a paid quiet zone. Of 186 who reply, 112 say they would pay for one, and the most common answer to the price question is £40 a month.
Rather than build the zone everywhere, the operator turns one meeting room at its largest site into a bookable quiet room for a trial month, at £8 a day or £35 a month. A baseline month of meeting room use at that site is recorded first.
During the trial, 23 members book at least one day and 6 take the monthly option. Most day bookings fall on Tuesdays and Wednesdays, and 4 of the monthly buyers had not answered the survey at all. Short interviews with bookers find the reason is calls, not concentration: they want somewhere to take a call without being overheard.
The operator prices the zone from bookings, not from the survey: day passes stay at £8, the monthly option is dropped, and the next site adds two phone booths instead of a large quiet room.
Limits and criticisms
Behaviour shows what people chose from what was on offer, not what they would choose from options that do not exist yet. A business that relies only on revealed preference can refine what it already sells and never discover the product customers would have preferred, because nobody could choose it.
Behavioural data rarely says why. A drop in upgrades could mean the price is too high, the benefit unclear, the upgrade path hidden or the season quiet. The data shows the choice; explaining it needs people to say something.
Stated answers are the only source for some questions. Reasons, feelings, the alternatives considered and the words customers use do not appear in a transaction log, and dismissing everything people say as unreliable throws away the evidence that explains the behaviour.
Observed behaviour is shaped by the prices, availability and defaults the business set. Reading it as pure preference can simply confirm the business's own earlier choices: the default plan looks popular because it was the default.
And the distinction gives no rule for how much of either kind of evidence a decision needs. That remains a judgement about the size and reversibility of the decision.
Related frameworks
Jobs-to-be-Done is largely a method for getting stated evidence right: it reconstructs a real past switch instead of asking what someone would want, which turns a stated answer about the past into something close to revealed behaviour. The value proposition canvas lays out customer jobs, pains and gains beside the offer; filling its customer side from behaviour as well as interviews keeps the fit it shows from resting on what people merely said. The marketing frameworks library lists the other measurement frameworks by the decision they serve.
Common mistakes
- Reading intentions as forecasts. "Would buy" is a reaction to a description, not a purchase.
- Asking the price question in the abstract. Without the trade-off attached, everyone prefers more for the same.
- Counting interest as demand. Clicks and sign-ups cost little; deposits and orders show more.
- Treating sales data as pure preference. Buyers can only choose what the business offered, at the price it set.
- Dismissing what people say. Behaviour shows the choice; only people can explain it.
- No comparison point. Without a baseline or a group that did not get the change, behaviour proves little.
- Personas built from self-description alone. Check each one against what that group actually bought.
