The concept goes back to the economist Paul Samuelson, who introduced it in 1938 to describe consumer behaviour without assumptions about inner utility. Today it is the counterpart to stated preference, the preference expressed when asked.
Why this matters for your decision
The idea is simple: what people choose says more about their preference than what they say about their choice. Samuelson formulated this for purchases at a given income and given prices. For companies, it follows that the most robust answer to the question "Which variant do customers want?" is an observed choice.
Traditionally, such data comes from the market: sales, scanner data, tariff switches, cancellations. For existing offers it is highly informative. For decisions still to be made it is missing: a new tariff, a new price or a new product has no market data yet. Anyone who waits until launch only gets revealed preference data after the investment has been made.
Behavioural tests bring this observation forward, before the investment. In Painted Door Tests, Horizon measures how real people respond to realistic offers with a price, which they see via Google and Meta ads in their familiar online environment. Nothing is sold, but the choice between variants is observed rather than asked for.
Example
A bank is considering offering a current account under one of two models: a fixed monthly fee with no conditions, or free of charge above a minimum monthly deposit. In a survey, a narrow majority prefers the free model. In the behavioural test, each person sees only one offer page. The fixed-fee variant achieves the higher measured sign-up intent because the condition puts many people off. Here the revealed preference differs from the stated one, and what counts for the decision is which model is chosen at the moment of the offer.
How it differs
Stated preference is the expressed preference from surveys, conjoint analyses or choice experiments. It can examine many attributes at once and explain the why, but it remains a statement without consequence.
Purchase intent can be either: asked on a scale (stated) or measured in behaviour (revealed). The Say-Do Gap describes the distance between the two. An A/B test also produces revealed preference data, but in the live system and on an existing offer.
Limitations
Revealed preference shows what was chosen, not why. Whether an offer was chosen because of the price, the wording or the image can only be separated by a clean test design in which the variants differ in exactly one respect, or by a complementary survey.
Observed behaviour also depends on the situation. If the offer is unrealistic or the page unclear, behaviour reflects these shortcomings. And a behavioural test describes the people reached through the chosen channels. The measured choice applies to this situation and this target group, not automatically to all channels.
Evidence
Samuelson 1938: Founds the theory of revealed preference: preferences are inferred from observed purchases at given prices and income, not from statements of utility. A Note on the Pure Theory of Consumer's Behaviour, Economica (context via Britannica). Source
Schmidt & Bijmolt 2020: 77 studies, 115 effect sizes: hypothetical willingness to pay is on average 21% higher than real willingness to pay; more so for higher-value products, specialty products and within-subject designs, and higher for indirect methods than for direct questions. Accurately measuring willingness to pay for consumer goods: a meta-analysis of the hypothetical bias, Journal of the Academy of Marketing Science. Source
Frequently asked questions
What is the difference between revealed and stated preference?
Revealed preference is inferred from observed behaviour, stated preference from statements in surveys. The first shows the choice, the second explains motives and can cover many attributes.
How do you measure revealed preference for a product that does not exist yet?
With a behavioural test before the market: people see a realistic offer and decide whether to choose it. Nothing is sold.
Is revealed preference data always better?
For the question of which variant is chosen, it is usually more robust. For the why and for many attributes at once, surveys are often the better tool.
Which preference does your offer reveal in behaviour?
Bring your decision question, and we will outline a possible test design.
You will speak with Daniel Putsche
Founder & CEO, 30 minutes
Read more