Skip to content
HIPOPOTESISStart a conversation

Willingness to pay research: what people say, and what they’d pay

Willingness to pay research, explained. Why people overstate what they’d pay, which methods come closest to real purchases and how to price a big decision.

Abdo LópezCo-founder & CEO
Published
Reading time
9 min read
Format
Article
Say or pay21%higher
How far stated willingness to pay ran above answers with real money at stake, on average, across 77 studies of consumer products and services.
TopicGrowth & Market

In short

Three points | 30 seconds
  1. Asking people what they’d pay tends to give you a number that’s too high. In a 2020 meta-analysis of 77 studies of consumer products and services, what people said they’d pay was on average about 21% above what they were willing to pay when their answer had real financial consequences.
  2. In a Swiss experiment with 1,124 consumers and an inexpensive cleaning product, the methods where people might really have to buy came in about 6% to 11% above the figure estimated from real purchases, while a conjoint survey with nothing at stake came out 76% too high.
  3. People judge a price against where they buy and what feels fair, so good willingness to pay research finds out what people compare you with before it asks them for a number.

Willingness to pay research tries to answer one of the most expensive questions in pricing research: what will people actually pay for this? Asking them gets you part of the way, but the answer tends to come back too high, and how far too high depends on how you ask.

Picture the meeting. The product is nearly ready, the margin model needs a number and someone has a survey saying customers would pay €49 on average. It’s tempting to type that figure straight into the plan. The research below suggests reading it as probably on the high side, and asking a few more questions before anyone builds a forecast on it.

What is willingness to pay, and why is it hard to measure?

Willingness to pay (WTP) is the most a person would pay for a given quantity of a product or service, as Jonas Schmidt and Tammo Bijmolt define it in their meta-analysis.1 The definition is simple, but many people don’t carry that number around, and in our experience they work it out when they need it, from whatever is in front of them.

Klaus Miller and his colleagues list a few reasons the answers drift. It’s hard to know what something is worth to you, especially when the product is new. Some people answer strategically, hoping a low answer will lead to a low price, and others overstate it because of the self-image they’d like to project.2

How much do people overstate what they’d pay?

Researchers call the gap between what people say and what they’d really pay hypothetical bias. A hypothetical answer has no financial consequences. A real one does: the participant may have to pay the price they named.1 In the meta-analysis below, most real measures come from an incentive-compatible mechanism, such as an auction or a BDM lottery, where a stated price can become binding.1

Schmidt and Bijmolt pooled 115 comparisons from 77 studies of consumer products and services, published in 47 papers, with 24,347 hypothetical and 20,656 real observations. On average, hypothetical willingness to pay was about 21% higher than real willingness to pay.1 The gap was bigger for specialty goods than for convenience goods, when the same people answered both ways and, with weaker support, for higher-value products.

The finding the authors themselves call surprising was about method. Many researchers assume that indirect methods, such as a conjoint survey where people choose between versions of a product at different prices, are less biased than simply asking for a price. In the meta-analysis it went the other way: indirect methods overestimated real willingness to pay more than direct questions did, by about 10 percentage points in the full model, which the authors note fits companies’ habit of relying on direct survey questions.1

They suggest using the predicted bias for each kind of product and method to adjust decisions. For specialty goods, for example, they estimate that direct questions overstate by 28% and indirect methods by 40%.1 That’s a long way from taking 21% off your own survey and calling the result a price. Every study in the sample was about consumers, too, so it can’t tell you how big the gap is for a business-to-business price, and the hint that it grows with product value is weak. For a high-value decision, it’s safer not to assume the gap shrinks.

Which willingness to pay methods get closest to real purchases?

Miller and his colleagues sort the common methods with two questions.2 Does the person name a price, or choose between options from which a price is worked out? And does their answer cost them anything?

  • Ask for the price. “What’s the most you’d pay for this?” Direct, and hypothetical.
  • Choose between versions. A choice-based conjoint survey shows several versions of a product at different prices, plus the option to buy none, and works out willingness to pay from the choices. Indirect, and hypothetical.
  • Name a price, and pay it. In the Becker-DeGroot-Marschak (BDM) method, named after the three researchers who described it in 1964,3 you state your maximum and a price is then drawn at random. If the drawn price is at or below your number, you have to buy at that price. If it’s higher, you don’t buy. In theory, stating anything other than your true maximum can never help you.2
  • Choose, and pay. An incentive-aligned conjoint survey works like the conjoint above, but the choices have real consequences.

The same team tested all four against real purchases. They randomly assigned 1,124 Swiss consumers to five groups. Four answered with one method each about an inexpensive cleaning product for computer keyboards and other high-tech equipment. The fifth saw the product in a test online shop, at a randomly assigned price, and decided whether to buy it.2

Figure | Say or pay Which way of asking came closest to what people really paid? Guess first, then pick each method to compare it with the figure estimated from real purchases of the same cleaning product. Switch views to see the price and the profit each method would have forecast.

Results from an experiment by Klaus Miller, Reto Hofstetter, Harley Krohmer and John Zhang with 1,124 Swiss consumers and an inexpensive cleaning product for high-tech equipment. Average willingness to pay estimated from real purchases in a test online shop: CHF 8.46. Naming a price with a real obligation to buy (the BDM method): CHF 8.96, about 6% above. Conjoint choices with real consequences: CHF 9.39, about 11% above. Asking for the price with nothing at stake: CHF 11.03, about 30% above. Conjoint choices with nothing at stake: CHF 14.92, about 76% above. The profit-maximising price from real purchases was CHF 8.50, with a 95% confidence range from CHF 6.87 to CHF 12.30, so wide that all four methods' recommended prices fell inside it: CHF 8.16 (BDM), CHF 7.93 (conjoint with consequences), CHF 9.68 (asking for the price) and CHF 11.49 (conjoint with nothing at stake). The profit forecasts differ more: compared with the forecast from real purchases, BDM was about 7% too high, conjoint with consequences 30%, asking for the price 43% and conjoint with nothing at stake 112%. Only the two forecasts with nothing at stake differed significantly from the real one, though every method's error was large.

Which of the four do you think landed closest to real purchases? Pick one.

Data: Klaus M. Miller, Reto Hofstetter, Harley Krohmer and Z. John Zhang, GfK Marketing Intelligence Review, 2012, tables 3 and 4 (an updated version of their 2011 study in the Journal of Marketing Research). One inexpensive product, Swiss consumers, a test online shop. The real-purchase figures are estimated from each shopper’s decision to buy or not at a random price. Prices in Swiss francs; percentages rounded.

The two methods with something at stake landed closest: BDM about 6% above the figure estimated from the shop’s purchases, and incentive-aligned conjoint about 11% above. The open question came in 30% higher and the conjoint with nothing at stake 76% higher. One reason the two conjoints differed so much, the authors suggest, may be the “none” option: 19% of participants chose it in the incentive-aligned conjoint, against 5% in the hypothetical one.2

Every method pointed to a profit-maximising price inside the 95% confidence range from the real purchases, but that range was wide, from about CHF 6.87 to CHF 12.30. The profit forecasts showed more of a difference. Against the forecast from real purchases, BDM came out about 7% too high, incentive-aligned conjoint 30%, the open question 43% and the conjoint with nothing at stake 112%. Only the two hypothetical forecasts differed significantly from the real one, though the authors note that every method’s error was large, and they argue that a biased method can still point to a good price.2

This is one inexpensive, fairly unique product, in one country. Citing earlier studies, the authors suggest conjoint may do better for products that face more competitors and involve a longer decision, such as a digital camera.2 They also point out that real stakes can be hard to use when the price is very high, the product is highly individualised, there’s no prototype yet, or cost or the law gets in the way, which covers many of the prices a board worries about most.

What is the Van Westendorp price sensitivity meter?

The Van Westendorp price sensitivity meter is a short survey that asks people at what price a product would be too expensive to consider, so cheap they’d doubt its quality, a bargain, and getting expensive. It has been used in commercial pricing research to find critical price ranges for new or relaunched products.4 A range is useful early on, when nobody knows where to start. Its four questions are about price thresholds, though, so they don’t measure how many people would buy at a given price, and they share the weakness of every hypothetical question: nobody pays anything for their answer.

Why is the same product worth more in one place than another?

This is the part of willingness to pay that a spreadsheet tends to miss, and where an anthropologist starts asking questions.

In a study published in 1985, Richard Thaler asked regular beer drinkers on an executive development programme to imagine lying on a hot beach, longing for a cold beer. A friend offers to fetch one and asks for a price: if the beer costs that or less, he’ll buy it, and if it costs more, he won’t. Their median answer was $2.65 if the beer came from a fancy resort hotel and $1.50 if it came from a small run-down grocery store, though it would be drunk on the beach either way.5 Thaler explained the difference with what he called transaction utility: people judge the deal as well as the thing, against what seems a normal price in that setting.

A sense of what’s fair moves the number too. In telephone surveys of randomly selected residents of Toronto and Vancouver between 1984 and 1985, Daniel Kahneman, Jack Knetsch and Thaler asked about a hardware store that raises the price of snow shovels from $15 to $20 the morning after a large snowstorm. Of the 107 people asked, 82% said it was unfair.6 The authors explained it as a dual entitlement: customers feel entitled to the terms they’re used to, and firms to their usual profit, but not to exploit a jump in demand.

Both are hypothetical questions asked decades ago, one of them in Canada, so they show a pattern, and the idea goes back much further. In 1971 the historian E. P. Thompson wrote about the moral economy of the English crowd: people in eighteenth-century England who rioted over food prices, acting on a shared sense of what was legitimate in the food trade, including a just price.7 People judge a price by the kind of exchange they think they’re in, and in our experience a price that breaks that sense can put off people who could easily afford it.

How to measure willingness to pay for a big decision

If the price you’re setting carries real money, here’s the order we’d work in.

  1. Start from the decision. A launch price, a price rise and a new tier need different evidence. As in any go-to-market research, ask how expensive it would be to get it wrong, and how quickly you could change it.
  2. Find the reference first. Interviews and observation won’t give you a reliable number, but they show what people compare you with, what words they use for “too expensive” and what would feel unfair. In our Amazon study, going into shops with each shopper showed us when they checked prices on their phones.
  3. Put something at stake where you can. A BDM question, an incentive-aligned conjoint or real pre-orders. A price test on part of your online traffic can work too, if you have the traffic, though it shows short-term sales, and given what the fairness studies above suggest, it’s worth asking how visitors would feel about seeing different prices. Where nothing can be at stake, assume stated answers lean high, and use a conjoint for what Miller and his colleagues suggest it for: comparing how much each feature and the price matter.2
  4. Look at the spread and the segments. Miller and his colleagues warn that when different groups value a product differently, the average can mislead.2 That’s one more reason customer segmentations need to be built around real differences.
  5. Check it against what people do and say unprompted. Sales data, reviews and public conversations show the reference in people’s own words. In our experience, social listening leans towards complaints and the people who post most, so it shows how people talk about your price, and it can’t tell you how many would pay it.

This is what we do

For a question like this, our pricing and value perception work brings together all three lenses of our H3™ method: interviews to find the reference and the lines people won’t cross, a survey with something at stake wherever possible to find the range people accept and the spread of what they’d pay, and social listening to hear how they talk about price when nobody is asking. When people call a price unfair in interviews and still buy at it in the survey, that disagreement is often the finding. If a price test on your own traffic would answer your question faster and cheaper, you’ll hear that from us first.

Ask people what they’d pay and you’ll get an opinion. Give the answer a cost and a context, and you’ll get much closer to a price.

Sources

Where the figures and studies in this article come from.

  1. Accurately measuring willingness to pay for consumer goods: a meta-analysis of the hypothetical biasJonas Schmidt and Tammo H. A. Bijmolt, Journal of the Academy of Marketing Science, 2020
  2. Measuring Consumers’ Willingness to Pay: Which Method Fits Best?Klaus M. Miller, Reto Hofstetter, Harley Krohmer and Z. John Zhang, GfK Marketing Intelligence Review, 2012
  3. Measuring utility by a single-response sequential methodGordon M. Becker, Morris H. DeGroot and Jacob Marschak, Behavioral Science, 1964
  4. A Review of Methods for Measuring Willingness-to-PayChristoph Breidert, Michael Hahsler and Thomas Reutterer, Innovative Marketing, 2006
  5. Mental Accounting and Consumer ChoiceRichard H. Thaler, Marketing Science, 1985
  6. Fairness as a Constraint on Profit Seeking: Entitlements in the MarketDaniel Kahneman, Jack L. Knetsch and Richard H. Thaler, American Economic Review, 1986
  7. The Moral Economy of the English Crowd in the Eighteenth CenturyE. P. Thompson, Past & Present, 1971

Field Notes

Get the next one in Field Notes

One email a month from Abdo and Norah. The month’s articles, a scene from the field and one thing worth reading.

Hi, I’m from . Please send Field Notes to

Written by Abdo and Norah. Knowing your company helps us send what fits your sector. Unsubscribe in one click. Privacy policy.