Last updated: 1 September 2026
Do not ask people what they would pay. Ask them to choose between options that each carry a price, and read the price out of what they gave up. The direct question produces a number that is reliably too high, and the size of the gap is large enough to invalidate a business case built on it.
The clearest published measure comes from a systematic review of criterion validity in willingness-to-pay methods, covering 50 papers and 159 comparisons of hypothetical against actual payment. That review found hypothetical willingness to pay averaging 3.2 times actual willingness to pay, across a range running from 0.7 to 11.8. That is not a bias you can correct with a rule of thumb, because the spread is wider than the average is useful.
An earlier meta-analysis of 28 stated-preference studies, covering 83 observations, put the median ratio of hypothetical to actual value at just 1.35. That the two reviews land so far apart is itself the useful finding: the size of the inflation is unsettled, which is why no single correction factor works.
What makes pricing research tractable despite that is a subtler finding, and it is the one that should drive method selection. The bias lands hardest on absolute price levels and considerably more lightly on the relative trade-offs people make between features. Design around that asymmetry and you get usable answers; ignore it and you get a confident number that will not survive contact with a checkout page.
Why Direct Price Questions Overstate What People Will Pay
Three mechanisms, and they compound.
- No budget constraint is engaged. Saying yes to a hypothetical price costs nothing. Real purchases displace other purchases, and that displacement is what actually caps willingness to pay.
- Respondents answer the question they think you are asking. A price question in a research context reads as a question about whether the product is good. People who like a product say yes to prices they would not pay.
- Social and self-presentation effects. Saying a product is worth very little feels dismissive; saying you would pay a lot signals discernment. Both push the same direction.
The evidence is not uniformly damning, and the variation is informative. A synthesis of hypothetical bias in stated choice experiments, published as an integrative review of the empirical evidence, reports significant bias as ubiquitous in consumer-behavior and transport studies while health-related choice experiments frequently find negligible bias. Context, incentive structure and how consequential the choice feels all move the number. The companion paper on bias mitigation methods covers what has actually been shown to reduce it.
Which Pricing Methods Survive the Bias?
| Method | What it asks | Robustness to hypothetical bias | Best used for |
|---|---|---|---|
| Direct WTP question | "What would you pay for this?" | Poor. Inflates, unpredictably | Nothing load-bearing. Occasionally useful as a rough upper bound |
| Van Westendorp price sensitivity meter | Four price points: too cheap, cheap, expensive, too expensive | Moderate. Anchors on the respondent's own scale rather than yours | Early-stage range finding on a new category |
| Gabor-Granger | Purchase intent at a series of set prices | Moderate. Still hypothetical, but constrained | Demand curve shape for one product |
| Conjoint and discrete choice | Choose between full product profiles that trade features against price | Best available. Bias hits levels more than trade-offs | Feature and price architecture, portfolio and bundle decisions |
| In-market price test | Real prices, real transactions | Highest. It is behavior, not stated preference | Final validation where the operational cost is acceptable |
Best when you can run it: the in-market test, and it is worth saying that plainly because it is the row that needs no research vendor at all. If you can vary price in a live channel and measure conversion, the survey question is a proxy for something you could simply observe. Pricing research earns its place when a live test is impossible, too slow, or too expensive to run across the range you need.
Van Westendorp deserves a specific caution. It is cheap and it produces a chart. That is why it is popular.
It answers a narrower question than the chart implies: it identifies a range of prices the market finds credible, not a revenue-maximizing price. Treating its intersection points as a recommended price is a routine misreading.
Why Do Trade-Off Methods Hold Up Better?
Because of what the bias does and does not touch.
The stated-choice literature reports a consistent pattern: studies find hypothetical bias in total willingness-to-pay estimates and in opt-in rates, while frequently not finding it in the hypothetical marginal rates of substitution once scale is corrected. In plain terms, respondents inflate what the whole thing is worth and what proportion of them would buy it. They are far more accurate about relative value: how much more a larger size is worth than a smaller one, or how much one feature is worth against another.
That is why a well-built conjoint is more useful than a direct question even though both are hypothetical. It reads price out of relative choices rather than out of an absolute claim, and it forces the respondent to give something up, which is the mechanism the direct question is missing.
The practical implication for a brief is twofold. Use trade-off methods to establish price architecture: which feature belongs at which tier, and how much a step up is worth. Then treat any absolute price level from a survey as needing external validation before it sets a list price. Where pricing sits in the wider launch order is set out in the CPG launch research sequence.
How Do You Design a Pricing Study That Holds Up?
- Show a realistic competitive context. A price evaluated in isolation has no reference point. Prices are judged relatively, and a study without competitors measures nothing a market will reproduce.
- Make the choice consequential where you can. Incentive-compatible designs, where a respondent has some real stake in their answer, consistently reduce bias, and the mitigation literature above covers the mechanisms.
- Include a no-purchase option. Forcing a choice among products manufactures demand that will not exist.
- Test the range you might actually charge, not a range centered on your current price, since the design constrains what the analysis can find.
- Segment before you conclude. An average willingness to pay across a heterogeneous market frequently describes nobody, and the price that maximizes revenue in one segment can be the price that eliminates another.
Professional expectations for disclosure, respondent treatment and data handling in commercial studies of this kind are set out in the ESOMAR code and guidelines and the AAPOR standards and ethics materials.
Who Is Missing From a Pricing Study?
More consequential here than in almost any other method, because price sensitivity is not distributed evenly across a population and neither is the ability to take part in research.
Pew Research Center's mobile technology fact sheet reports that 16 percent of US adults are smartphone-only internet users. Pew puts that at 34 percent among adults in households under $30,000 a year against 4 percent above $100,000. A pricing study fielded only through browser-based panels systematically under-samples lower-income households.
Since those households are the most price-sensitive part of the market, the study will return a higher willingness to pay than the market actually holds, on top of the hypothetical bias already inflating it. The general form of that failure is mapped in sample validity and who you miss.
The two errors point the same way.
Globally the gradient is steeper again, and the ITU's connectivity statistics remain the reference for how uneven access is by country and income band. For any product priced for a mass market, or priced differently across markets, the sampling frame is a pricing decision in itself.
Where a study has to reach past that frame, the channel is what moves it. Alchemic fields interviews natively inside WhatsApp with no link and no app to install, and by outbound phone call for respondents reachable by voice rather than browser. Alchemic publishes 57+ languages including Hindi, Tamil, Telugu, Bangla, Arabic and Indonesian. Recruitment runs as managed fieldwork or bring your own, across fourteen markets that include the USA and the UK, and in India from metros and Tier 1 through Tier 2 and Tier 3.
It runs conjoint and full quantitative work alongside moderated qualitative in the same study, which is what lets a trade-off exercise sit next to probing on why a respondent chose as they did.
Where Pricing Research Gives You the Wrong Answer
Five limits worth stating before a study is commissioned.
It measures stated preference, not behavior. Every method above except the in-market test infers a decision from a proxy. The 3.2 average ratio is a property of that gap, not a flaw in any one technique.
It cannot anticipate competitor response. A price that wins against today's competitive set may not survive a rival's reaction to it, and no survey respondent knows what that reaction will be.
Brand equity moves willingness to pay and moves slowly. A pricing study run on an unfamiliar brand measures the price of an unfamiliar brand. Results do not transfer to the same product after brand-building work.
Conjoint output is only as good as the attribute list. If the attribute that actually drives choice is not in the design, the model will confidently allocate its influence to whatever is. This is the failure that most often makes a technically clean conjoint useless.
Segment averages hide the decision. A revenue-maximizing price computed across the whole sample can be worse than a segmented structure in every segment, which is why the segmentation and the pricing work belong in the same conversation.
One further limit sits outside the method entirely. Pricing research tells you what a market will bear; it does not tell you what you are permitted to say about the price once you set it. Comparative and reference-price claims, introductory pricing and discount framing are all regulated advertising conduct, and the Federal Trade Commission's advertising and marketing guidance is the governing reference in the United States.
A pricing architecture that tests well and cannot be described lawfully in market is not a finished decision. Where household budget context matters to the read, the Federal Reserve's survey of household economics and decisionmaking gives a public baseline for how much financial headroom households actually report having. It is a useful sanity check on any willingness-to-pay figure that implies discretionary spending a large share of the market does not hold.
Teams working out which stage of the launch this belongs to will get more from the qualitative research primer and the concept testing service overview. Pricing questions asked before a proposition is settled tend to measure confusion rather than value.