Last updated: 18 September 2026
Quick Answer: The Van Westendorp price sensitivity meter is a four-question survey that maps the price range buyers find credible. The four ask where it becomes too cheap to trust, cheap, expensive but worth buying, and too expensive. Plotted as cumulative curves, their crossings give a range, not a recommended price.
A published study of Amazon Prime Video in India ran this instrument on 263 respondents and reported an optimal price point of 1,300 rupees inside an acceptable range of 1,000 to 1,500. The annual plan listed at 1,499 rupees on 1 April 2022, the top of that band, which is the range read as permission rather than as a map of where credibility runs out.
Neither the four Van Westendorp questions nor the acceptable range is a single fixed object, and no explainer says so. Van Westendorp's own wording asks where a price perception begins; the copied version asks whether a price is a bargain. Two published definitions read the lower bound off different curve crossings, so two analysts with identical data can publish different ranges, both following a documented convention.
Why the Wording of the Four Questions Changes the Answer
The wording does more work here than the arithmetic. Peter van Westendorp introduced the method at the 29th ESOMAR congress in 1976, and the original phrasing asks at which price the respondent begins to experience the product as cheap, expensive, too expensive and too cheap. Each is a threshold question, asking where a perception starts.
The version in general circulation is not that. Its second question asks at what price the product would be a bargain, or a great buy for the money. A recent review of willingness-to-pay measurement states the four questions in that form, which is what the industry fields.
The substitution moves the curve you plot.
- "Beginning to experience as cheap" asks for a boundary. The respondent reports where one perception gives way to another.
- "A bargain, a great buy for the money" asks for a verdict. The respondent reports a price they would feel good about paying.
A verdict price sits lower than a boundary price for most people. Pull the cheap curve down and the lower bound moves with it, and nothing in the output flags it.
- Fix the wording before fielding and record it with the result. A range compared against a wave run on the other wording is not like-for-like.
- Keep the four questions in a fixed order across waves.
Pew Research Center's work on question construction sizes that kind of effect: in 2008, 58 percent named the economy the most important election issue from a list, against 35 percent who volunteered it. Disclosure expectations for commercial work sit in AAPOR's standards and ethics.
How Do You Turn Four Price Answers Into Four Curves?
Each question becomes a cumulative distribution across the price range, counted in respondents, not averaged. By convention the "too cheap" and "cheap" curves are inverted, counting respondents at or below a given price while the expensive curves count upward. Without the inversion the lines do not cross where the method says they should.
Respondents whose four answers run out of order are removed before analysis. The R pricesensitivitymeter package applies that rule and reports invalid cases alongside the sample. Report that count. Open-ended numeric answers lose people upstream: a Census Bureau web experiment left 8.1 percent of plain utility-cost write-ins blank, and an opt-out pushed missing down-payment answers from 7.0 to 26.9 percent.
Each crossing answers a narrower question than its name suggests. The rows run from the lowest crossing upward, and in the Prime Video study the outer two landed at 1,000 and 1,500 rupees.
| Point | Which curves cross | What it is read as | What it does not tell you |
|---|---|---|---|
| Point of marginal cheapness (PMC) | Too cheap against the inverse of cheap | Lower bound of the range | Not a profitability floor, a perception boundary |
| Indifference price point (IDP or IPP) | Cheap against expensive | Equal numbers call it cheap and expensive. Van Westendorp read it as the median price paid, or a leader's | Not a recommendation. It describes where the market is |
| Optimal price point (OPP) | Too cheap against too expensive | Equal numbers reject it as too low and as too high, so resistance is lowest | Not the revenue-maximizing price. Minimum resistance is not maximum return |
| Point of marginal expensiveness (PME) | Too expensive against the inverse of expensive | Upper bound of the range | Not a ceiling you can charge up to. Where credibility ends |
None of the four points is a revenue forecast. What a pricing study itself costs, and which units it bills on, is in market research costs and pricing models; why a stated answer overstates the purchase that follows is in what purchase intent scores actually predict.
Which Acceptable Price Range Did Your Vendor Report?
Two documented definitions of the acceptable range exist, they use different curve crossings, and they produce different numbers from the same data.
- The original definition, from van Westendorp's paper. The lower bound is the intersection of "too cheap" with the inverse of the "cheap" curve; the upper bound is "too expensive" against the inverse of "expensive".
- The narrower definition, applied by some research firms. The lower bound is "expensive" against "too cheap"; the upper bound is "too expensive" against "cheap". The package documentation notes this yields a narrower range, and that the optimal price point can land above its upper bound.
This is not a theoretical split. The Prime Video study cited at the top uses the second convention, defining the point of marginal expensiveness as the crossing of "too expensive" with the bargain curve. Its point of marginal expensiveness lands at 1,100 rupees, below the same analysis's 1,300-rupee optimal price point. The R package defaults to the first. Both are defensible. Without a label they are not comparable.
Ask any result for the wording fielded, the range definition applied, and the count dropped at validation.
Whose Reference Price Are You Measuring?
The four questions only work on someone who already has a price for the category. A respondent without one still types four numbers that plot like an informed answer. Anchoring fills the gap: a PLOS ONE experiment on 799 shoppers found a high anchor lifted stated willingness to pay 48 percent in hypothetical decisions, against 14 percent when real money changed hands. The Prime Video sample was 59.5 percent aged 18 to 24 and 58.3 percent students.
Price sensitivity concentrates where a browser panel reaches least well, with 16 percent of US adults smartphone-only rising to 34 percent in households under $30,000. Alchemic fields interviews natively inside WhatsApp with no link or app, and by outbound AI phone call where an out-of-order set can be queried in session rather than deleted.
Where the Price Sensitivity Meter Breaks Down
Two limits belong to the instrument, not to any study.
- It measures stated perception with no purchase consequence. A meta-analysis of 77 studies puts hypothetical willingness to pay 21 percent above what people really pay. The chart maps price credibility, not demand.
- It carries no competitive context. Two respondents can give identical numbers while pricing against different alternatives, and the method averages them.
Pricing questions asked before the proposition is settled measure confusion, not value, which is what concept testing is for. The full comparison of pricing methods sets out when Gabor Granger or a discrete choice design is the better tool.