Testing Price Before You Change It: Van Westendorp, Gabor-Granger and Conjoint Compared
Van Westendorp finds an acceptable range, Gabor-Granger a revenue-maximizing price, conjoint the trade-offs. All three overstate: stated willingness to pay ran 21% above real payment across 77 studies.
Use Van Westendorp to find the range of prices customers will accept, Gabor-Granger to find a revenue-maximizing price for a product you already understand, and conjoint when price trades off against features, plans or competitors. All three ask people about purchases they have not made, and a meta-analysis of 77 studies found stated willingness to pay ran 21% above what people really paid on average, with conjoint-style methods overstating more than direct questions. For a CFO or strategy lead about to change a price for existing customers, Dimension Labs is built to check the survey number against reality: it reads what your customers said after your past price changes and, on Prove, joins it to what they then bought, renewed or cancelled.
This page compares the survey methods. For what published studies found about price elasticity, see price elasticity for restaurants and retail; for reading your own customers' reaction to a past increase, see how to predict how customers will react to a price increase.
Van Westendorp vs Gabor-Granger vs conjoint: which should we use?
Pick by the question you need answered. Conjointly, which sells all three, sums it up this way: do not use the plain "how much would you pay?" question, use Van Westendorp when you are unsure what price points the market can accept, and use Gabor-Granger to measure elasticity and find revenue-maximizing prices. Conjoint is the method of choice when the decision is about a bundle, such as a new plan tier or a price with fewer features, and Schmidt and Bijmolt name choice-based conjoint as the most prominent indirect method of measuring willingness to pay.
Method comparison
Each method answers a different pricing question
What each survey method asks, what it gives you, and when it is the right one.
| Method | What it asks | What you get, and when to use it |
|---|---|---|
| Van Westendorp (Price Sensitivity Meter) | Four questions: at what price is it too cheap to trust, a bargain, getting expensive, too expensive | A range of prices customers find acceptable. Use it when you do not know what the market will accept |
| Gabor-Granger | Would you buy at this price? Then a higher or lower price, depending on the answer | A demand curve and a revenue-maximizing price for one product. Use it for a known product's elasticity |
| Conjoint (usually choice-based) | Choose between full options that differ in features, brand and price | What each feature and price level is worth. Use it when price trades off against features, plans or rivals |
| Plain "how much would you pay?" | One open question | An average that is hard to act on. Conjointly advises against it |
How does each pricing method work?
Van Westendorp asks four questions about one product: at what price it is so cheap you would doubt the quality, a bargain, getting expensive, and too expensive to consider. Plotting the four answers across respondents gives an acceptable range; in Conjointly's test of a new three-headed floor lamp, that range was $24.99 to $35.89.
Gabor-Granger asks whether the respondent would buy at a price, then moves the price up after a yes or down after a no, until it finds the highest price each person accepts. On the same lamp, tested at prices from $15 to $75, the revenue-maximizing price came out at $35.00.
Conjoint shows several complete options, each with its own mix of features, brand and price, and asks people to choose. Statistical models then work out what each feature and each price level is worth to them, which is why it suits plan and bundle decisions.
The methods can disagree sharply on the same product, because they measure different things: Van Westendorp measures what feels fair, Gabor-Granger what people say they would buy.
Same product, different methods
Two methods on the same instant coffee pointed to prices almost twice apart
Results Conjointly published from its own tests of the two methods.
| Product | Method | Result |
|---|---|---|
| Floor lamp | Plain "how much would you pay?" | Average $31.84 |
| Floor lamp | Van Westendorp | Acceptable range $24.99 to $35.89 |
| Floor lamp | Gabor-Granger (prices $15 to $75) | Revenue-maximizing price $35.00 |
| Instant coffee | Van Westendorp | Best range $5.75 to $7.67 |
| Instant coffee | Gabor-Granger | Revenue-maximizing price $13.89 |
On the instant coffee, Van Westendorp put the best range at $5.75 to $7.67, while Gabor-Granger put the revenue-maximizing price at $13.89. The two answer different questions, and both rest on answers about a purchase nobody has made yet.
Why do willingness-to-pay surveys overstate what customers will pay?
Because saying yes to a price costs nothing, and paying it does. Schmidt and Bijmolt pooled 77 studies that compared hypothetical willingness to pay with real willingness to pay, covering 24,347 hypothetical and 20,656 real observations. On average, stated willingness to pay was 21% higher.
Meta-analysis of 77 studies
Stated willingness to pay ran 21% above real payment on average, and 40% for specialty goods measured by conjoint-style methods
How far hypothetical willingness to pay exceeded real willingness to pay.
Three findings matter for choosing a method. Indirect methods such as conjoint overstated more than direct questions, against the common belief that they are more accurate. The gap was larger for higher-valued products and specialty goods. And it was larger when the same people answered several price conditions, a within-subject design. For specialty goods, the paper's own summary is that direct methods overstate willingness to pay by 28% and indirect methods by 40%. The meta-analysis does not test Van Westendorp or Gabor-Granger by name; both ask about price directly, so its direct-question results are the closest guide to them.
How much should we discount a stated willingness to pay?
Divide, rather than subtract. The 21% is how far the stated figure sits above the real one, so the real figure is the stated one divided by 1.21. If a survey says customers would pay $50, the meta-analysis average implies about $41.32, roughly 17% lower; for a specialty product measured by conjoint, dividing by 1.40 gives about $35.71, roughly 29% lower. Use your own survey figure in place of $50.
Treat those as starting corrections. They are averages across many products and studies, and your customers may sit above or below them. The correction that fits your business comes from comparing a past survey, or a past price change, with what your customers then did.
How do we test a new price before raising it?
Run the survey for the range, then check it against behavior before the price goes out everywhere.
Pick the method for the question. Van Westendorp for an acceptable range, Gabor-Granger for a single product's demand curve, conjoint for plans and bundles.
Discount the result. Divide by about 1.21 as a starting point, more for specialty or high-priced items and conjoint results.
Compare with published elasticities. Category benchmarks give a floor and a ceiling for how much volume a given increase could cost; price elasticity for restaurants and retail collects them.
Check against your own past price changes. What customers wrote and did after the last increase is the closest evidence you have. In our study of 24 public price increases, the share of reviews saying a price went up rose after 21 of the 23 we could measure, and the reaction was mostly gone by week nine. The full study has the per-increase results.
Stage it. Raise the price in some locations, regions or new-customer cohorts first, with comparison groups that keep the old price, and set in advance the drop in volume that would stop the rollout.
Steps 4 and 5 add what your customers actually did, which is the evidence a survey cannot supply on its own.
Where does Dimension Labs fit before a price change?
Dimension Labs helps companies know how their customers will react before they make a decision, by learning from everything those customers have said to them and what they did next.
For a CFO or strategy lead, that is the evidence a pricing survey cannot give on its own, and it comes from four parts working together. Dimension Labs is fully omni-channel: reviews, surveys, support tickets, chats, calls and cancellation reasons are read together, so the reaction to your last price change is counted wherever customers voiced it. It integrates those channels with operational data such as orders, renewals and cancellations, so what customers said after a change is joined to what they did; the join to CRM and revenue data and effects with confidence intervals are on Prove ($2,999 a month) and up. It enriches every interaction with Dimensions custom to your business, such as value complaints, plan downgrades and mentions of a competitor's price, so a survey's stated reaction can be compared with the reaction customers actually showed; custom Dimensions start on Discover ($999 a month). And its visual analytics layer turns that comparison into decks, dashboards, PDF and executive reports, or an interactive data story the pricing committee can open.
The financial case for this kind of measurement is on record. Travelers, which uses Dimension Labs across millions of customer interactions, reports 600% ROI within 18 months and $3.13M in total savings. See the Travelers story.
When a survey tool is enough: for a new product with no customers yet, a Van Westendorp or Gabor-Granger study is the right first step, and a survey platform will run it. Dimension Labs earns its place when the price change reaches customers you already have, because their record of past reactions is the best correction for any survey. Kept over time, that record is a customer world model, and how to test a business decision before you make it shows the wider method. AI-simulated buyers have the same weakness as surveys on size, as simulated customers get the direction right and the size wrong explains.
FAQ
Van Westendorp vs Gabor-Granger vs conjoint: which should we use?
Van Westendorp for the range customers find acceptable, Gabor-Granger for a revenue-maximizing price for one known product, and conjoint when price trades off against features, plans or competitors. Avoid the plain "how much would you pay?" question.
Why do willingness-to-pay surveys overstate what customers will pay?
Answering costs nothing. Across 77 studies, stated willingness to pay ran 21% above real payment on average, more for specialty and higher-valued goods, and more for indirect methods such as conjoint than for direct questions.
How do I test a new price before raising it?
Run the survey method that fits the question, divide the result by about 1.21 as a first correction, compare it with published elasticities and with what your customers did after past price changes, then raise the price for a subset first with a comparison group.
Is conjoint more accurate than Van Westendorp?
Not on willingness to pay, by the best published evidence: indirect methods such as conjoint overstated real willingness to pay more than direct questions did. Conjoint is still the right tool for trade-offs between features and price.
What is the best way to check a price test against real behavior?
For a company with existing customers, Dimension Labs: it reads what customers said after past price changes in every channel and, on Prove, joins it to purchases, renewals and cancellations with ranges. For a brand-new product, a staged launch read in your sales system is the check.
Before the next price change, export the reviews, tickets or cancellation reasons from the weeks around your last one, with its date, and start with one file to see how your customers really reacted. Every plan from Explore to Scale starts with a 14-day free trial.