The Gabor-Granger pricing method
A ladder of prices, one yes-or-no question at each, and two curves out the other end: how many would buy, and what you would earn. Paste your ladder below — and see how much of the peak is real.
Short answer
What is the Gabor-Granger method?
A pricing survey that asks each respondent whether they would buy at a series of prices. The share saying yes at each price is a demand curve; price multiplied by that share is a revenue curve, and its peak is the revenue-maximising price.
The short version
- Developed by André Gabor and Clive Granger at the University of Nottingham in the 1960s. Granger shared the 2003 Nobel Prize in Economics.
- Input is two columns: the price, and how many of your respondents said they would buy at it.
- The peak is the revenue-maximising price — not the price most people accept, which is always the cheapest one tested.
- It measures stated intent, not purchase, and shows no competitor beside your product. Rank prices with it; do not forecast income.
Free · No signup · Nothing leaves your browser
Paste the ladder, get the demand and revenue curves
One row per price you tested: the price, then how many people said they would buy at it. Paste straight from a spreadsheet — extra columns are ignored.
Example: a SaaS plan tested on 60 people across eight prices.
8price points read
60respondents
$39
60% would buy, giving $23.40 of expected revenue per person you asked.
$29 – $59
With 60 respondents these prices cannot be told apart from the peak. Treat the band as the answer; the single number inside it is sharper than your data.
| Price | Would buy | Range, 95% | Revenue each |
|---|---|---|---|
| $9 | 55 · 92% | 82% – 96% | $8.25 |
| $19 | 50 · 83% | 72% – 91% | $15.83 |
| $29 | 44 · 73% | 61% – 83% | $21.27 |
| $39 | 36 · 60% | 47% – 71% | $23.40 |
| $49 | 27 · 45% | 33% – 58% | $22.05 |
| $59 | 20 · 33% | 23% – 46% | $19.67 |
| $79 | 11 · 18% | 11% – 30% | $14.48 |
| $99 | 5 · 8% | 4% – 18% | $8.25 |
# GABOR-GRANGER: [What you are pricing] Date: 2026-09-12 Respondents: 60 ## The ladder $9 55 buyers 92% (82%–96%) revenue/respondent $8.25 $19 50 buyers 83% (72%–91%) revenue/respondent $15.83 $29 44 buyers 73% (61%–83%) revenue/respondent $21.27 $39 36 buyers 60% (47%–71%) revenue/respondent $23.40 $49 27 buyers 45% (33%–58%) revenue/respondent $22.05 $59 20 buyers 33% (23%–46%) revenue/respondent $19.67 $79 11 buyers 18% (11%–30%) revenue/respondent $14.48 $99 5 buyers 8% (4%–18%) revenue/respondent $8.25 ## Result Revenue peaks at: $39 Expected revenue per respondent at the peak: $23.40 Prices this sample CANNOT tell apart from the peak: $29 – $59 ## What this does NOT say Respondents said they would buy. Nobody paid. Stated intent runs ahead of behaviour, so read the revenue curve as a ranking of prices, not as a forecast of income. The ladder also shows one product at one moment, with no competitor on the shelf next to it. ## Next [Would the peak survive a real checkout page?] [Is the band narrow enough to matter, or do you need more respondents?] Calculated with the free Gabor-Granger tool at https://gonogo.team/gabor-granger
What this cannot know. Everyone in the ladder said they would buy. Nobody paid. Stated intent runs ahead of behaviour, reliably and in one direction, so read the revenue curve as a ranking of prices rather than a forecast of income. The ladder also shows your product alone, with no competitor beside it on the shelf and no discount in the next tab.
Want the price people find believable rather than the one that maximises this curve? Van Westendorp answers that, and the two together bracket the question better than either alone. And whether anyone pays at all is one of the three critical criteria in our Go/No-Go rubric.
Everyone in the ladder said they would buy. Nobody paid. A 15-minute session ends with a written GO / WAIT / NO-GO and the reasoning behind it. Free tier, no card.
How to run the survey
- 1
Describe the product once, concretely
Same paragraph, same feature list, same unit for everyone — per month, per seat, one-off. Respondents price what they can picture; a vague description produces answers anchored to nothing.
- 2
Choose the ladder before you start
Five to eight prices, spanning wider than you think is sensible on both sides. The method can only find a peak among the prices you show, so a ladder that stops at your favourite number will return your favourite number.
- 3
Ask one question per price
“Would you buy it at [price]?” — yes or no. Randomise the order across respondents: walking the ladder upward makes people anchor on the first price and say yes for longer than they would.
- 4
Count, do not average
For each price you need the number of people who said yes, and the total number you asked. That pair is the entire input — everything else is arithmetic.
Where the peak comes from
The demand curve is a count: at each price, the share of respondents who said yes. It always falls as price rises, which is why the cheapest rung is never interesting on its own — of course more people would buy at $9.
The revenue curve is that share multiplied by the price. It rises while the extra money per sale outweighs the customers you lose, and falls once it does not. The turning point is the revenue-maximising price.
And then the honest part. The share at each price is an estimate from a sample, with real uncertainty around it. Run the arithmetic on that uncertainty and the peak usually stops being a point: with sixty respondents, a peak at $39 is typically indistinguishable from $29 and $49. The calculator draws that uncertainty as a band, because a band is what the data supports.
Gabor-Granger or Van Westendorp?
They are not rivals. They answer different questions, and the answers are most useful together.
| Gabor-Granger | Van Westendorp | |
|---|---|---|
| The question | “Would you buy at $X?”, repeated | Four open questions about what a price feels like |
| You supply | The prices to test | Nothing — respondents name the numbers |
| You get | Demand and revenue curves, a revenue peak | A range of believable prices, four named points |
| Blind to | Prices you did not think to test | How many would buy — no demand at all |
| Use it when | You have candidate prices and need to choose | You have no idea what the number should be |
The natural order is Van Westendorp first to bound the field, then Gabor-Granger on a ladder that spans that range. The Van Westendorp calculator is free too.
What Gabor-Granger cannot tell you
Stated intent is not purchase. Saying yes in a survey costs nothing and people say yes more often than they buy. The bias runs in one direction, so the curve ranks prices well and forecasts revenue badly.
There is no competitor in the room. Your product appears alone at one price, with no alternative to switch to and no discount in the next tab. Real buyers price by comparison. Conjoint analysis exists precisely to fix this, at the cost of a much heavier design.
It cannot find a price you did not test. If revenue peaks at the top rung of your ladder, the true optimum is above it and nobody was ever shown it. The calculator says so out loud rather than reporting the edge of your questionnaire as a market fact.
Order changes the answer. Prices shown in ascending order anchor respondents and inflate the apparent optimum. If your ladder was not randomised, the peak is biased upward and you cannot tell by how much.
What it does well is narrow the field. A defensible band, arrived at by a method you can explain in a sentence, beats a number chosen because a competitor charges it.
Four ways the ladder goes wrong
❌ Reading the peak as “the price”
The peak is one number computed from a sample. Move three people between answers and it moves. Calculators that print a single optimal price are reporting a precision the data does not contain.
✓ Instead: Read the band. If the sample cannot tell $29 from $49, then your evidence supports a range, and picking inside it is a business decision rather than a research finding.
❌ Presenting prices from cheapest to dearest
Anchoring. The first price sets the reference, and agreement runs on further than it should. An ascending ladder reliably produces a higher apparent optimum than a randomised one.
✓ Instead: Randomise price order per respondent. If your survey tool cannot, at least alternate ascending and descending halves of the sample.
❌ Treating “would buy” as “will buy”
Stated intent overstates behaviour, consistently and in one direction. Nobody in the survey opened a wallet, and the survey contained no competitor, no budget, and no reason to say no.
✓ Instead: Use the curve to rank prices, not to forecast revenue. If you need an absolute number, discount stated intent heavily, or run the price on a real checkout page.
❌ Smoothing a curve that goes the wrong way
Sometimes more people accept a higher price than a lower one. That means the data has a problem — a typo, different people at different rungs, or too few respondents. Smoothing makes the chart look right while making it less true.
✓ Instead: Find out why. The calculator on this page flags rising demand instead of smoothing it, because the flag is the useful output.
The question underneath the price
A revenue peak assumes there is demand to price. If the survey says $39 and nobody has ever paid you anything, the number is arithmetic on a hypothesis.
Will they pay at all is one of three critical criteria in our Go/No-Go rubric — the ones no amount of strength elsewhere can compensate for. A 15-minute session works through your idea across all seven and ends with a written verdict and the reasoning behind it.
15 min · free tier, no card
Frequently asked questions
What is the Gabor-Granger method?+
How is Gabor-Granger different from Van Westendorp?+
How many respondents do I need?+
What if the revenue peak is at the highest price I tested?+
Does Gabor-Granger account for competitors?+
Is this calculator free, and does my data leave the browser?+
Related guides