Gabor-Granger pricing: find the revenue peak, then assume it runs high

The Gabor-Granger method is a pricing survey that shows each person a price, asks whether they would buy at it, and then moves the price up or down until it finds the highest price they would accept. Add those answers up and you get a demand curve (the share of people who would buy at each price) and a revenue curve, whose peak is the "best" price. Use Gabor-Granger to compare a handful of prices for one product you can describe precisely, and treat the peak as an upper bound that needs a real buying test before you print it on the pricing page.

The method is quick to run and easy to explain to a finance team. Its weak spot is that it measures what people say about prices you chose, for a product shown with no competitors next to it.

1. Know where it came from, because the history explains the blind spots

André Gabor and Clive Granger developed the technique in the 1960s, according to Conjointly's product page and the Wikipedia entry on the method. Both worked on how shoppers react to prices. Their 1961 paper "On the Price Consciousness of Consumers" in the Journal of the Royal Statistical Society, Series C, asked shoppers in Nottingham whether they knew the price of an item at the time they bought it, across 15 grocery products (RePEc listing).

Granger spent 22 years at the University of Nottingham and shared the 2003 Nobel Memorial Prize in Economic Sciences for work on time series data, not for this survey (Wikipedia). So yes, a Nobel laureate's name is on your pricing survey, and no, that does not make the survey Nobel-grade.

The pair also wrote about price as a signal of quality. Their paper "Price as an Indicator of Quality: Report on an Enquiry" (Management Decision, volume 17, 1979) argues that price affects choice both as a cost and as a hint about quality. Keep that in mind. It is one reason a very low price in your survey can lose buyers instead of winning them.

2. Run it as a short ladder of prices with a random starting rung

The standard version, as described by Wikipedia and Sawtooth Software, works like this:

1. Describe one product exactly, including what the price buys and the unit (per seat per month, per order, one-time).
2. Pick a set of prices. Wikipedia's version and SurveyMonkey's guide use five, evenly spaced. Conjointly suggests 8 to 15 levels (Conjointly).
3. Show the respondent one price chosen at random and ask a five-point purchase question, from "definitely would buy" to "definitely would not buy."
4. If they pick one of the top two answers, show a higher price. If not, show a lower one.
5. Stop when you find the highest price they would accept. Wikipedia calls that number the Gabor-Granger variable.
6. People who would not buy even at the lowest price are usually coded as zero or dropped.

Sawtooth explains the random start as a way to reduce anchoring, where the first number someone sees pulls later answers toward it. Start everyone at the same price and you are partly measuring that price.

Use 6 to 8 prices, spaced evenly, with the lowest one below anything you would actually charge and the highest one above anything you would dare to charge. The survey can only ever pick a price from your list, so a list that stops too low will crown its own top rung.

3. Read the revenue curve, then look at how flat its top is

Each person's highest accepted price tells you they would buy at that price and every lower one. So the share willing to buy at a price is the share whose highest accepted price is that price or above. Multiply each price by its share and you get expected revenue per respondent (Sawtooth).

Here is a made-up example. Priya runs pricing at a small company that sells shift-scheduling software to dental clinics, priced per clinic per month. She tests six prices and gets 150 completed answers from clinic owners, above the 100 per segment that SurveyMonkey suggests. Her results, and the revenue for every 100 respondents:

The revenue peak is $80. But $100 earns only $160 less per 100 respondents, about 4% below the peak, with 12 fewer customers. With 150 answers, the 52% at $80 has a 95% margin of error of about 8 points either way, so revenue at $80 could plausibly sit anywhere from about $3,520 to $4,800 per 100. A gap of $160 is well inside that noise, so the honest reading is "somewhere from $80 to $100."

Now add cost. Sawtooth notes that if you know your costs you can find the profit-maximizing price too. Say each clinic costs Priya $30 a month to serve. Profit per 100 respondents is $2,600 at $80 ($50 margin times 52) and $2,800 at $100 ($70 margin times 40). Once you count the cost of serving each customer, the best price moves up a rung.

Short version: report the flat top of the curve as a range, and redo the math with margin before picking a point.

4. Assume the answers run high, especially for pricier products

People say they will pay more than they do. A 2020 meta-analysis by Schmidt and Bijmolt in the Journal of the Academy of Marketing Science pooled 77 studies from 47 papers and found stated willingness to pay ran about 21% above real willingness to pay on average (abstract). The bias was larger for higher-valued products and for specialty goods, and the authors say those answers need a bigger downward adjustment. If your product is expensive or specialized, expect the gap to be wider than average. For the wider picture, see why every willingness to pay answer is a ceiling.

There is a pull in the other direction too. Conjointly notes that because the price is shown openly, people can guess the study is setting a price and may understate what they would pay. It says this offsets the usual habit of overstating purchase intent, and that the two biases mostly cancel out. That is a vendor's view, and nobody has shown they cancel for your market. Plan as if the net effect is upward.

5. Remember what the survey cannot see

Sawtooth lists the main limits (Sawtooth): the method asks about one product in isolation, ignores competitors, says nothing about how features other than price shape the choice, and assumes a stable market where people know the category.

Three practical consequences follow:

1. Competitors are invisible. A clinic owner who already pays for a rival tool at $60 may say "probably would buy" at $90 in a survey, then keep the rival tool in real life.
2. Price changes what the product seems to be. Gabor and Granger's own work on price as a quality signal means a very low rung can look suspicious rather than attractive, which bends the low end of the curve.
3. The best price is limited to your list. If you tested $80 and $100, the survey cannot tell you that $89 is better.

If competitors or feature tradeoffs drive the decision, use a choice-based method instead. Conjoint analysis shows people full product options side by side, with price as one attribute.

6. Pick Gabor-Granger over Van Westendorp when you need to know how many would buy

The two methods answer different questions. The Van Westendorp price sensitivity meter lets people name prices that feel too cheap, a bargain, getting expensive and too expensive, and gives you a range of acceptable prices. Gabor-Granger hands people your prices and gives you a share who would buy at each, so you can compute revenue (Sawtooth).

They can disagree sharply. In a vendor write-up, Conjointly's Hugh Zhao reports a floor lamp study where Van Westendorp gave an acceptable range of $24.99 to $35.89 and Gabor-Granger put the revenue-maximizing price at $35. In an instant coffee case, Van Westendorp's range was $5.75 to $7.67 while Gabor-Granger landed on $13.89. The post does not report sample sizes, so treat these as illustrations of disagreement, not benchmarks.

When you have no idea where the price should sit, run Van Westendorp first to set the range, then run Gabor-Granger with rungs inside and slightly beyond that range. Conjointly also says Gabor-Granger can supplement a Van Westendorp study.

7. Skip it when the buyer is a committee or the market is crowded

Gabor-Granger fits a product people can picture from a description, sold to individuals who decide on their own, in a category without a dominant competitor to compare against. Consumer subscriptions and simple self-serve software fit.

It fits poorly in three cases:

1. Enterprise sales. The person answering may not control the budget, and the final price comes out of a negotiation. Settle who pays and what you charge for first.
2. Crowded categories. Buyers will compare, and the survey does not let them.
3. Products people cannot imagine yet. A "probably would buy" for something nobody has used is a guess about a guess. SurveyMonkey also says the method works better for familiar product types.

In those cases, talk to the people who set and approve prices in your market. A few conversations with B2B pricing leaders or chief revenue officers can tell you how buyers actually budget for a tool like yours. Instant Expert lets you invite specific people like these to a short paid call, and you pay only if they book or answer. For the interview itself, use this guide to researching willingness to pay without asking for a guess.

8. Confirm the winning price with a real buying step

The survey narrows the choice. It does not make it. Take the range from section 3, offer the price at the top of it to real prospects, and see who signs, pays or asks for a quote. A stated "probably would buy" at $100 is a hypothesis until someone pays $100. This guide to price testing ranks the live tests by how much traffic and risk each needs.