Gabor-Granger Method: Building a Demand Curve from Yes/No Answers

The Gabor-Granger method is a price-sensitivity research technique developed by André Gabor and Clive Granger, two British economists, in the mid-1960s. Each respondent sees one product at one price and answers a simple yes/no (or 5-point intent) "would you buy this at $X?" question. If they say yes, the next question shows a higher price. If they say no, the next shows a lower price. The yes/no staircase continues until each respondent's individual price ceiling is found.

The output is a cumulative purchase-probability curve at the sample level. Plot the percentage who said yes at each price point against price, and you have a price-demand curve for the product. Multiply by predicted market size and you get a volume forecast at each price. Multiply by price and you get a revenue curve, with a peak that signals the revenue-maximising price.

The mechanic is monadic. Each respondent evaluates one product, in isolation, with no competitor on the screen. That keeps the survey short and easy to field, but it is also the method's main limitation. Real shoppers compare. Gabor-Granger asks them to commit before the comparison.

How it differs from the Van Westendorp PSM. PSM asks four direct questions (too cheap, cheap, expensive, too expensive) and produces an acceptable price RANGE, not a demand curve. Gabor-Granger produces the curve but does not test the quality-floor concerns the PSM picks up. The two methods are complementary, not interchangeable. Use PSM to find the corridor, use Gabor-Granger to locate the optimum within it.

How it differs from conjoint. Conjoint forces respondents to trade off bundles of features at different prices, and lets you decompose willingness-to-pay attribute by attribute. Gabor-Granger does not. If the product is finished and the question is "what price?", Gabor-Granger is faster and cheaper. If the question is "which features are worth paying for?", conjoint is the only method that answers it.

At each tested price point P:

Purchase Probability(P) = share of respondents who said yes (or "definitely" / "probably" on a 5-point intent scale)

Predicted demand:
Q(P) = Market Size × Awareness × Distribution × Purchase Probability(P) × Purchase Frequency

Predicted revenue:
R(P) = P × Q(P)

Revenue-maximising price P* solves dR/dP = 0.
Profit-maximising price P** solves d[(P − VC) × Q(P)]/dP = 0.

P* is always at or above P. Once contribution margin enters the math, every dollar of additional price drops to profit while volume falls more slowly than 1:1 in the relevant range.

Hypothetical-bias correction (Juster scale). Stated purchase intent overstates real-world buying behaviour. Industry practice in FMCG concept testing is to multiply the raw "yes" share by intent-class-specific correction factors before treating it as a probability:
- "Definitely would buy" → multiply by ~0.70
- "Probably would buy" → multiply by ~0.30
- "Might or might not" → multiply by ~0.05
- "Probably/definitely would not" → 0

These weights are widely used in FMCG to bring stated intent closer to real-world conversion. Without this step, raw Gabor-Granger output regularly overstates demand by 40 to 60 percent.

A juice manufacturer tests a new cold-pressed-style line extension at six price points with 400 respondents in the target consumer group.

Raw "definitely" plus "probably" would-buy:
- $3.99: 72 percent
- $4.49: 65 percent
- $4.99: 54 percent
- $5.49: 38 percent
- $5.99: 22 percent
- $6.49: 11 percent

After Juster-scale correction (0.70 / 0.30 weights):
- $3.99: 38 percent
- $4.49: 33 percent
- $4.99: 27 percent
- $5.49: 18 percent
- $5.99: 10 percent
- $6.49: 5 percent

Assumed market size 200,000 target consumers, 80 percent awareness, 75 percent distribution.

Predicted demand (units per month) and revenue at each price:
- $3.99: 45,600 units, $181,944 revenue
- $4.49: 39,600 units, $177,804 revenue
- $4.99: 32,400 units, $161,676 revenue
- $5.49: 21,600 units, $118,584 revenue
- $5.99: 12,000 units, $71,880 revenue
- $6.49: 6,000 units, $38,940 revenue

Revenue maximises at $3.99. But variable cost is $2.40 per pack, so contribution per unit changes the answer.

Profit-maximising price: $4.49.
Profit at $4.49 = 39,600 × ($4.49 − $2.40) = $82,764 per month.
Profit at $3.99 = 45,600 × ($3.99 − $2.40) = $72,504 per month.

The 50-cent step from $3.99 to $4.49 costs 13 percent of volume and adds 14 percent to profit. The Gabor-Granger demand curve made that trade-off visible in numbers a CFO can sign off on. Without the curve, the team would likely have launched at $3.99 (the headline revenue maximum) and left around $10K of monthly profit on the table.

Price optimisation is profit optimisation, not volume optimisation. A demand curve and a contribution-margin number, read together, are usually enough to find the right answer.

Use Gabor-Granger when the product is finished, the price set is narrow, and the team needs a defensible demand curve fast. Common cases: pricing a line extension, validating a list-price increase before a customer negotiation, sense-checking a planned promotional depth.

Skip Gabor-Granger when one of these is true:
- The product is still being defined. Conjoint is the right tool, because it lets you value features individually.
- The competitive context is the whole story. Gabor-Granger evaluates one product in isolation. If the shopper's real choice depends on what is on the shelf next to it, no monadic test will catch that.
- The category is genuinely novel. With no reference frame, respondents anchor on whatever number you show first, and starting-price bias dominates the result.

Three execution rules separate good Gabor-Granger studies from bad ones.

First, randomise starting prices across respondents. Otherwise the first price you show becomes an anchor and your "demand curve" is partly a measurement artefact.

Second, choose at least five price points and bracket the realistic range generously on both sides. The output is bounded by the prices you tested. If your highest tested price still gets a 40 percent yes share, you have not found the ceiling and your revenue curve has no peak.

Third, segment the results by consumer type before drawing the demand curve. Aggregate curves average out the heterogeneity that determines actual behaviour. A 60 percent yes share at $4.99 might be 80 percent for premium loyalists and 30 percent for value shoppers, and those two segments need very different commercial strategies.

What Gabor-Granger does not catch:
- Stated-intent bias. Real-world purchase is swayed by promotion, distribution, shelf adjacency, and a hundred other factors not visible inside the survey.
- A gameable structure. The yes/no staircase is transparent to repeat respondents, who learn to say "no" first to bring the price down.
- Competitive context. The number you read off the curve is what the product is worth to consumers in isolation, not what it is worth versus the competitor SKU sitting next to it.

Treat Gabor-Granger as one input among several, not as a price oracle. Pair it with scanner-data elasticity (revealed preference) and with a Van Westendorp PSM corridor before locking a final price.