Willingness to pay can be measured, but only if the method fits the question and the sample fits the market. We design and run pricing research for B2B companies, from customer interviews to conjoint studies, and translate the results into prices, tiers, and segment rules.
The most expensive mistake in pricing research is choosing the method first. A company decides it needs a conjoint study, spends three months on it, and discovers the real question was whether one segment was being overcharged, which its own invoice data could have answered in a week.
So a pricing research consultant should begin by asking which decision the evidence has to support, and how precise it needs to be. “Should the premium tier cost 20% or 40% more than the standard one?” needs a different study from “Which of our five features actually drive the purchase?” or “Will our installed base accept a 7% increase?”
Three constraints then shape the design:
Use the question you are trying to answer to find the starting method. Most engagements combine two: one qualitative, one quantitative.
| The question you have | Start with | Add if needed |
|---|---|---|
| Why do we win and lose on price? | Win and loss interviews, CRM deal data | Transaction analysis by segment |
| Which features drive the purchase decision? | Customer interviews | MaxDiff |
| What price range will this market accept for a new offer? | Van Westendorp | Customer interviews to explain the range |
| Which price maximizes revenue for a defined offer? | Gabor-Granger | In-market price test |
| How much will buyers trade for each attribute, including price? | Conjoint analysis | Interviews to set the attributes |
| Are similar customers paying very different prices? | Transaction analysis | Interviews with the outliers |
| Will an increase stick with the installed base? | Transaction and churn history | Controlled price testing on a subset |
Every method measures something slightly different, and each has a characteristic way of going wrong. The sample figures below are typical ranges for B2B studies, not rules.
| Method | What it measures | Typical B2B sample | Time | Where it misleads |
|---|---|---|---|---|
| In-depth customer interviews | Value drivers, buying process, reasons behind price reactions | 15 to 25 per segment | Two to four weeks | Small samples cannot size an effect; interviewer bias if questions lead |
| Van Westendorp price sensitivity | The range of prices buyers consider acceptable | 100 or more per segment | Three to five weeks | Ignores competitors; respondents anchor on what they pay today |
| Gabor-Granger | Stated purchase likelihood at set price points | 100 or more per segment | Three to five weeks | Stated intent overstates real buying, often heavily |
| Conjoint analysis | Trade-offs between attributes and price, share at each price | 150 to 400 in total | Six to 10 weeks | Only as good as the attribute list; long surveys tire busy buyers |
| MaxDiff | Relative importance of features or messages | 100 or more | Three to five weeks | Ranks importance but says nothing about price levels |
| Transaction analysis | Revealed willingness to pay in real deals | All invoice lines, usually 24 months | Two to four weeks | Shows what was accepted, not what might have been |
| In-market price test | Actual behavior at a new price | Depends on deal volume | One to two quarters | Slow in low-volume B2B; results can be distorted by rep behavior |
The pattern in that table is the reason a willingness to pay survey on its own rarely settles a B2B price. Stated methods are quick and cheap and systematically optimistic; revealed methods are reliable and slow. The strongest studies use one to check the other.
In most B2B markets customer interviews produce more usable pricing insight per dollar than any survey. They also set up every quantitative study that follows, because they tell you which attributes and price levels are worth testing.
How we run them:
Sample across roles and outcomes.
Current customers, lapsed ones, and prospects who chose someone else, and within each account the user, the economic buyer, and where possible procurement. Fifteen to 25 conversations per segment is usually where new interviews stop adding themes.
Never ask “what would you pay?”
Buyers answer that question strategically. We ask what they compared the offer against, what it replaced, what went wrong the last time a supplier failed them, and what that cost. Price sensitivity emerges from the answers.
Quantify the value drivers in the buyer’s own units.
Hours of downtime, days of delay, the cost of an error. Those numbers feed the value based pricing calculation directly. Buyers rarely dispute figures they supplied.
Separate the buying reason from the renewal reason.
What wins the first contract and what keeps the account are often different, and pricing should reflect both. Switching risk dominates at signing; service reliability dominates at renewal.
Interviews are also where the findings no dataset shows tend to appear: a service attribute buyers quietly pay extra for, a competitor nobody had listed, a budget cycle that makes a price increase easy in March and impossible in October.
Van Westendorp pricing asks four questions about an offer: at what price would it be so cheap you would doubt its quality, at what price a bargain, at what price getting expensive, and at what price too expensive to consider. Plotting the cumulative answers produces four curves whose crossing points define a range.
A B2B software add-on, annual subscription, 140 respondents from the target segment. The table shows the share of respondents choosing each answer at or beyond each price.
| Annual price | Too cheap | A bargain | Getting expensive | Too expensive |
|---|---|---|---|---|
| $6,000 | 41% | 74% | 8% | 2% |
| $7,000 | 24% | 55% | 17% | 5% |
| $8,000 | 12% | 36% | 31% | 11% |
| $9,000 | 6% | 21% | 49% | 22% |
| $10,000 | 3% | 11% | 66% | 38% |
| $11,000 | 1% | 5% | 80% | 57% |
Interpolating between the rows gives four points:
Point of marginal cheapness, about $7,270, where “too cheap” crosses “getting expensive”. Below this, more buyers doubt the offer than find it costly.
Optimal price point, about $8,060, where “too cheap” crosses “too expensive”: the price at which the fewest respondents reject the offer outright.
Indifference price point, about $8,150, where “a bargain” crosses “getting expensive”: the price the market treats as normal.
Point of marginal expensiveness, about $8,960, where “a bargain” crosses “too expensive”. Above this, resistance climbs quickly.
The acceptable range is therefore roughly $7,300 to $9,000. How to use it: the range is a guard rail, not an answer. It excludes prices that would damage the offer’s credibility or stall adoption, and it tells you the market’s current reference point. It does not account for the alternatives a buyer weighs, so a price near the top of the range is only justified when the value case supports it.
Gabor-Granger shows respondents a sequence of prices and asks whether they would buy at each. Multiplying each price by the share who say yes gives a revenue index.
Five price points for a unit sold in volume.
| Price | Would buy | Revenue index (price x share) |
|---|---|---|
| $80 | 78% | 62.4 |
| $90 | 70% | 63.0 |
| $100 | 61% | 61.0 |
| $110 | 45% | 49.5 |
| $120 | 30% | 36.0 |
Revenue peaks at $90, but the curve between $80 and $100 is almost flat: the index moves from 62.4 to 63.0 to 61.0. That flatness is the most useful finding. It means the choice between $90 and $100 should be made on margin, not revenue, and at $100 each unit sold earns more contribution. The steep fall above $100 is the real signal: that is where demand breaks.
The warning: stated purchase intent is optimistic, often heavily. Use Gabor-Granger to find the shape of the curve and where it breaks, not to forecast volume.
Conjoint analysis for pricing shows respondents a series of choices between offers that combine different features and prices, and infers how much each attribute is worth from the choices they make. It is the most informative stated method, and the one most often under-sampled in B2B.
A widely used rule of thumb for the minimum sample in a choice-based study is:
where n is respondents, t is choice tasks per respondent, a is alternatives per task, and c is the largest number of levels on any attribute.
Price is tested at five levels, the largest of any attribute, so c = 5. Each respondent completes 10 tasks with three alternatives each. Then n ≥ 500 x 5 / (10 x 3) = 83.3, so at least 84 respondents. That minimum applies to every segment you want to read separately. With three segments, the study needs about 252 completes, and most practitioners would add a margin above that.
In a B2B market of 2,000 reachable buyers, 252 completes is one in eight of the population, which is rarely achievable. The design responses, in the order we usually try them:
Cut the levels on price. Four well-placed levels instead of six reduce c and the required sample.
Reduce the attribute list to the ones the interviews showed actually matter.
Pool segments where the interviews suggest similar trade-offs, and read the others qualitatively.
Replace conjoint with interviews plus Van Westendorp when the population simply cannot support it.
A smaller study read honestly beats a large one read on 40 responses.
Almost every Acustrategy engagement includes some form of customer research, and the findings are rarely what the client expected.
The same product, three different sensitivities.
Survey and interview work for a global industrial chemicals and equipment maker found price reactions split by geography. Accounts in Asia-Pacific pushed back hard on price; North American and European accounts mostly did not. A blanket increase in the past had therefore hurt exactly where it should have been avoided. Priced region by region afterwards, the business saw realization improve by 8%, according to the client.
What customers would pay extra for, and what nobody had sold them.
Perception research across a subscription services base of about 18,000 US accounts showed a willingness to pay more for quicker response, low awareness of the full service range, and unmet demand for analytics. The transaction side of the same study surfaced a price gap of more than two to one between similar accounts: $7.7K against $3.7K a year.
Speed and stock outranked price.
Buyer research for a rigging hardware distributor asked what would make customers switch and what they would pay to avoid it. The answer was fast quotes, parts on the shelf, and good support, for which buyers accepted higher prices, with sensitivity that shifted by account size. The pricing segments were built from those answers.
Research that ends as a slide of charts has not finished its job. Our pricing research consultants hand over outputs that feed decisions directly:
Studies are scoped to the decision, which keeps most of them within our standard engagement length of six to 10 weeks.
Tell us the pricing decision you are facing and what you already know. We will tell you which method fits, how large the sample needs to be, and whether your own data can answer it first.
An agency runs fieldwork well. A pricing research consultant designs the study around a pricing decision, chooses the price levels and attributes so the answer is usable, and turns the results into a price structure. We also interpret stated results against your transaction data, which agencies usually do not see.
Yes, and it is often the best sample you have. Two cautions: current customers anchor on what they pay today, and a survey from you signals that prices may change. We handle the invitation wording and usually include lapsed customers and prospects to balance the view.
It is enough to set a range and rule out prices that would harm credibility. It is not enough to choose a price within that range, because it does not account for competitors or for how demand changes. We pair it with interviews or a value calculation.
By not asking them directly what they would pay. Choice-based methods such as conjoint make gaming hard because the respondent trades off features against price rather than naming a number. Interviews focus on past behavior and the cost of problems, which are harder to shade.
Yes, under your confidentiality terms, and it usually matters more than any survey. Invoice-level data shows revealed willingness to pay across thousands of real decisions, where a survey shows a few hundred hypothetical ones. Where stated answers and real invoices disagree, the study needs a second look.
Enough to change the answer. A conjoint study read on 40 responses can produce confident-looking price elasticities that are mostly noise. We state the sample behind every finding and the margin of error where it applies, and we say when a study cannot support a conclusion.
