Deep diveBusiness Building
Willingness-to-pay research methods for pricing a new B2B venture
A new B2B venture usually has to set a price before it has the sales history to justify one, and asking customers what they would pay produces numbers nobody should trust. This deep dive covers the methods that work better before revenue, what each can and cannot tell you in a small market, how to choose a pricing metric for AI products whose costs grow with use, and how to turn a pilot price into a list price.
On this page
- Why pricing belongs in venture validation, not after launch
- The trouble with asking customers what they would pay
- Willingness-to-pay methods and terms
- Comparing willingness-to-pay methods before revenue
- From the customer's alternative to a defensible price range
- Small samples and buying committees in B2B pricing research
- Choosing a pricing metric for an AI product
- Hypothetical example: an AI compliance tool for mid-size insurers
- Turning a pilot price into a list price without the free-pilot trap
- Questions and answers
- Sources
Why pricing belongs in venture validation, not after launch
A venture can prove that customers want its product and still find it has no business, because the price they will pay does not cover the cost of serving them. Price also shapes who buys: the same product priced for enterprises and for small teams finds different customers, channels and sales cycles. Pricing evidence therefore belongs in validation, alongside evidence about the problem and the solution.
In ColdAI's studio method, validation sprints test core assumptions before significant resources are committed, and product-market fit work iterates on how customers respond1. Willingness to pay is one of those core assumptions. This page covers price only; validation sprints cover whether the problem and solution are real.
The trouble with asking customers what they would pay
“What would you pay for this?” fails for predictable reasons. The answer is hypothetical, so it costs the respondent nothing. It anchors on whatever they pay for something similar, even when the new product does far more. B2B buyers also answer strategically: anyone who expects to negotiate with you later has a reason to name a low figure. And the person answering is often not the one who controls the budget.
None of this makes customers' views useless. It means stated prices are a starting point to be tested, and the strongest evidence is a commitment that costs the customer something.
Willingness-to-pay methods and terms
- Van Westendorp price sensitivity meter
- A survey asking at what price the product would be too expensive to consider, expensive but still worth considering, a bargain, and so cheap that quality is in doubt. Plotting the answers shows a range of acceptable prices.
- Gabor-Granger
- A survey that shows respondents a sequence of prices and asks whether they would buy at each, producing a rough demand curve and a revenue-maximizing price for one defined offer.
- Choice-based conjoint
- A survey in which respondents repeatedly choose between bundles that vary in features and price, so the value of each feature and the sensitivity to price can be estimated statistically.
- Economic value to the customer
- The cost of the customer's best alternative, plus the value of the new product's advantages, minus the cost of its disadvantages. It sets an upper reference point for price.
- Pricing metric
- The unit the customer pays for: a seat, a transaction, a volume of usage or an outcome. Also called a value metric.
- Priced letter of intent
- A written, usually non-binding statement that a buyer intends to purchase at a stated price if agreed conditions are met. Weaker than a contract, much stronger than an interview.
Comparing willingness-to-pay methods before revenue
| Method | Decision it supports | Sample it needs | Main bias risk |
|---|---|---|---|
| Direct question | Rough orientation only | Any | Hypothetical, anchored and strategic answers |
| Van Westendorp | The acceptable price range for a defined offer | Dozens of respondents who understand the product | No purchase intent; anchored on current alternatives |
| Gabor-Granger | Demand at specific price points for one offer | More respondents than Van Westendorp for a stable curve | Overstates purchase intent and ignores competing offers |
| Choice-based conjoint | Trade-offs between features, packages and price | Usually more respondents than a niche B2B market contains | Complex to design; respondents simplify hard choices |
| Value-based interviews | The economic value behind a price, and how to frame it | A few deep interviews per buyer role | Customers may not know, or share, their real costs |
| Behavioral tests | Whether buyers will pay a specific price now | As many real prospects as the venture can reach | Small numbers; early adopters pay more than the mainstream |
Behavioral tests include pre-sales, priced letters of intent, paid pilots and price tests on a landing page. Pair one survey or interview method with at least one behavioral test.
From the customer's alternative to a defensible price range
- Best alternative's cost
What the customer spends today on the closest substitute, including staff time.
- Add differentiation value
Savings or gains the new product delivers beyond that alternative.
- Subtract switching costs
Integration, retraining, risk and anything the product does worse.
- Economic value to customer
The ceiling a rational buyer would pay.
- Choose the value share
Leave customers enough of the value that switching is clearly worth it.
- Test the range
Confirm with surveys, interviews and priced commitments.
Small samples and buying committees in B2B pricing research
Many B2B ventures sell into markets with a limited number of potential buyers, which rules out the statistical precision consumer pricing studies expect. It does not rule out good decisions. Deep value interviews with each buyer role, a Van Westendorp survey across the people you can reach and a few priced commitments can bound the answer well enough to set a first price.
Buying committees complicate things. The user judges usefulness, the budget holder judges value against other spending, procurement compares price with benchmarks and security or compliance can veto the purchase outright. Interview each role separately, and expect procurement to anchor on the cheapest comparable product even when it does far less.
Choosing a pricing metric for an AI product
Each use of an AI product carries an inference cost, so the metric has to track the value customers get and keep the margin positive as usage grows. AI total cost of ownership covers the cost side.
- If
Value grows with the number of people using the product, and usage per person is predictable
ThenPrice per seat, with fair-use limits on the most expensive operations.
Seats are easy to budget for, but unlimited heavy users can erase the margin.
- If
Value and inference cost both grow with the volume processed, such as documents, calls or transactions
ThenPrice by usage, with a committed minimum or volume bands.
The metric tracks both value and cost, and the commitment gives each side predictability.
- If
The outcome is measurable, attributable to the product and agreed by both sides
ThenConsider an outcome-based element, usually on top of a base fee.
It aligns price with value, but attribution disputes and delayed revenue make it risky on its own.
- If
Buyers need a fixed annual figure for procurement
ThenOffer tiered platform fees mapped to usage bands.
It fits procurement while keeping a link to cost.
Hypothetical example: an AI compliance tool for mid-size insurers
Turning a pilot price into a list price without the free-pilot trap
Free pilots teach customers that the product is free and teach the venture nothing about price. Charge something for every pilot, and write the conversion price and success criteria into the pilot agreement before it starts. Crediting the pilot fee against the first year's contract keeps the commitment without making it feel like a penalty.
Publish a list price early, even if most early deals are discounted, and record every discount as a discount from that list. Give design-partner terms an end date. Otherwise the first customers' prices become the market's reference point, and every later negotiation starts from them.
Questions and answers
Is the Van Westendorp method valid for B2B products?
It is useful for finding the range of prices respondents consider acceptable, provided they understand the product well enough to judge it and their opinion affects the purchase. It is weaker in B2B than in consumer markets because samples are small and users, budget holders and procurement answer differently. Use it to bound the range, then test specific prices with commitments.
How many respondents does willingness-to-pay research need?
It depends on the method. Value-based interviews can be informative with a few conversations per buyer role, because they aim to understand costs rather than measure averages. Van Westendorp and Gabor-Granger need enough respondents for the curves to settle. Choice-based conjoint needs more than most niche B2B markets can supply, so new B2B ventures usually rely on interviews, simple surveys and priced commitments.
Should a new venture launch with a low price to win its first customers?
Usually not as a list price. Low launch prices are hard to raise, attract customers who buy on price and set the reference point for every later deal. If early customers need an incentive, offer a time-limited discount from a published list price or extra service, and record it as such. Charging for pilots also turns the first revenue into pricing evidence.
When does outcome-based pricing make sense for an AI product?
When the outcome can be measured cleanly, can be attributed to the product rather than other factors and is agreed by both sides before the contract starts, such as documents processed to an agreed accuracy. Even then, most ventures combine an outcome element with a base fee, because purely outcome-based revenue arrives late and invites disputes over attribution.