Stated willingness to pay overstates achieved price in B2B, and the gap widens with the size of the decision unit — the person answering is often not the one who approves the spend.
Does willingness-to-pay research work?
Directionally, yes. It orders segments by price tolerance reliably and estimates the level badly, which is the same pattern that appears in stated switching intent and for the same underlying reason.
The techniques are not the problem. Conjoint and van Westendorp do what they claim; the difficulty is who is answering.
The level fails because a price question asked in isolation has no budget attached to it. A respondent answering what they would pay is describing a preference in a world with no competing claims on the same line, no procurement cycle and no internal case to write. The ordering survives that abstraction because relative judgments do not depend on the constraint. The absolute number does not survive it at all.
Who actually approves the spend?
In B2B, usually not the person the research reached. Users and specifiers have views on value; procurement and budget holders set ceilings, and the two produce different numbers — a mapping the sales team needs before the research does.
That makes decision-unit mapping a prerequisite for pricing research rather than a refinement of it.
Mapping the unit is a short exercise and a neglected one. For a representative recent purchase, establish who raised it, who specified it, who negotiated it, who signed it, and who could have stopped it. Five names, sometimes fewer. Everything the pricing study does afterward depends on whether it reached the fourth and fifth of them, and most fieldwork reaches the first and second because they are the easiest to recruit.
How big is the stated gap?
Large enough to misprice a launch, and it scales with how many people must approve. Where three or more sign off, the person answering the research is rarely the one who decides — and reaching the one who does runs straight into the incidence problem.
The corollary is that pricing research in a single-approver market is much more reliable than the same study in a committee market.
Correcting for the gap is possible where it has been measured rather than assumed. If a segment's stated ceiling has been compared against what that segment actually paid on a comparable purchase, the relationship between the two is stable enough to apply. Where no such comparison exists, the honest move is to report the ordering and decline to report the level, rather than deflating the number by an amount nobody can defend.
How should the question be framed?
Around an approval rather than a preference. "What would clear your approval process at this value, and who else signs" produces something closer to a price than "what would you pay" does.
It also surfaces the process, which is frequently more actionable than the number.
The framing has a cost worth naming: it slows the interview and it will not survive a self-completion survey. An approval question needs follow-ups — what the threshold is, who else sits in the chain, what happened the last time something at that value went through — and that is an interview rather than a form. Teams needing both usually run the interviews first and use them to write the closed questions the survey then asks at scale.
What beats asking about price?
Asking buyers what they currently spend on the alternative, including the internal cost of doing nothing. Buyers describe existing spend accurately and predict future spend poorly.
Existing spend also anchors the conversation in something checkable, which is a useful discipline for both sides.
Existing spend also gives you a check the respondent cannot easily game. Where someone describes what they pay today and the figure sits well outside what the rest of the segment reports, the divergence is almost always definitional — a bundled service, an internal cost counted or not counted — and resolving it teaches you more about how the category is actually bought than the price answer would have.