Willingness to pay

A pricing team can obtain a confident answer to the wrong question. Asking users what a feature is worth may miss the budget holder's alternative. Studying only current subscribers misses customers deterred by onboarding, payment access or the existing offer. Define who can buy, what they would receive and what declining would mean before selecting a survey or experiment.

Brzozowicz and Krawczyk compared hypothetical valuations with consequential bids and varied price anchors. Their experiments found higher hypothetical valuations and stronger anchoring in hypothetical settings. Their incentive mechanism gave a bid a potential spending consequence; it was not an ordinary shop's posted-price conversion test. Some regression interaction tests did not establish an interaction even where other analyses supported their overall interpretation. That makes the paper evidence about elicitation design, not a universal correction factor. [Brzozowicz and Krawczyk, methods, results and conclusions](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0262130)

Aoki and Akai supply a useful counterexample. In their Japanese mandarin-orange study, participants chose among offers differing in price and carbon emissions. A consequential laboratory treatment involved real oranges and an endowment; other treatments used imagined spending in the lab or online, with an additional online reminder about hypothetical bias. The comparison concerned the marginal value of a carbon-emission reduction, not the total maximum price of an orange. [Aoki and Akai, Sections 2.1–2.5](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0261369)

The paper reports estimated marginal willingness to pay of 0.53, 0.52, 0.54 and 0.58 JPY per gram of carbon-emission reduction for the consequential lab, hypothetical lab, hypothetical online and online-with-reminder treatments respectively. Its pairwise tests did not detect hypothetical bias. This challenges the rule that every survey necessarily inflates valuation; failure to detect a difference does not establish that the methods are equivalent. [Aoki and Akai, Section 4.2 and Figure 4](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0261369)

This is not one contemporaneous randomized market trial. The lab data were collected in 2012 and online data in 2016; recruitment and participant composition differed. The authors explicitly discuss selection and estimation limits. For a manager, the useful inference is to test whether the particular incentive, product attribute and sample alter the answer. The numerical proximity of these estimates cannot justify importing the orange valuation into another category. [Aoki and Akai, Sections 2.1, 4.1 and 5.4](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0261369)

Consequential research can still be unrepresentative. An experimental endowment, unfamiliar product, repeated questioning or a narrow sample may change attention and spending. Incentives address one source of bias; they do not recreate every commercial constraint. Preserve the experiment's choice conditions and compare them with the intended market.

Observed conversion also has competing explanations. A promotion may change traffic quality, a price change may coincide with product improvements and a renewal can occur under switching costs. Treating their aggregate differences as a price effect would conflate exposure, offer and preference. A credible comparison holds the important conditions stable or reports their remaining confounding.

A marginal attribute estimate is not a total price ceiling. Adding feature estimates mechanically can ignore interactions, substitution or the buyer's budget. Likewise, the highest price named by a few respondents is not the price that maximizes contribution. The offer architecture needs demand across feasible prices, delivery costs and likely movement between tiers.

For an unnumbered hypothetical team purchase, ask separately what the employee values, what the budget holder can authorize and which alternative the organization would choose. If the feature is useful but purchasing is blocked, a lower price might leave the decisive constraint untouched. That is a research question to test, not a claim about a named company's customers.