A test should teach which assumptions hold. It cannot isolate the effect of price if the product, audience and distribution all change at the same time without a way to separate them.
State the decision in plain language
A fictional maintenance service is considering a move from $90 to $100 per monthly customer. The variable service cost is $30 per customer. With 200 customers, current monthly contribution is 200 × ($90 − $30) = $12,000.
At the proposed price and unchanged unit cost, preserving that total requires $12,000 / $70 = 171.428… customers, or at least 172 whole customers. This defines a volume boundary; it does not forecast retention or justify applying the change to any real customer.
Choose who is exposed
A new-customer price test answers a different question from an existing-customer renewal change. Existing users have experience, commitments and switching work. New prospects compare alternatives before joining. Do not combine their results and call the average “price acceptance.”
Where appropriate, compare similar groups seeing comparable offers. If random allocation is impractical, use a clearly described rollout and record differences between the groups. Seasonality, channel changes and customer mix can otherwise explain the apparent result.
Honor existing terms and communicate what changes clearly. The research objective does not remove contractual commitments or the importance of customer trust.
Measure the full bridge
| Measure | Why it belongs in the test |
|---|---|
| Qualified inquiries and purchases | Reveals how new demand responds |
| Renewals, cancellations and downgrades | Shows what happens to the starting base |
| Contribution after variable service cost | Connects price and demand to economics |
| Refunds, support effort and complaints | Catches hidden cost and experience changes |
| Later retention | Tests whether an early acceptance persists |
Suppose 180 fictional customers remain at $100. At a $30 variable cost, contribution is $12,600, or $600 above baseline. If those remaining customers instead cost $35 each to serve, contribution is $11,700. The same customer count can support opposite economic conclusions.
Set a window and a stopping condition
Choose a period that can observe the decision you care about. A monthly service may need renewal data; an annual contract may show effects much later. Set a limit on undesirable outcomes and name who reviews them.
A small sample can reveal implementation trouble or a severe response, but often cannot reliably distinguish modest effects. Avoid adjusting the observation window repeatedly until the result looks favorable. Keep the uncertainty visible and use the business-experiment guide to document it.
Interpret what happened
A contribution gain in this group does not establish the optimal price for every segment. A lost customer may reflect a weak offer, a credible substitute or a service failure. Ask what changed, what buyers compared and what result they expected.
Use the customer-promise guide to make the offer explicit, the price calculator to compare cost scenarios, and cohort retention to inspect the timing of departures.
Sources and limits
The underlying contribution distinction follows OpenStax. The service, figures and test design are original fictional HBW analysis. This guide gives no universal sample size, acceptable churn rate or optimal price.