Run a pricing experiment for an early-stage product

A pricing test should answer one decision

Do not start with “What is the perfect price?” Write a decision you can act on, such as: “For independent consultants who send at least five client reports a month, should the launch plan be $18 or $28 per month, with the same included workflow?” State what evidence would change your mind and when you will stop collecting it.

Price is only one part of the offer. A change in price, included features, trial length, billing period, or onboarding can each change behavior. If you change several together, you will not know what caused the result. Begin with the first-price guide to choose a plausible model and clear offer framework to define exactly what each buyer receives.

Start with behavior and buying context

In customer conversations, ask about the most recent time the person did this job, what they used, what it cost in money or time, who chose and paid, what approval was needed, and what happened when the problem was left alone. Ask for concrete examples, invoices, budgets, or renewal decisions where appropriate and volunteered. Separate the user, economic buyer, approver, and payer when they are different people.

Then show one specific offer and ask the person to think aloud: What do they believe is included? Which limit matters? What would prevent them from starting? Who else must approve? What would they compare it with? A hypothetical “Would you pay $X?” is weak evidence because agreement costs nothing. Record the objection and next action, not just the respondent’s stated price ceiling.

When appropriate, invite a real next step: start a disclosed trial, schedule a paid pilot, request procurement review, or purchase. Explain that the product is early, what is and is not available, the exact charge and renewal behavior, delivery date, cancellation and refund terms, and any uncertainty before someone commits. A clear no or a stalled approval is useful evidence about this offer; it does not, by itself, prove the entire market will reject it.

For interview sampling and analysis, see customer interviews and analyze a small customer-research sample. A pilot that tests delivery as well as willingness to pay should use the transparent paid-pilot workflow.

Choose a test that fits your traffic and sales cycle

Method	Use it when	What it can tell you	Main limit
Offer interviews	You are still learning the buyer, value metric, or package	What people understand, compare, and object to	Opinions are not purchases; interviewees are not a representative sample by default
Paid pilot or preorder	A real outcome can be delivered with a bounded scope	Whether a specific buyer will commit under stated terms, and what delivery costs	A founder-led sale may not predict self-serve conversion or repeatable acquisition
Sequential public offer	Traffic is low and you can honor a clear launch offer for a defined enrollment window	Whether actual conversion, objections, or deal progress changes between comparable periods	Seasonality, channel mix, and learning can differ between periods; treat the result as directional, not causal
Randomized online test	You have enough eligible, comparable traffic and can keep the rest of the offer stable	A stronger estimate of how a price change affects a defined audience and time window	Needs adequate sample size, stable assignment, clean instrumentation, and follow-up long enough to observe relevant outcomes

Published field experiments show why outcomes beyond the first click matter. In one subscription-services market experiment, price affected usage and customer attrition differently; ignoring attrition understated price sensitivity in that particular setting. A large software freemium experiment found that restricting free features increased conversion but reduced usage, and the authors cautioned that conversion and viral activity alone could overstate the economic value of a free tier. These studies are examples, not forecasts for your product: their markets, scale, and customers differ. Use them as a reminder to measure the behavior your business depends on, not to copy their prices or effect sizes (Danaher, Marketing Science; Runge et al., ESMT working paper record).

For a low-traffic indie product, do not imitate a large-company A/B test with a handful of visitors. Prefer structured interviews, a real paid pilot, and successive clearly defined offers. Keep notes on who saw each offer and through which channel. If the customer mix changed, say so. You may learn which offer deserves the next test without claiming a statistically reliable winner. There is no universal sample-size threshold: it depends on the outcome, baseline rate or variability, smallest difference worth detecting, and acceptable false-positive and false-negative risks. NIST’s sample-size guidance makes these assumptions explicit; for a consequential high-volume test, get a suitable power calculation rather than choosing a convenient round number. NIST’s experimental-design overview also starts from the objective and factors to be tested.

Write the test brief before showing prices

Copy this brief and fill it in before launch:

Decision:
Target buyer and qualifying situation:
Offer and included limits:
Price, currency, tax treatment, billing interval:
Renewal, cancellation, and refund terms:
Current alternative and buyer's cost of inaction:
Primary outcome and observation window:
Guardrails (margin, activation, refunds, support time, retention):
Offer variants (change one variable only):
Audience and acquisition channel:
Assignment or enrollment rule:
Minimum useful evidence / stopping rule:
Known confounders and unresolved questions:
Decision date and owner:

Set the price range from three constraints: the value and alternatives customers describe, the way the buyer approves and pays, and what it costs you to deliver and support the promise. Use the contribution-margin worksheet to model low, expected, and heavy-but-plausible use. A price that converts but loses money or creates unbounded manual work is not a successful test.

Write a falsifiable hypothesis. For example: “Among inbound solo consultants who currently prepare at least five reports monthly, the $28 plan will produce more contribution per qualified opportunity than the $18 plan, without reducing first-month activation or creating more than 30 minutes of support per new account.” The numerical thresholds are yours to set from your constraints; this example is not a benchmark.

Choose a primary outcome that matches the decision:

• Early discovery: qualified buyers who take the agreed next step, with reasons for refusal coded consistently.
• Checkout: completed purchases divided by eligible visitors who saw the full offer; also record checkout starts and failures.
• Paid pilot: paid commitments, delivery hours, direct cost, completion, and whether the buyer would renew or buy the next defined scope.
• Subscription: paid conversion plus activation, refund, cancellation, usage, and renewal over a period long enough to observe them.

For comparing plans, calculate contribution per qualified opportunity as well as conversion. A simple first pass is:

contribution per qualified opportunity
= total customer contribution over the observation window
  ÷ eligible opportunities shown the offer

Use contribution after variable delivery cost, payment costs, support labor you actually provide, refunds, and other customer-dependent costs. Do not confuse booked annual recurring revenue, sign-ups, or top-line revenue with collected cash, profit, or retained customers. Show cohort age and the number of eligible opportunities beside every rate.

Run the test without confusing customers

1. Freeze the offer. Write the exact audience, product state, limits, price, currency, billing cadence, renewal, cancellation, and refund terms. Change one variable between versions.
2. Choose an assignment rule in advance. For a sequential offer, set a start and end date and record channel and customer type. For a randomized test, randomly assign eligible visitors and keep each person’s version stable. Do not switch assignments after seeing early results.
3. Make the displayed terms real. Do not advertise a price you will not honor, invent a deadline, hide recurring charges, or make cancellation harder to increase conversion. Disclose material limits and the total obligation before payment. The FTC’s dark-patterns report documents deceptive patterns such as hidden costs, fake urgency, unclear subscription terms, and cancellation roadblocks in a U.S. context. Check requirements that apply to your customers and jurisdictions; the report is not a global legal standard.
4. Instrument the path. Record eligible exposure, offer version, qualified opportunity, checkout start, successful payment, activation, refund, support time, cancellation, and renewal where applicable. Avoid collecting personal data you do not need.
5. Wait for the outcome window. A signup conversion test cannot establish retention. Define how long a buyer must use the product or remain paid before evaluating recurring value.
6. Review by segment and channel only when the data supports it. Keep the raw counts visible. Tiny subgroup rates swing sharply; do not search through many slices until one looks favorable.
7. Make and record the decision. Keep the current offer, choose a variant for a defined reason, or gather a specific missing piece of evidence. Note what result would cause you to revisit it.

If you quote different prices to different people, explain the rule and make sure it is fair, consistently applied, and lawful for the market. A simple public launch price or a clearly bounded, disclosed pilot price is often easier for a small team to administer than individualized opaque pricing. Honor written quotes for their stated validity period and define how existing customers are treated before changing their price.

Read the result without overclaiming

If one price has a higher purchase rate, ask whether it also changed the customer mix, plan mix, usage, support burden, refunds, or renewal. If a lower price brings in more customers but each needs extensive setup, price alone may not be the bottleneck. If buyers accept the offer in interviews but do not take the next step, inspect trust, urgency, authority, delivery risk, and the channel before concluding the price is wrong.

When samples are small, report counts (for example, “3 of 24 qualified buyers started a pilot”) rather than presenting a percentage as a stable market rate. State the audience, dates, channel, offer, and what was measured. Do not claim an A/B-test winner unless assignment, sample size, and analysis justify that inference. A small experiment decision guide can help distinguish a useful signal from an inconclusive result.

Keep a simple decision log:

Date and audience	Offer shown	Observed behavior	Costs and follow-up	Decision / unresolved question

Revisit the decision when buyer behavior, delivery costs, retention, or the value metric changes—not merely because another company changed its pricing page.