Pitching
Pricing Experiments: Design a Test That Answers One Question
Test one pricing decision with a clear customer, comparison and outcome. Work through a fictional contribution example before choosing your next move.
What you will learn
- Write a testable pricing hypothesis.
- Compare contribution per eligible prospect, not conversion alone.
- Separate an exploratory result from a rollout decision.
A pricing experiment changes one clearly defined part of an offer and measures the result against a decision you chose in advance. Start with the customer and value delivered, then test a price, package or billing unit. Changing all three at once makes the result difficult to interpret.
Choose the question before the price
“Will customers pay more?” is too vague. A useful question is: for new five-person design agencies that need monthly reporting, does a $120 offer produce more contribution per qualified prospect than a $100 offer, without increasing early cancellations? That names the buyer, comparison, outcome and guardrail.
Use YC’s pricing lesson for founder context. The test card below is an original planning exercise, not a claim about a universally correct startup price.
Build a pricing test card
An original, fictional pricing-test brief
| Field | Decision to record |
|---|---|
| Customer | New agencies with five staff and a recurring reporting job |
| Single change | Monthly price: $100 versus $120; same features, onboarding and contract |
| Assignment | Random assignment of eligible new prospects; retain assignment in analysis |
| Primary outcome | Contribution per eligible prospect after the same 30-day observation window |
| Guardrails | Cancellation, refunds, support load and complaints |
| Stopping rule | Predetermine duration and sample needs; do not stop when one arm briefly leads |
| Decision owner | Founder reviews results with product and finance; existing contracts unchanged |
Work the economics, not just the conversion rate
In this fictional first cohort, each arm receives 100 eligible prospects. Variable service cost is assumed to be $30 for each customer during the measured month. Taxes, acquisition spending and fixed costs are excluded. No refunds have matured yet, so this is a preliminary comparison.
Fictional one-month test, before refund and retention readout
| Measure | Offer A | Offer B |
|---|---|---|
| Monthly price | $100 | $120 |
| Eligible prospects | 100 | 100 |
| Customers who pay | 20 | 18 |
| Observed conversion | 20% | 18% |
| Revenue | $2,000 | $2,160 |
| Assumed variable service cost | $600 | $540 |
| Contribution | $1,400 | $1,620 |
| Contribution per eligible prospect | $14.00 | $16.20 |
Offer B converts fewer prospects but produces $2.20 more contribution per eligible prospect in this sample. That is a useful observation, not proof that B is superior for the wider market. Refunds, retention, support and sampling uncertainty can change the conclusion.
Keep the comparison interpretable
- Measure from assignment, rather than removing prospects who disliked the price.
- Run comparable groups at the same time when possible; traffic and seasonality can otherwise confound the result.
- Do not mix existing renewals with first purchases without a separate analysis.
- A sales-negotiated price test needs a record of exceptions and who actually saw which offer.
- Plan sample size around a meaningful difference and available traffic; a handful of purchases is directional learning.
Practice: change one assumption
Suppose offer B adds $15 of extra onboarding cost per customer. Its contribution becomes $1,350, or $13.50 per eligible prospect. A $0.50 disadvantage replaces the apparent $2.20 advantage. The price alone did not explain profitability; delivery cost mattered.
Your worksheet should retain the original hypothesis, assignment method, numerator, denominator, observation window, exceptions and unresolved risks. Do not rewrite the hypothesis after seeing the winner.
Turn the result into a next move
Keep the test evidence separate from your company’s business-model explanation. Bring a clear result, its limitation and the next test into investor conversations. In Finta, keep that evidence with the relevant fundraising materials and record who needs an update. The experiment should sharpen the conversation, not create another unsupported slide.