Business
Pricing A/B Test Plan AI Prompt
Plan pricing A/B tests with this AI prompt. Test price points scientifically without angering customers. Includes variables, examples, and pro tips.
The prompt
ROLE: You are an experimentation lead who runs pricing tests at scale. CONTEXT: Product: [PRODUCT]. Current price: [CURRENT_PRICE]. Hypothesis: [HYPOTHESIS]. Traffic: [TRAFFIC]. Sales cycle: [SALES_CYCLE]. Constraints: [CONSTRAINTS_TEST]. TASK: Design a rigorous pricing A/B test. FORMAT: 1. Test Design: variants (control + 2 challengers max), randomization unit, sample size and duration calculation logic, primary metric (revenue per visitor) and guardrail metrics (conversion, churn, support tickets). 2. Price Presentation: how each variant is shown (keep everything except price identical). 3. Risk Controls: traffic cap, kill criteria, customer communication if someone sees different prices. 4. Analysis Plan: how to call a winner (statistical and business significance). 5. Rollout Plan for the winner. End with 3 pricing tests to run after this one. CONSTRAINTS: Never test different prices on existing customers without disclosure. Minimum 2-week duration to cover weekly cycles. Require both statistical significance and meaningful business impact before declaring a winner. If any variable is unclear or missing, ask up to 3 targeted clarifying questions before producing the output, and clearly mark every assumption you make with [ASSUMPTION].
Variables to fill in
Replace each [variable] in the prompt with your own details:
[PRODUCT] [CURRENT_PRICE] [HYPOTHESIS] [TRAFFIC] [SALES_CYCLE] [CONSTRAINTS_TEST]
Example input
PRODUCT=language learning app, CURRENT_PRICE=$9.99/month, HYPOTHESIS=$14.99 increases revenue per visitor without hurting trial starts, TRAFFIC=50k visitors/month, SALES_CYCLE=7-day trial, CONSTRAINTS_TEST=cannot show different prices to logged-in users
How to use this prompt
Define the hypothesis sharply, set kill criteria before launch, and resist peeking at results daily; let the test run its course.
Tips for the best results
- Test price presentation (annual default, charm pricing) before testing price levels; framing moves numbers cheaper.
- One variable at a time; testing price and packaging together tells you nothing.
- Document every test; a pricing test library compounds into institutional knowledge.
Frequently asked questions
Is it ethical to show different prices to different visitors?
Standard practice in e-commerce if done transparently at scale. Never vary prices for the same logged-in customer, and honor advertised prices.
How long should a pricing test run?
At least two full business cycles (usually 2-4 weeks) with enough traffic for statistical power on revenue per visitor.