A/B test design
Designs an A/B test — hypothesis, variants, primary metric, MDE, sample size, duration.
Stops "let's just see if it converts" experiments that yield nothing.
The prompt
Copy this into ChatGPT, Claude, Gemini or any assistant you already use — then paste your own change being tested + product context underneath it.
Design an A/B test. Output:
**Hypothesis** ("If we X, then Y, because Z")
**Variants** (control + 1-3 treatments)
**Primary metric** (single, with definition)
**Guardrail metrics** (2-3 that shouldn't regress)
**Minimum detectable effect** (target lift %)
**Sample size** (rough estimate, with assumptions)
**Duration**
**Stop criteria** (early-stop rules)
**Analysis plan** (how we'll cut the data)
Rules:
- One primary metric. Don't change it mid-test.
- State assumptions out loud.Or run it here
Clicking copies the prompt to your clipboard and opens the chatbot in a new tab. Gemini doesn't accept URL params — paste manually with Ctrl/Cmd+V.
Related prompts
Breaks a top-line metric (e.g. ARR) into a tree of contributing inputs the team can act on.
Takes a vague stakeholder question and turns it into a precise, answerable analytical query.
Reads cohort data + surfaces retention shape, decay rate, and "is this curve flat enough?".
Given an unexpected metric change, lists 5 most-likely causes + the order to investigate.
Designs a 6-tile dashboard layout for a stakeholder, with each tile's metric + viz type.
Analyses a conversion funnel — identifies the leakiest step + 3 hypotheses for why.
Browse the full free AI prompt library or see more data & analytics prompts.