analysis · GPT-4o · free
Turn my vague product idea into a proper A/B test hypothesis. The idea: [DESCRIBE — 2 3 SENTENCES] Surface: [PAGE / FEATURE / EMAIL] Baseline metric today: [VALUE] Business goal: [WHAT I'M TRYING TO MOVE] Deliver: Testable hypothesis in "Because [insight], if we [change], then [metric] will [direction] by [amount] because [mechanism]" form Primary metric + why it's the right one (not a proxy) Guardrail metrics (2 3) to protect against downstream damage Segment I expect the effect in, and where I don't expect it (falsification test) Minimum practical effect — under which I'd still kill it Risk of interference with other running tests If the test wins, the follow up test. If it loses, what I learn. No test without a killable hypothesis.
#hypothesis #ab-test #experimentation