experiment-readout

★ 7.2 · data

experiment-readout is a Claude Code skill that turns A/B test and product experiment data into a structured readout with a clear decision: ship, iterate, extend, or stop. The skill produces nine required sections: a header with a prominent decision badge and primary metric delta, a verifiable hypothesis, experiment setup (audience, variants, duration, sample size, primary and guardrail metrics), a result snapshot with lift and absolute delta, a control-vs-variant metric table, an interpretation that separates signal from noise and unknown, a justified decision, 2–4 follow-up experiment proposals with hypothesis and effort estimate, and instrumentation notes covering data gaps and sample bias. Three visual styles are available — product-readout, lab-notebook, and growth-console — selected automatically based on the material. No data is fabricated: p-values and sample sizes come exclusively from user input, and the skill explicitly flags inconclusive results as "directional" or "needs more data". Built for PMs, growth teams, and data analysts who need raw experiment results turned into an actionable product decision.