Run Ab Experiments · shahriarfarzadi bundle Use this skill when a user wants to plan, review, launch, manage, intervene in, ramp, monitor, analyze, or interpret an online controlled experiment or A/B test; choose hypotheses, units, variants, exposure, metrics/OEC, guardrails, sample size, duration, stopping rules, SRM checks, or ship/iterate/stop decisions. Manage the full lifecycle in three modes—plan, manage, and interpret—with explicit cognitive-load reviews and concise, non-repetitive outputs grounded in Kohavi, Tang, and Xu. Always begin with a mandatory question-first interview and context-and-assumptions confirmation before work. Do not use it for observational-only causal claims unless deciding whether an experiment is feasible.
