Stopping Rules and Significance
Define your stopping rule before launch, not after peeking at early lifts. Use power analysis to estimate required sample size, commit to a significance threshold, and monitor guardrail metrics. Seasonality, promotions, and outages can distort results; annotate timelines, revalidate after rollout, and watch for regression once novelty and curiosity decay.