WhyLabA/B causal impactFeature impact

Experiment readiness checklist

Before launching an experiment, confirm that assignment, logging, power, and decision rules are strong enough to support a business decision.

Design

  • Define primary and guardrail metrics.
  • Pre-register segments and minimum detectable effect.
  • Set stop rules before launch.

Instrumentation

  • Verify assignment events and exposure events separately.
  • Confirm user identity stitching across devices.
  • Check bot, employee, and QA traffic exclusions.

Decision

  • Agree on rollout, rollback, and inconclusive outcomes.
  • Track novelty effects and post-test monitoring windows.
  • Archive assumptions for future causal reviews.

Use this page as the intake checklist for growth, activation, retention, or monetization experiments.

Review causal assumptions