When to pause a growth experiment
Experiments earn their keep when they change a decision. If a test has already shown a clear direction for onboarding length or trial pricing, continuing it only to chase statistical elegance can delay shipping the better path.
Pause when measurement itself is broken. Missing events, duplicated user IDs, or a release that mixed cohorts will not become cleaner with more time. Fix tracking, then restart with a clean date stamp.
Pause when the audience shifted underneath you. A holiday campaign, a viral mention, or a store feature can flood an experiment with people who are not your usual visitors. Note the external event and decide whether the result still applies.
Write the stop decision down. Include what you learned, what you will ship, and what remains unknown. That short record is often more valuable than another week of inconclusive charts.