Conversion

When to stop an A/B test

Write the rule before you launch: minimum sample, primary metric, what “win” means. If you did not write it down, you will stop when B looks nicer.

Peeking every morning and calling the test at the first green day inflates false winners. Lucky weeks become process.

If traffic is thin, you do not have a test. You have a sequence of anecdotes. Either wait, or change one thing on the page without pretending it was statistically decided.

Use the A/B calculator against the numbers you actually have - visitors and conversions per variant - not a screenshot of a “confidence” badge you cannot define.