A/B testing is an experiment that compares two versions of a page, email or ad to find out which one performs better. Traffic is split randomly between version A (the control) and version B (the variation), and the result is judged on a single metric, such as conversion rate or click-through rate.
To be reliable, a test needs enough visitors, a clear hypothesis and a fixed duration. Stopping early because one version looks ahead is one of the most common mistakes.