A/B Test Duration Calculator: How Long to Run Your Test

Tell it your daily traffic, your baseline conversion rate, and the lift you want to detect, and it works out how many days the test must run to reach a valid result. Stopping early is tempting and almost always wrong; a duration decided up front is the simplest guard against calling a winner that is not one.

Run for whole weeks, not part-weeks. Weekday and weekend behaviour differ, so a test that ends mid-week can over-weight whichever days it caught. Even when the maths says four days, take it to a full seven, and never to fewer than one complete business cycle.