Sample size per group by the lift you want to detect
How often a test with no real effect declares a winner
Each extra look is another chance for noise to cross the line. The fix is to decide the sample size first, or use a sequential method built for repeated looks.