In practice, if we were going to talk about that -- variations which perform statistically significantly better one day may perform worse the next. For example,

In practice, if we were going to talk about that -- variations which perform statistically significantly better one day may perform worse the next. For example, because happy people like to click on one variation versus another which is preferred by stressed people. Keep A-B tests running even after you've decided the best variation to make sure that the uplift you've observed is real! See http://john.freml.in/ab-testing-significance http://john.freml.in/ab-testing-significance