A statistical method in which two variants of the same activity are compared against each other (typically with several variants), one at a time, and the most effective variant is selected.