The use of a randomized experiment with two variants (termed A and B) which differ in a single particular.