¿Qué es A/B test, randomisation unit, interference, sample ratio mismatch?

An A/B test compares two versions of a component by assigning units of live flow to them at random and comparing an outcome. The randomisation unit is what is assigned: an order, a stock-day, a day, a client. Interference occurs when the outcome of one unit depends on the assignment of others, so that the difference between the arms of a test differs from the effect of switching every unit. A sample ratio mismatch (SRM) is an observed split between the arms that differs, beyond chance, from the planned one (Fabijan and co-authors).