The different versions are made to compete against each other by being submitted simultaneously and randomly to different visitors. This makes it possible to erase context-linked effects (seasonality, sales promotion…). A/B testing should be distinguished from multivariate testing (MVT), where several elements are made to vary across the different versions; these two kinds of tests are part of a discipline called web testing.
An A/B test is a scientific method which, when applied to marketing, consists in displaying several variants of a same web page to different users. These variants are called versions. The aim of such a test is to compare the performance of the different versions based on a set of predefined KPIs.