Variant

A variant is one of the competing versions in an experiment; each variant shows a different experience to a share of users so their performance can be compared.

In the simplest experiment there are two variants: the current version (which is the control) and the changed version. But the number of variants can be higher, and here lies a mathematical trap: the more variants, the more the traffic is split among them, and the slower each one reaches enough volume for a meaningful result. A variant should not be confused with a feature; a variant is a temporary state for measurement, not a permanent capability. The healthy rule is to keep the number of variants small so the experiment reaches an answer in reasonable time.