Data Drift

A change in the distribution of input data relative to the reference used to develop or validate a system. It can appear in vocabulary, user profiles, category frequency, or collection quality even when the relationship between input and outcome has not changed.

Detecting statistical drift signals a need for investigation; it does not by itself prove quality loss. Monitoring should connect the change to segments, task metrics, and operational effects. When the relationship between inputs and targets itself changes, the problem is commonly treated as concept drift, a useful distinction for choosing a response.