Glossary · Methodology

Data Drift

A change in the statistical properties of input data over time relative to what a model was trained or evaluated on, which can silently degrade a deployed model's real-world accuracy.