Model Drift Glossary Definition

Model drift is a decline or change in model performance when real-world data, behavior, conditions, or relationships shift after deployment.

Monitoring for drift helps teams decide when to retrain, recalibrate, investigate data changes, or adjust deployment controls.

MLOps teams monitor model drift to decide when to retrain, recalibrate, investigate data issues, or adjust model governance controls.

Additional Resource: Authoritative source.

Related concepts include MLOps.