Model Drift
In plain terms
Degradation of model accuracy as real-world data changes.
Going deeper
Model drift occurs when the data a model sees in production diverges from its training data, eroding performance over time. The world shifts; the model staling.
Under the hood
Data drift (inputs change) and concept drift (relationships change) require monitoring and retraining.
Examples
Production monitoring, retraining triggers, fraud models, recommendation decay.
Think of it like...
Like a map slowly going out of date as roads and cities change.