Model Drift

A model's real-world accuracy quietly degrading over time as the data it sees diverges from what it was trained or tuned on.

A model that performed well at launch can degrade months later — not because the model changed, but because the real-world inputs it's handling have shifted (new product categories, new customer language, a changed process) away from what it was originally validated against. Without ongoing evaluation, this degradation is silent until someone notices the outputs have gotten worse.