Knowledge MaxxingConcept drift
Concept drift is when the real-world relationship between input features and the correct outcome changes over time, making a previously accurate model gradually less reliable.
Prerequisite: Online inference
What matters
Because the model itself does not change, drift is invisible unless performance is actively monitored after deployment; a model can silently degrade without any code changing.
Example
A spending-fraud model trained before a major shift in shopping habits may start missing new fraud patterns that did not exist when it was trained.
Common misconception
Concept drift is not the same as data drift; concept drift is a change in the relationship between features and outcome, not just a change in the input data's distribution.
Recall question
Why can a model silently become less accurate over time with no code changes at all?