Data drift is when the data a deployed model sees changes from what it was trained on; concept drift is when the link between inputs and right answers changes.
In plain words
A model learns patterns from training data, but the world keeps moving. If incoming data starts to look different, such as new customer types or new slang, that is data drift. If the correct answer for the same input changes, that is concept drift. Either way, predictions quietly get worse. Teams watch for drift by comparing live data with a baseline and retrain when it shifts.
A quick example
Google's glossary gives a car example: a model labels car models as fuel efficient or not. But the idea of a fuel-efficient car keeps changing over the years, so the old model's labels grow less and less useful.
Where you'll meet it
- Monitoring models in production.
- Deciding when to retrain.
Related
Sources
- DocsMachine Learning Glossary, Google for Developers · read 29 Sept 2026
- DocsIntroduction to Vertex AI Model Monitoring, Google Cloud docs · read 29 Sept 2026
- DocsData and model quality monitoring with Amazon SageMaker Model Monitor, AWS docs · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.