What is model drift?
Model drift is the gradual loss of accuracy in a predictive model because the data or relationships it learned from no longer match the real world. It happens when customer behavior, markets, or tracking change.
Last reviewed .
What causes model drift
Data drift happens when the inputs change, for example when a new channel brings a different kind of visitor. Concept drift happens when the link between inputs and outcomes changes, for example when buyers respond differently to the same offer.
Tracking changes can cause drift as well. If a tag breaks or a consent rule shifts what you collect, the model sees different data than it was trained on.
How to catch and manage it
Compare predictions with actual outcomes on a regular schedule, and watch for shifts in the inputs. A steady decline in accuracy is the clearest sign.
The usual fix is model retraining on recent data. Set an alert threshold so you know when to retrain and don't discover the problem after a bad decision.
Common questions
What is the difference between data drift and concept drift?
Data drift is a change in the inputs a model sees. Concept drift is a change in how those inputs relate to the outcome. Either can reduce accuracy.
How do you detect model drift?
Track the model's accuracy against actual results over time, and monitor whether input data looks different from the training data. Automated monitoring can flag large changes.
How often does model drift happen?
It depends on how fast your market and data change. Fast-moving or seasonal businesses tend to see it sooner.
Does model drift affect ad platform bidding?
It can, if the signals feeding the platform's models change, for example after a tracking problem. Keeping conversion data accurate helps those systems stay well calibrated.
Want help putting this to work?
Tell us where your marketing stands today, and we'll show you what to do first.