Definition When a model gets worse over time because the world it was trained on has changed.
Example A model trained on last year's purchasing patterns begins performing worse as customer behaviour changes.
See how it works Record baseline performance Monitor inputs and outcomes Investigate performance changes Review, update or retrain as justified Check your understanding Common mistake: Every performance decline is not necessarily model drift; data defects or system changes may also be responsible.
What should be monitored besides model predictions?
Show explanation Input distributions, relevant outcomes, quality of incoming data and performance against an agreed baseline.
For your role Your role Choose your role Accounts payable Credit controllers FP&A analysts Finance business partners Finance directors and CFOs Financial controllers Management accountants Payroll specialists Account executives Sales managers Sales operations analysts Sales representatives Bid and tender managers Commercial directors New to business Pricing analysts Revenue managers Senior leaders Campaign managers Communications managers Customer insight analysts Digital marketing managers Marketing analysts Marketing directors HR business partners HR directors Learning and development managers People analysts People managers Recruiters Operations analysts Operations directors Operations managers AI product managers Heads of product Product analysts Product managers Product owners Machine learning engineers Analytics engineers Data analysts Data architects Data engineers Data governance leads Data privacy officers Data scientists Information architects Platform engineers Report builders Business analysts Corporate development managers Heads of strategy Strategy analysts
Choose your role to see what this term means in your work.
Why it matters Where it is used Any model left running for months without being checked.
