seniorSupervised Learning
What is model drift monitoring in production machine learning systems?
Updated May 17, 2026
Short answer
Model drift monitoring tracks performance degradation of ML models in production over time.
Deep explanation
In production, models degrade due to changes in data distribution, user behavior, or external environment. Monitoring involves tracking input data distribution (data drift), prediction distribution, and ground truth performance metrics. Tools detect drift using statistical tests like PSI (Population Stability Index) or KL divergence.
Real-world example
E-commerce recommendation systems losing accuracy due to changing user preferences.
Common mistakes
- Only monitoring accuracy without tracking input distribution changes.
Follow-up questions
- What is PSI (Population Stability Index)?
- How often should drift be monitored?