Skip to content
Latchkey

Model Monitoring - CI/CD Glossary Definition

Model monitoring tracks a live models inputs, predictions, and outcomes to catch drift and performance decay.

Model monitoring continuously tracks a deployed models predictions, inputs, and outcomes to detect performance decay, drift, and data issues. Alerts trigger retraining or rollback.

A model can pass every deploy check and still degrade as the world changes; monitoring is how you notice before users do.

What it watches

Monitoring compares live input distributions to training data (data drift), tracks prediction quality over time (model drift), and measures accuracy once ground truth arrives.

Related guides

Run this faster and cheaper on Latchkey managed runners. Start free →