Concept drift is the gradual or sudden change in the relationship between what happens on calls (language, intent, outcomes) and how your analytics, models, or QA rules interpret it. When drift occurs, the same words or behaviors can start meaning something different, or new patterns appear that weren’t present in the training data.
Operationally, drift matters because it can quietly degrade accuracy in things like intent detection, sentiment, compliance flags, and auto-dispositioning. That leads to missed risks, noisy alerts, incorrect reporting, and coaching based on outdated signals.
Common causes include new products or promotions, policy updates, seasonality, changes in customer mix, and shifts in agent scripts or workflows. Monitoring for drift and refreshing labels, rules, and models helps keep dashboards and QA findings aligned with what’s actually happening on calls.
