Signal decay is when a conversation signal (like a keyword rule, intent model, sentiment score, or compliance detector) becomes less reliable because the underlying conversations and operations have changed. New product names, updated policies, shifting customer phrasing, channel mix changes, and agent scripting updates can all cause the signal to drift away from what it was designed to measure.
Operationally, signal decay matters because it can quietly distort dashboards, alerts, and QA sampling. Teams may miss emerging issues, overreact to false spikes, or make coaching and staffing decisions based on stale indicators. Managing signal decay typically means monitoring signal performance over time, validating against human review, and retraining or updating rules when drift is detected.
