Sampling, calibration and coaching cadence — the three mechanics that keep quality from drifting.
Quality does not degrade suddenly; it drifts. Without sampling, drift is invisible until a customer escalates.
Sample deliberately: a fixed number of interactions per agent per week, scored against a rubric that reflects the SOP. Random sampling beats cherry-picked reviews every time.
Calibrate the scorers. Once a week, several reviewers score the same interaction independently and discuss variance. Without calibration, scores measure the reviewer, not the agent.
Then coach on patterns rather than incidents. One low score is noise; three in the same rubric category is a training gap worth a session.