When application analytics and on-call alerts disagree

During Monitoring Health Checks we often find a split: application analytics charts show rising error rates on an endpoint, while on-call stays quiet — or the pager fires on deploy noise while users report nothing.

Common causes

• Thresholds set against averages instead of percentiles for that endpoint
• Alerts scoped to infrastructure hosts rather than the application path users hit
• Analytics labels that changed after a release while alert queries did not

A reconciliation pass

List the endpoints that matter to revenue or compliance. For each, note the analytic signal, the alert condition, and the last time a human investigated. Gaps in that table are the health check’s real output.

Endpoint performance monitoring is not only collection — it is whether the right people hear about the right failure while there is still time to act.