Alerting

Problem 

Your API error rate jumps from:
0.1% → 18%
Metrics contain the information. But nobody is staring at dashboards at 03:00.

Solution — Alerting

Define conditions that automatically notify operators. 
Error rate > 5%
       ↓
Alert
       ↓
On-call engineer
Core idea: alerting turns important telemetry conditions into human attention.