Symptoms, not causes

Alert on high error rates, slow responses and failed checkouts. CPU at 90% is not a problem if nobody is waiting; a checkout failing at 3% is a problem even at low CPU.

The three questions any dashboard should answer

Every monitoring setup should immediately answer:

  • Is the site up right now
  • Is it slower than usual
  • Did anything change recently

Test the alerting path

Trigger a real alert quarterly. Alerting configurations decay: people leave, phone numbers change, and integrations expire silently.