SOP: Push alerts vs pull dashboards for monitoring

SOP: Push alerts vs pull dashboards for monitoring

Run this whenever you set up how the team learns something went wrong.

Stage 1 — Classify the signal (owner: on-call owner)
☐ Does it require action within minutes, or is it a trend you review weekly?
☐ What's the false-positive rate if you alerted on it?

Stage 2 — Choose
☐ Push (page/alert): interrupts a human now. Use only for signals that are urgent, actionable, and rare. Cost: every noisy alert trains people to ignore the channel.
☐ Pull (dashboard/report): waits to be checked. Use for trends, capacity, and anything not minute-critical. Cost: nobody looks unless it's on a schedule.

Stage 3 — Guardrails (owner: on-call owner)
☐ Every alert links to its runbook. No runbook → not an alert.
☐ Dashboards get a named owner and a review cadence on the calendar.
☐ Done-when: alert volume is low enough that every page is real.

Trigger: setting up monitoring. Done-when: each signal routed push or pull with a runbook.

Save this. Run it every time.