| name | escalation-runbook |
| description | Use when an issue outgrows the support queue or threatens an account — assigns a single owner, sets severity and communication cadence, drives resolution, and closes the loop with the customer plus a prevention step so the same fire doesn't start twice. |
Escalation Runbook
A playbook for handling the issue that jumped a team — a bug, an outage, an account-threatening problem — so it gets one owner, honest communication, and a real close instead of a silent drift. The failure mode it prevents: an escalation that everyone is "aware of" and no one owns, going dark on the customer until the account is lost.
Own it and size it
- Assign one named owner on contact. The customer must know exactly one person now owns their problem — ambiguity about ownership is how escalations rot.
- Assess severity by impact and blast radius: is one account or many affected, is it degraded or fully broken, is a renewal or a reference at stake. Severity sets the pace and who gets pulled in.
- Confirm the facts before you act — who's affected, since when, reproducible or not — so the response targets the real problem, not the loudest symptom.
Communicate and drive
- Set a communication cadence sized to severity and tell the customer what it is. A predictable "still working, here's where we are" beats a silent wait every time.
- Never give a false ETA and never go dark. If you don't know when it'll be fixed, say what you're doing and when you'll next check in — and hit that check-in.
- Drive the fix, don't just track it. Pull in the people who can resolve it, clear their blockers, and keep pressure on the path to resolution rather than narrating it.
Close the loop and prevent the repeat
- Close only when the customer confirms it's resolved from their side — an internal fix or a merged change is not a close.
- Run a short post-incident review: root cause, whether the signal was visible earlier, and what would have caught it sooner.
- Route every prevention item as a Task to the team that owns it — a documentation gap to knowledge, a triage or macro fix to support, a relationship repair to success — and log the whole trail on the escalation's Task.
Quality bar
- Exactly one owner, named to the customer.
- Severity assessed and cadence communicated, not improvised.
- No false ETAs, no silence.
- Closed by customer confirmation, and ending in an assigned prevention Task.
Output format
Escalation record: owner + severity + impact → communication log with timestamps → resolution confirmed by the customer → post-incident review with root cause and prevention Tasks assigned to named owners.