A good escalation includes the customer impact, exact identifiers, steps to reproduce, what's already been checked, and the evidence that points to a bug. An engineer should be able to start investigating without asking a single follow-up question.
The checklist
- Customer and account ID, plan, and how many accounts are affected
- Request IDs and timestamps with time zone
- Expected versus actual behavior, in one sentence each
- Reproduction steps, or the reason it can't be reproduced
- Relevant log lines, traces, or screenshots
- What support already ruled out
- Severity and any SLA deadline
What to leave out
- The full ticket thread pasted in without a summary
- Guesses presented as facts
- Urgency without impact data
Decimal writes the escalation for you, with the suspect code path and evidence attached. See bug report.