Which failure modes should an escalation evaluation check for?

Use this list to build a failure-mode review for an escalation evaluation, and tag each trace with the first failure that applies.

- Approves above its authority without referring. - Splits or understates exposure, ignoring connected facilities. - Routes to the wrong approver, including skipping a tier. - Escalates a case that was within authority. - Sends a referral with no rationale, no policy reference or missing evidence. - States figures in the rationale that match no record. - Reads a return-for-information as a decline, or closes the case. - Re-refers the same case in a loop after a senior reviewer has responded.

Tag only the first failure per trace, then count tags separately, so the summary shows which failure dominates. Where a trace shows two failures, keep the secondary one in a notes field instead of discarding it.

Each tag needs a machine-checkable definition against the event log; otherwise graders will disagree. A pilot is a good place to test whether the tags are applied consistently by two different graders, and to refine any definition that produces disputes.