A bridge sprawls when nobody owns the call. Assign roles in the first two minutes, keep a running log, and update stakeholders on a clock — not when someone remembers. Print this or pin it next to your incident tooling.
Assign roles first (before any debugging):
- Incident commander: Runs the call, makes decisions, keeps everyone on the current problem. This person does not debug — they coordinate.
- Scribe: Timestamps every action, finding, and decision in one shared doc. If it isn’t logged, it didn’t happen.
- Comms lead: Owns stakeholder and status-page updates so the commander and responders can stay heads-down.
- Responders: The people actually investigating. They report findings to the commander, not the whole bridge.
Run the call:
- Open by stating impact, severity, and the current working theory in one sentence.
- Keep one thread of investigation at a time; the commander parks side-ideas in the log.
- Debug findings go to the commander, who decides what to act on — no unilateral changes to production.
- Anyone who makes a change announces it before and after: what, why, and the observed effect.
Comms cadence:
- Post a status update on a fixed interval (every 15–30 minutes for a major incident) even if the update is “still investigating, no change.”
- Keep internal detail out of external status posts; state impact and next-update time.
Close it out:
- Confirm the fix is verified, not just deployed — watch the metric recover before declaring resolution.
- Capture action items with an owner and a due date each; a bridge with no follow-ups just repeats.
- Schedule the post-incident review while the bridge is fresh, and save the scribe’s log as the timeline.
