After a bad outage someone asks for "a quick summary for the board." Paste a Slack thread and you look chaotic. Give a ten-page RCA and you look evasive. Use a short script.
The four beats
- What customers felt: could not log in for 42 minutes; checkout failed for EU users; etc.
- How long: detect → mitigate → full restore.
- Why, in one sentence: bad deploy; vendor DNS; certificate; capacity.
- What we change: better check; maintenance windows; rollback drill; capacity headroom.
If you do not know (4) yet, say when the postmortem lands, then keep that promise (small-team postmortem).
Words that help
- "We were down" or "degraded" with the definition you use publicly
- Minutes and scope, not "briefly"
- "Our monitoring caught it in X minutes" or "a customer caught it first" (both honesty)
Words that hurt
- Blame storms about a junior engineer
- Untranslated acronyms
- Guaranteed never-again promises
- Fake precision ("exactly 99.992%") when your monitors cannot support it
Metrics they might ask for
Translate nines into hours. Mention MTTD/MTTR only if you will explain them in a clause each. Point to the public status history if it exists; third-party visible truth beats a slide.
Calm, specific, forward-looking. The product is trust. Downtime is a withdrawal. The explanation is how you deposit again.
