Opening the book…
During an incident, silence is its own failure, because the people affected, teammates, and anyone waiting on the system, fill an information vacuum with worst-case guesses and a stream of 'any update?' pings that distract the very people trying to fix it. Regular updates, even ones that say 'still investigating, next update in fifteen minutes,' are what keep everyone calm and out of the responders' hair. Communicating during the fire also creates a real-time record that makes the postmortem accurate instead of reconstructed from foggy memory, and it lets people make good decisions, like whether to tell a customer or hold a launch, based on current reality. The instinct under pressure is to go heads-down and communicate nothing until it's fixed, which feels focused and actually makes everything harder, because it leaves everyone anxious and uninformed. A steady drumbeat of honest updates is part of the response, not a distraction from it.
Assign someone, often not the person hands-on-keyboard, to own communication during an incident. Post updates on a regular cadence to a known place, even when the update is just 'no change yet, still on it,' because the absence of news reads as chaos. Say what you know, what you're doing, and roughly when the next update will come, and keep the tone calm and factual rather than either falsely reassuring or alarmist. Tell affected people early rather than hoping to fix it before anyone notices, since being upfront buys trust and being caught hiding spends it. When it's resolved, say so clearly, and flag that a postmortem will follow. Keep a rough timeline as you go, because it's the raw material for learning later.
[14:05] INVESTIGATING — checkout returning 500s
for ~30% of users. Team on it. Next update 14:20.
[14:20] IDENTIFIED — payments provider timing out;
rolling out retry + failover. Next update 14:35.
[14:35] MONITORING — error rate back to normal after
failover. Watching for 30 min. Postmortem to follow.For a very short, contained blip that's resolved in minutes and affected no one, a full comms cadence is overkill, a quick after-the-fact note suffices. And internal updates should stay honest and detailed even when external, customer-facing communication is necessarily more measured.