Terravek publishes a post-incident review for every incident that caused customer impact, on status.terravek.com, usually within five working days.
The structure is fixed: what customers experienced, the timeline in UTC, the technical cause, the contributing factors, what went well in the response, what did not, and the specific actions with owners and dates. Actions without an owner and a date are aspirations and we do not publish them.
There is no individual named as a cause. This is not politeness. An engineer who can make a change that takes down a production service has been handed a system where that was possible, and the interesting question is always why the system permitted it, not who happened to be at the keyboard.
The hardest section to write honestly is 'what did not go well in the response', because it is where the organisation is exposed rather than the technology. Our March review said that we took forty minutes to page the right team because the service ownership record was out of date. That is embarrassing and it is the most useful sentence in the document.
Customers occasionally tell us these reviews are the reason they renewed. That was not why we started publishing them, but it is a reasonable argument for continuing.