A blameless account of an outage for the whole company: impact, timeline, cause and fixes with owners.
A postmortem is how one team's bad morning becomes something the whole company learns from. Read the finished one here, about a made-up outage in which an expired certificate stopped payments for 51 minutes, and take the prompt to write up your own incident. Write it within a few days, while the chat log still makes sense.
Who it's for
An engineering lead or on-call engineer writing for the whole company after something customers noticed: an outage, lost data, a release that had to be rolled back. Support, sales and the executives will read it, so it has to make sense to someone who has never heard of a load balancer.
Customers sometimes need an account too. Don't forward this one. Write them a shorter page with the impact, the cause in two sentences and what you've changed.
What a good one contains
- A three-sentence summary. What happened and for how long, why, and what you've done. Many readers stop here.
- Impact in numbers a customer would feel. The sample counts 1,280 failed checkouts, 448 orders never completed and 412 shops. It also says what wasn't touched: no card details exposed.
- A timeline to the minute, in one time zone. Fourteen entries run from 9:12 to 10:40. The gaps tell the story: 14 minutes passed before anyone was paged.
- The root cause in plain words. The sample says what a certificate is before it says why this one expired. If only engineers can follow the cause, only engineers learn from it.
- Why nobody noticed sooner. This is often a second cause. Here the alarm counted failed requests, and an expired certificate produces none.
- What went well. Support paged after six matching tickets. Writing down what worked tells people to keep doing it.
- What didn't, described as systems. "Thirty warnings were sent and none reached a person" points at a broken route for alerts and at no person.
- Action items with an owner, a date and a status. Six in the sample, two already done. An item with no name on it is a wish.
Keeping it blameless
Blameless doesn't mean vague. Say exactly which decision led where: the DNS move left the renewal job off its checklist. What you leave out is a verdict on a person. If a sentence could be read as "and that was Sam's fault", rewrite it around the system that let one slip reach customers.
Check every time against the logs and add up the minutes. The sample's 14, 8 and 29 come to 51. When an action item closes, change its status, and the link people already have shows where things stand.
How to send it
To named people. An internal postmortem lists gaps you wouldn't publish, so only people at your company should be able to open it. Customers get a shorter version, written for them.
Every page you publish gets one link that opens in any browser. You choose who can open it: anyone with the link, anyone with the password, or only the people and company domains you name, who sign in with Google or Microsoft. Change the page later and the link stays the same.
