Status page incident update templates

A status page post is read by worried customers in a few seconds. This page gives a copy-paste template for each stage of an incident (investigating, identified, monitoring, resolved), a short rule for what each one must and must not say, and a worked example. The incident in the example is invented.

The four stages

Most status page tools, including Atlassian Statuspage, use the same four labels: Investigating, Identified, Monitoring and Resolved. Use them in order and post a new update each time the stage changes, or when you promised the next update, even if nothing has changed.

1. Investigating

Investigating: [service or feature] is [not loading / slow / returning errors] for some [customers / users]. We are looking into it. Next update by [time and zone].

Must say: what customers will notice, and when you will next write. Must not say: a cause. At this stage you do not have one, and a wrong guess is hard to take back.

2. Identified

Identified: We have found the cause of the [symptom] affecting [service or feature] and are [applying a fix / rolling back a change / waiting on a provider]. Next update by [time and zone].

Must say: that you know what is wrong and what you are doing about it. Must not say: a fix time unless you are sure of it. "We expect an update by 15:00 UTC" is a promise about writing, not about repairing.

3. Monitoring

Monitoring: A fix has been applied and [service or feature] is working again. We are watching it to make sure it stays that way. Next update by [time and zone].

Must say: that service is back and that you are still watching. Must not say: that the incident is over. Many repeat failures happen in this window.

4. Resolved

Resolved: [service or feature] has been working normally since [time and zone]. It was affected from [start time] to [end time] ([duration]). [One sentence on what happened, only if confirmed.] [A full report will follow by DATE / No further action needed.]

Must say: the start and end of the impact, in one time zone. Must not say: "it will not happen again", a credit you have not agreed, or the name of the person who made the change.

A worked example (invented)

09:20 UTC  Investigating: Online booking is returning an error for some customers. We are looking into it. Next update by 09:50 UTC.
09:36 UTC  Identified: We have found the cause of the booking errors, a full server disk, and are clearing space. Next update by 10:00 UTC.
09:50 UTC  Monitoring: Online booking is working again. We are watching it to make sure it stays that way. Next update by 10:30 UTC.
10:30 UTC  Resolved: Online booking has been working normally since 09:48 UTC. It was affected from 09:12 to 09:48 UTC (36 minutes). A full report will follow by Tuesday.

Notice that the start time (09:12) is the time the problem began, not the time the first update was posted (09:20). If you only know when you noticed it, say "from about 09:12 UTC" or leave the start as a blank until you have checked your logs.

Seven quick checks before you post

One time zone throughout, written next to every time.
Plain words for the symptom: "cannot log in", not "authentication degradation".
No cause in the first two updates.
A next-update time in every update that is not Resolved, and keep it.
No blame, no names, no "human error".
No promises about credits, refunds or the future unless someone with the authority has agreed them.
Nothing a customer could not safely forward to their own boss.

After the incident

The Resolved post is short on purpose. Larger incidents also need a fuller report for the customers who were affected. Our guide to customer incident reports covers the seven parts of that report, and the free timeline calculator and checker works out detection time, repair time and total duration from your timeline and checks a draft report for missing parts. It runs in your browser and uploads nothing.

If you would like a first draft of the report, a status page post and a cover email made from your own incident notes, with every time or cause you did not give left as a marked blank, see the $9 draft service (AI-written by Claude, not legal advice).