What should an automation failure alert say?
A failure alert should name the system and the step that failed, give the reason in plain words, and say the one next action, so whoever owns the alerts can act without opening anything. It ends with a run reference, and the resolved message repeats it. This drill steps through an example card, one line at a time.
Walk through one alert
This alert is an example in made-up words, with placeholder references. Press Next to go through it one line at a time, from the failure to the fix.
[ ALERT ] DEMO · Lead intake failed at sending the approved reply
What failed: the email account's login expired, so the approved reply did not go out.
Next: reconnect the email account, then send the approved reply again.
#demo · run demo-run-id
The failure
A lead was approved for a reply, and the step that sends it failed. Without an alert, nobody hears about it until the lead has gone elsewhere. This card is the message whoever owns the alerts gets instead.
After the fix
[ RESOLVED ] DEMO · Approved reply sent after reconnecting
#demo · run demo-run-id
If one of your automations could fail without anyone hearing about it, we can go through what its alert should say on a call.
How this works
- First, the failure. A lead's reply was approved, and the step that sends it failed. Without an alert, nobody hears about it until the lead has gone elsewhere.
- The first line names the system and the step, so whoever reads it knows which automation failed and where without opening anything.
- The second line gives the reason in plain words: the email account's login expired, so the approved reply did not go out.
- The Next line gives one action, written so whoever owns the alerts can act on it: reconnect the email account, then send the approved reply again.
- The last line is the run reference. Once the account is reconnected and the reply goes out, the resolved message carries the same reference, so the log shows the failure and the fix as one run.
Assumptions and limits
- The card is an example. Its system code DEMO, its tag #demo and its run reference demo-run-id are placeholders, and the failure is made up.
- A drill shows the format of a useful alert. It does not prove that any system sends alerts like it.
- Sending the reply again is this card's instruction for this one failure. Whether a step is safe to run again depends on how it was built, so ask to see that on a test record.
- The card names nobody, because who gets the alerts is set for each system.
Questions
Who should get the alerts?
A person by name who can fix the step or call the person who can, with a named backup for when they are away. A group chat where everyone assumes someone else is on it counts as nobody.
What does a good Next line look like?
One action, in words the reader can act on without asking anyone: reconnect the email account, then send the approved reply again. A line that only says an error happened leaves the reader to work out the rest, and that is where the time goes.
What should I ask whoever builds my automations?
Ask them to break a test run on purpose, by expiring a test login, and show you the alert it produces. Check that it names the system and the step, gives the reason and the next action, and carries a reference you can search for. Then ask who gets it, and who covers when they are away.
Why repeat the run reference on the resolved message?
So the failure and the fix read as one record. Anyone who searches for the reference later finds the alert, what was done and when it was resolved, without piecing it together from a chat.
Related
- What a reliable automation looks likeFive properties any automation should have, checkable by an owner who doesn't build: logged runs, named alerts, one send per approval, safe reruns, an owner.
- Automations that fail without telling anyoneAn automation can stop for weeks before anyone notices. The usual causes, what not knowing costs, and what an alert should say when a step fails.
- How an install runs, from workflow map to handoverHow I install a business system: map one workflow, put numbers on it, build the simplest fix, measure it, and hand it over running in accounts you own.
- Records from forms and emailed PDFs, kept in one place that stays rightA system I built that keeps the right version of each customer record in Airtable. Forms and emailed PDFs feed it, and HubSpot copies from it.