Every automation logs every run, retries on its own when something hiccups, and texts a person when it can't recover. At 6:30 every morning the owner gets one email: what ran, what broke, and which jobs need someone.
A run fails, then retries on its own with longer waits each time. It either recovers, or it texts a person with the error, what's held and a suggested fix. Work is held, never dropped.
Each bar is one day of the last 14. Amber means a retry fixed it, red means a person had to step in.
A separate watchdog checks that each automation actually ran when it should. An automation that stops quietly gets caught within 5 minutes, even when nothing errored.
| Automation | Expected | Last run | Check |
|---|
The watchdog runs on n8n, and a small Apps Script checks the watchdog itself, so the checker can't stop quietly either.
Every run, retry and check is written down, including the boring ones. If it isn't in the log, it didn't happen.
| Time | Automation | Result | Items | Took | Details |
|---|