I page when the exception queue ages past a threshold during operating hours, when a job that always has rows returns none on a weekday, and when write error rates spike.
I do not page on Saturday night because a vendor did not send a file they never send on Saturday.
Quiet hours are a product decision
The warehouse has a clock. Alerts should too. A digest at open is better than a siren at 2am for a report nobody reads until 8.
Make the alert the next action
A link to the failed rows beats a Slack message that says "error." If I cannot click to the queue, I have only created anxiety.
