Support · Alerts
Beating alert fatigue
Monitoring dies by notification. The mute-test governs design: if a senior engineer would mute this channel in a week, redesign before shipping.
The mute-test
Before shipping any alert, ask: would a senior engineer mute this channel within a week? If yes, the alert is designed wrong - fix the trigger or the threshold, not the engineer.
Practical levers
Raise immediate-email thresholds to true materiality (the $500 default exists for a reason). Let digests carry below-floor visibility instead of interrupts. Scope chronically noisy accounts into their own digest rather than deafening the shared channel.
Signals you are getting it right
Alerts get acted on within the payout window; nobody has muted the channel; the digest gets read because it contains only things worth reading. All three are measurable - check monthly.