The Acceptable-Flapping Policy for Noisy Alerts
Some alerts flap. Most are noise; some are signal. The policy that turns flapping into structured response.
Track flaps
An alert that fires more than 3 times in 24 hours without a sustained incident is a flap.
Flaps are auto-tagged. Dashboard shows the top 10.
Each flap is a candidate for tuning, suppression, or removal.
The decision tree
Real signal but threshold too tight: tune the threshold. The alert fires only when the issue is meaningful.
Real signal but transient: aggregate. Five flaps within 5 minutes = one alert.
Not real signal: remove. The alert was wrong.
Weekly review
Top 10 flapping alerts. Each gets a decision: tune, aggregate, remove.
Decision is logged. The same alert flapping next month should not require re-deciding.
Result: the team's alert pool converges on signal-bearing alerts only.