CX-201a · Module 2

Alert Threshold Design

3 min read

An alerting system with too many alerts is worse than no alerting system at all. Alert fatigue is real — when every score fluctuation triggers a notification, the CSM stops reading them. Then the real alert arrives and it sits unread alongside forty false alarms. Threshold design is the discipline of configuring alerts that fire when action is needed and stay silent when it is not.

  1. Absolute Thresholds Fire when a score crosses a fixed level. Health score drops below 60: amber alert. Drops below 40: red alert. Absolute thresholds catch accounts that enter danger zones regardless of their trajectory. Simple, reliable, and the foundation of any alerting system.
  2. Trend Thresholds Fire when a score declines by a defined amount over a defined period. Health score drops 15 points in 30 days: trend alert. Trend thresholds catch declining accounts that have not yet crossed the absolute threshold — the account at 75 dropping to 60 that will be at 45 next month if nobody intervenes.
  3. Behavioral Triggers Fire when a specific behavioral event occurs regardless of score. Champion contact goes silent for 7 days: behavioral trigger. Stakeholder attendance drops to one person: behavioral trigger. Behavioral triggers catch the qualitative signals that scores may not capture — the events that indicate the relationship is changing.

Do This

  • Layer absolute, trend, and behavioral triggers for comprehensive coverage
  • Tune thresholds based on false alarm rate — if more than 20% of alerts do not require action, thresholds are too sensitive
  • Route different alert types to different response owners — the CSM handles behavioral triggers, the team lead handles red alerts

Avoid This

  • Set thresholds so sensitive that every minor fluctuation triggers an alert
  • Rely solely on absolute thresholds — they miss declining trajectories above the threshold
  • Send all alerts to the same person with the same priority — differentiation drives appropriate response