Multi-Signal Alert Design

Single-signal alerts are the primary cause of alert fatigue. A response time that exceeds 48 hours triggers an alert. A meeting cancellation triggers an alert. A usage dip triggers an alert. Each signal, in isolation, is ambiguous — the client may be on vacation, reorganizing priorities, or simply busy. Multi-signal alerts require corroboration before firing: response time increased AND meeting attendance declined AND usage dipped. The corroboration eliminates the majority of false positives because individual signal noise averages out across multiple independent measurements.

1. Define Signal Groups Organize leading indicators into signal groups by category: engagement signals (response time, meeting attendance, contact ratio), adoption signals (usage trends, feature depth, team breadth), and sentiment signals (tone, questions, proactive engagement). An alert fires when multiple signals within or across groups align — not when any individual signal fluctuates.
2. Set Corroboration Requirements For each alert level, define the minimum signal corroboration. Amber alert: two signals from the same group declining simultaneously. Red alert: signals from two or more groups declining simultaneously. Critical alert: signals from all three groups declining simultaneously. The corroboration requirement determines the alert's specificity — how confident you are that a real problem exists.
3. Time-Window Alignment Signals must be measured over the same time window to be corroborated. Response time declining over the past 14 days corroborates with meeting attendance declining over the past 14 days. Response time declining this week does not corroborate with a meeting cancellation from last month. Time-window alignment prevents temporal noise from creating false corroboration.