EI-301i · Module 1

Trigger Calibration

3 min read

Trigger calibration is the ongoing process of adjusting trigger thresholds to minimize false positives (alerts that turn out to be noise) and false negatives (genuine signals that did not trigger an alert). Start with sensitive triggers that may over-alert, then systematically tighten thresholds based on feedback from recipients. The calibration data: for each alert, was the recipient's response "this is actionable" or "this is noise"? After 30 days of feedback, you have enough data to adjust trigger thresholds for each alert type.

Do This

Start with sensitive triggers and tighten based on false positive data — it is safer to over-alert initially than to miss genuine signals
Collect feedback from alert recipients: was this alert actionable? — binary feedback on every alert calibrates the system fastest
Review trigger thresholds monthly during the first quarter, then quarterly once calibrated — thresholds drift as the ecosystem evolves
Maintain separate thresholds for different source types — a blog post change trigger should be different from a pricing page change trigger

Avoid This

Set triggers once and never adjust — the optimal threshold changes as the ecosystem's baseline noise level changes
Calibrate based on volume targets ("we want exactly 10 alerts per week") — calibrate based on actionability, not volume
Tighten triggers so aggressively that you miss genuine signals to eliminate noise — a missed signal is more costly than a false positive

The precision-recall tradeoff applies directly to alert trigger design. Precision is the percentage of alerts that are genuinely actionable (higher precision means less noise). Recall is the percentage of genuine signals that trigger alerts (higher recall means fewer missed signals). You cannot maximize both. For ecosystem intelligence, the correct bias is toward recall — it is better to receive an occasional false positive than to miss a genuine signal that could have driven a timely response.