Environmental Alert Action Rate (EAAR)

GMI-PH-008Pharmaceutical OperationsPilot-Ready

Definition

Percentage of environmental monitoring alerts receiving a documented investigation, physical intervention, or formal dismissal with rationale. Paired with a companion ratio comparing excursion probability after dismissed alerts against the no-alert baseline.

The Measurement Gap

The Gap: Environmental Excursion Rate shows excursions after they happen but cannot tell you that the early warning system generated alerts that were ignored beforehand. Alert Volume is tracked by IT as a system performance metric, not a quality metric. No one connects alert disposition patterns to excursion outcomes. The classic alert fatigue trajectory (initial engagement, rising alert volume, selective attention, missed genuine signals) plays out invisibly.

The Decision:

EAAR below 40% triggers an alert rationalization project to reduce false positive rate. DEC above 3.0 triggers immediate investigation of dismissed alerts as excursion root cause contributors. Together, EAAR and DEC become the effectiveness metric for the environmental monitoring AI investment: proving whether the monitoring system is generating value or generating noise the team has learned to ignore.

Formula

Notation:

EAAR = (Σ Actioned_Alerts / Σ Total_Alerts) × 100

Companion: DEC = P(Excursion within 72h | Alert Dismissed) / P(Excursion | No Alert)

Variables:

  • Actioned_Alerts = alerts with documented investigation, physical intervention, or formal dismissal with rationale
  • Total_Alerts = all AI-generated pre-emptive environmental alerts
  • DEC numerator: probability of excursion within 72 hours following a dismissed alert
  • DEC denominator: baseline excursion probability during any equivalent 72-hour period without a preceding alert

Worked Example: Last month: 92 total alerts. 23 investigated, 11 resulted in intervention, 8 formally dismissed with rationale, 50 received no documented response. EAAR = 42/92 × 100 = 45.7%. For DEC: 50 dismissed alerts, 4 followed by excursions within 72 hours. P(Excursion|Dismissed) = 8.0%. Baseline without alerts: approximately 0.3% per 72-hour block. DEC = 8.0/0.3 = 26.7. Dismissed alerts are 27 times more likely to precede excursions than random periods.

Edge Cases:

  • Formal dismissal with documented rationale counts as actioned: EAAR measures documented response, not physical intervention. Deliberate documented non-action is a quality decision; silence is not.
  • If DEC is approximately 1.0, dismissed alerts have no predictive value: the model's false positive rate is the problem, not the team's response
  • If DEC is much greater than 1.0, the signal is real and the organization is ignoring it. Fix the response process, not the model.

Sign convention:

Higher EAAR is better. Higher DEC is worse.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Monthly; per environmental alert
Category
Human-AI Trust & Override Governance
← Diversion Alert Investigation Quality (DAIQ)Reject Attribution Accuracy (RAA) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment