Definition
CLAPV measures the fraction of customer leak alert notifications sent by AI-based detection systems that are confirmed as actual leaks upon investigation.
The Measurement Gap
Measurement Gap and Decision Supported
The Invisible Failure Mode: A utility launches an AMI-based customer leak notification program, sending automated alerts when continuous consumption patterns suggest a service line or fixture leak. In the first year, the system sends 4,200 alerts. Customer service tracks complaint calls from alerted customers but does not systematically track confirmation outcomes. Marketing reports the program as a success based on alert volume and the handful of thank-you calls from customers who found real leaks. Nobody computes the false positive rate. When an independent audit samples 200 alert recipients, only 112 had confirmed leaks. The other 88 either found no leak, had already known about the issue, or changed their perception of the utility from helpful to intrusive. The program's CLAPV is 56% (barely better than a coin flip) and customer satisfaction survey scores in alerted areas have declined.
What Existing Metrics Miss:
Leak detection platforms report detection sensitivity (what fraction of real leaks are detected) but rarely report positive predictive value (what fraction of alerts are real). A system with 95% sensitivity and 40% CLAPV detects almost every leak but floods customers with false alarms. Without CLAPV, utilities optimize for detection sensitivity while ignoring the customer experience cost of false positives.
The Decision CLAPV Enables:
Is the customer leak alert program credible enough to maintain customer trust, or does the false positive rate require raising the alert threshold, adding confirmation steps before notification, or restructuring the notification workflow? CLAPV provides the evidence base for this program governance decision.
Cross-References:
CLAPV connects directly to Alert Signal-to-Noise Ratio (ASNR, GMI-WT-010) from the maintenance KPI set. ASNR measures alert quality for internal operational alerts; CLAPV measures alert quality for customer-facing notifications. Both address false positive governance but in different stakeholder contexts. CLAPV also connects to LDCC (GMI-WT-022): customer leak alerts contribute to demonstrated leak detection coverage, but only alerts with confirmed outcomes count toward LDCC validation. Cybersecurity Alert Scoring Accuracy and Relevance Fraction (CSARF, GMI-WT-033) addresses the same false-positive governance problem in the OT security domain.
Formula
Plain language: Of all customer leak notifications sent based on AI smart meter analysis, count how many were confirmed as actual leaks (by field investigation, customer-reported repair, or subsequent consumption pattern normalization consistent with repair). Divide by total notifications sent.
Notation:
CLAPV = Confirmed_leaks / Total_notifications_sent
Where:
- Confirmed_leaks = notifications followed by: (a) field-verified leak, (b) customer-reported repair within 30 days, or (c) consumption pattern normalization within 30 days consistent with repair
- Total_notifications_sent = all AI-triggered customer leak notifications in the measurement period
Units: Percentage. Track separately: true positives (confirmed leak), false positives (no leak found), and unresolved (customer did not respond, no investigation performed).
Proposed formula, pending validation pilots.