Prediction Economic Yield (PEY)

GMI-TEL-007TelecommunicationsIntegration-RequiredPattern O-4: Alert Quality / Signal-to-Noise Ratio

Definition

Net economic value of acting on predictive maintenance alerts, comparing the cost of preventive interventions against the failure costs those interventions avoided.

The Measurement Gap

Measurement Gap and Decision

What PEY Measures: PEY calculates the net economic value of acting on predictive maintenance alerts by comparing the cost of preventive replacement to the estimated cost of the failure that would have occurred, adjusted for the model's demonstrated accuracy rate per equipment class.

Decision Supported:

Are predictive maintenance interventions on network equipment economically justified, or is the organization replacing equipment that would have run for months? Without PEY, predictive maintenance programs report "prevented failures" without accounting for false positives or comparing intervention cost against actual avoided loss.

Why This Is Unmeasured:

Predictive maintenance dashboards track alert volume, response time, and "failures prevented." None of these answer the economic question. A model that generates 500 alerts and "prevents" 50 failures sounds productive: until you discover 400 of those alerts were false positives, each costing labor and parts for unnecessary replacements, and the 50 "prevented failures" included equipment with months of remaining useful life. The counterfactual problem (replaced equipment didn't fail, so you can't prove it would have) makes this measurement uncomfortable, but the alternative is flying blind on ROI (TM Forum, "AI in Network Operations: Maturity Benchmarks," 2024; Analysys Mason, "Predictive Maintenance Economics in Telecom Infrastructure," 2023).

Formula

Plain Language: For each predictive intervention, estimate the cost of the failure that was allegedly prevented. Using historical MTTR, service credit exposure, and revenue impact for that equipment type. Subtract the cost of the preventive action (parts, labor, planned downtime). Adjust by the model's demonstrated accuracy rate for that equipment class, because some "prevented failures" were false positives. Sum across all interventions.

Notation:

PEY = Σ [(Estimated_failure_cost_i × P(correct_i)) − Preventive_action_cost_i] / Σ Preventive_action_cost_i

Variable Definitions:

  • Estimated_failure_cost_i = modeled cost of unplanned failure for equipment i, computed as service credits + lost revenue + emergency repair cost (dollars)
  • P(correct_i) = estimated probability that the prediction for equipment i was accurate, derived from the model's historical accuracy rate for that equipment class (dimensionless, 0–1)
  • Preventive_action_cost_i = actual cost of the preventive replacement including parts, labor, and planned outage impact (dollars)

Units:

Dimensionless ratio. Greater than 0 = predictions are net-positive economically. Less than 0 = predictions are destroying value (cost of acting exceeds value of avoiding failures). PEY of 1.0 means every dollar spent on predictive prevention saves an additional dollar in avoided failures.

Sign Convention:

Higher is better. Negative PEY means the organization would save money by ignoring the predictions entirely.

Edge Cases:

  • Counterfactual problem: replaced equipment did not actually fail, so the "avoided failure" is estimated, not measured. Use historical failure rate for that equipment age, type, and operating conditions as the baseline probability.
  • Predictions where the equipment was replaced for other reasons (end-of-life refresh, capacity upgrade): exclude from PEY. These are not prediction-driven actions.
  • Minimum sample: 30 predictive interventions per equipment class for meaningful PEY per class.

Proposed formula, pending validation pilots.

At a Glance

Unit
Ratio
Trend Direction
Higher; above 0
Reporting Frequency
Quarterly
Category
Predictive Asset Economics
← Storm Readiness Prediction Accuracy (SRPA)Prediction Portfolio Equity Score (PPES) →

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