Definition
The dollar penalty of elapsed time between an AI alert and the final go/no-go decision on the resulting work order, calculated as decision latency multiplied by the asset's marginal cost rate of delay.
The Measurement Gap
Maintenance metrics designed for calendar-driven or reactive work have no view of the deliberation phase that AI alerts create. MTTR starts the clock at first wrench. Schedule compliance measures whether approved work happens on time: it does not measure how long approval took. Recommendation-to-Execution Translation Loss (RETL) measures the entire alert-to-completion leak as a single bucket, but does not isolate the decision-phase cost from scheduling, parts, or execution delays. Each remediation is different: fixing decision delay requires governance redesign; fixing scheduling delay requires planning tools; fixing parts delay requires inventory policy. Measuring the total leak without segmenting cause is like diagnosing "the patient is sick" without identifying the organ.
Decision Supported:
Whether to redesign governance (tiered approval authority, expedited review thresholds, auto-approval rules) and which approval chains or asset classes consume the most prediction window.
Formula
Plain language: Measure the hours between when the AI alert fired and when someone definitively said "do it" or "don't do it." Multiply those hours by the asset-specific cost rate that captures rising failure probability, unplanned-failure consequence, and lost scheduling flexibility. The product is what indecision cost the organization on that alert.
Notation:
IDC = Decision Latency × Marginal Cost Rate of Delay
DL = T_decision − T_alert
MCRD = (P_failure_increase_per_hour × C_unplanned) + C_scheduling_flex_loss
Variables:
| Variable | Definition | Unit |
|---|---|---|
| T_alert | Timestamp when AI anomaly/probability crossed action threshold | Datetime |
| T_decision | Timestamp when WO status transitioned to APPR (go) or CAN (no-go) | Datetime |
| DL | Decision Latency: T_decision − T_alert | Hours |
| P_failure_increase_per_hour | Rate of failure-probability rise as the prognostic window shrinks | Probability/hour |
| C_unplanned | Cost of unplanned failure (downtime + emergency premium + collateral damage) | $ |
| C_scheduling_flex_loss | Value of expiring scheduling options (overtime avoidance, outage coordination) | $/hour |
Proposed formula, pending validation pilots.