Definition
Measures the net economic value of acting on a predictive maintenance alert, calculated as the cost of the prevented failure (weighted by failure probability) minus the production cost of the maintenance interruption.
The Measurement Gap
Predictive maintenance programs are typically evaluated on prediction accuracy: what percentage of predicted failures actually occurred. This accuracy metric is necessary but insufficient for manufacturing, because it ignores the production context of each intervention.
Consider two PdM alerts, both at 80% confidence. The first predicts a bearing failure on a non-critical asset during a slow production week. Acting on this alert costs $500 in production interruption and prevents a $5,000 repair. The second predicts the same bearing failure on a critical asset during a high-volume contract run. Acting costs $50,000 in production interruption (lost output, changeover, expediting) and prevents the same $5,000 repair. Both alerts are equally accurate, but the first creates $3,500 in net value while the second destroys $45,000.
The gap is that PdM programs treat all alerts at the same confidence level identically, regardless of production context. Without MINPV, organizations cannot answer the most important question about their PdM program: is it creating net value when production disruption costs are included? They may be celebrating 85% prediction accuracy while the program is net-negative because too many interventions interrupt high-value production to prevent low-cost failures.
Formula
MINPV is calculated per intervention and aggregated across the portfolio:
MINPV_i = (C_failure_i x P_failure_i) minus C_intervention_i
Aggregate MINPV = Sum of MINPV_i / N
Where:
- C_failure_i = estimated total cost of the predicted failure if it occurred (dollars, including repair, downtime, quality scrap, and safety or regulatory exposure)
- P_failure_i = probability the failure would have occurred in the relevant time horizon if not addressed (0 to 1, derived from the PdM model's confidence or from historical failure data at that condition level)
- C_intervention_i = total cost of acting on the alert (dollars, including production interruption cost, maintenance labor, parts, and changeover or restart costs)
- N = number of interventions in the measurement period
Positive MINPV means the intervention created net value. Negative means it destroyed value. Aggregate MINPV is the portfolio-level answer to whether the PdM program creates value net of production disruption.
Key considerations: P_failure is estimated, not known. Use the PdM model's confidence score as a starting proxy, then refine with historical data showing what percentage of alerts at each confidence level actually resulted in failure when deferred. C_failure is also estimated, using historical average failure cost for the same failure type and asset class. The "do nothing" counterfactual is unknowable for any single event, so MINPV works as a portfolio metric over many events. Minimum sample of 30 PdM-driven interventions per measurement period.
Proposed formula, pending validation pilots.