Definition
Percentage of AI-predicted component replacements where post-removal inspection reveals the component had significant remaining usable life.
The Measurement Gap
Legacy maintenance metrics track whether work was completed on time and within budget, but they cannot distinguish between a justified predictive replacement and a wasteful one. Schedule compliance tells you the job got done; it says nothing about whether the job needed doing yet. Component cost tracking aggregates spend by category without separating premature pulls from end-of-life replacements. Mean time between replacement treats all intervals equally, whether the component was exhausted or barely worn. Even condition monitoring dashboards that display severity scores and remaining useful life estimates provide no feedback loop on whether those estimates proved accurate after the component was removed. Without post-removal assessment data linked back to the triggering prediction, organizations have no way to quantify the cost of model conservatism. The predictive system claims credit for every prevented failure while the cost of unnecessary replacements remains invisible in aggregate maintenance spend.
Decision Supported: Is the predictive maintenance program costing more in premature component pulls than it saves in prevented failures, and which component types are over-predicted?
Formula
Plain Language
Count the number of AI-triggered component removals where post-removal inspection showed the component still had meaningful life remaining. Divide that count by the total number of AI-triggered removals. Multiply by 100 to express as a percentage. Separately, sum the cost of each premature removal (component value consumed, labor expended, and production downtime incurred) to quantify the dollar impact of model conservatism.
Notation
PRR = (Premature_removals / Total_AI_triggered_removals) x 100
Premature_removal_cost = Sum(Premature_removals x (Component_cost + Labor_cost + Downtime_cost))
Variables
| Variable | Definition | Unit |
|---|---|---|
| Premature_removals | Components removed on AI recommendation where post-removal assessment shows remaining life above the site-defined "still serviceable" threshold | Count |
| Total_AI_triggered_removals | All component removals where the work order was created based on a predictive alert | Count |
| "Still serviceable" threshold | Site-specific, component-specific condition standard defining minimum remaining life that qualifies as premature | Component-dependent |
| Component_cost | Value of the component consumed by the premature removal | Currency |
| Labor_cost | Maintenance labor hours expended on the removal | Currency |
| Downtime_cost | Production hours lost during the removal activity | Currency |
Proposed formula, pending validation pilots.