Autonomous Reorder Confidence (ARC)

GMI-MRO-033MROIntegration-Required

Definition

ARC measures the calibrated probability that an auto-generated purchase order will result in the ordered parts actually being consumed within their planned maintenance window, and whether the complete material kit arrives correctly.

The Measurement Gap

Autonomous procurement systems make financial commitments (predictions with a price tag) but standard metrics cannot judge whether those commitments are justified. Fill Rate tells you the part was in the storeroom, not whether it should have been ordered. Inventory Turns cannot distinguish parts genuinely consumed from parts issued just to close a stale reservation. Maintenance metrics (PM Compliance, MTBF) have no concept of procurement accuracy. Even model-level metrics (precision, recall, F1) measure prediction quality, not the procurement action that followed, a model can be 85% accurate while generating waste on 40% of orders because the buying decision adds layers of uncertainty (right part? right quantity? right timing? complete kit?) the prediction score does not capture.

Decision Supported:

Whether the autonomous procurement system deserves the authority it has been granted, and which failure mode (wrong prediction, wrong timing, wrong specification, kit incompleteness) requires the corrective action.

Formula

Plain language: Combine two value-weighted measurements. Whether the auto-ordered part was actually consumed against the originating work order within the planned window (consumption hit rate), and whether the full kit of planned materials was correct and ready when work began (kitting accuracy).

Notation:

ARC = α × ConsumptionHitRate + (1 − α) × KittingAccuracy
ConsumptionHitRate = Σ(V_j × I(consumed_j within window_j)) / Σ(V_j)
KittingAccuracy = WOs_with_complete_correct_kits / Total_AI-triggered_WOs_with_planned_materials

Variables:

Variable Definition Unit
V_j Dollar value of auto-generated PO line j $
I(consumed within window) 1 if the part was issued against the originating WO inside the planned window Binary
window_j SCHEDSTART to SCHEDFINISH + configurable buffer (e.g., 30 days) Date range
α Weighting between consumption and kitting components (default 0.6) Dimensionless
Complete correct kit All planned material lines available, in correct quantity, at correct storeroom, before work began Yes/No

Proposed formula, pending validation pilots.

At a Glance

Unit
Score (0 to 1)
Trend Direction
Higher
Reporting Frequency
Quarterly
Category
Autonomous MRO & Supply Chain Elasticity
← Algorithmic Just-in-Time Arrival Accuracy (AJTAA)Cross-Site Inventory Arbitrage (CSIA) →

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