Lead Time Prediction Fidelity (LTPF)

GMI-MROSC-012MRO Supply ChainPilot-ReadyPattern O-3: Twin Fidelity

Definition

The agreement between AI-predicted lead time and actual realized lead time across vendor segments, tracked separately from vendor-stated lead time so the AI is graded against ground truth rather than a moving target the vendor controls.

The Measurement Gap

Legacy MRO supply chain metrics cannot isolate AI lead-time prediction quality from vendor performance. Vendor on-time delivery compares actual against vendor-stated lead time: the vendor controls the target. Lead time variance measures realized variability without identifying which prediction source failed. Forecast accuracy mixes lead-time prediction error into demand-side errors. When the AI predicts 21 days, the vendor quotes 28 days, and the part arrives at 39 days, three independent failures are occurring simultaneously: vendor-stated lead time drifted from vendor capability, AI prediction drifted from reality, and safety stock was sized against a signal that no longer reflects ground truth. Without clean separation of AI prediction fidelity from vendor adherence, the operation cannot diagnose whether to recalibrate the AI, renegotiate with the vendor, or both. Every stocking decision and expediting trigger downstream inherits this ambiguity. The decision LTPF supports: whether to plan replenishment against vendor-stated, AI-predicted, or a blended lead time, and when to expedite preemptively based on predicted-versus-stated divergence.

Formula

LTPF = 1 − mean( |LT_predicted − LT_actual| / LT_actual ) by vendor segment

Variables:

  • LT_predicted = AI-predicted lead time at PO placement time
  • LT_actual = days from PO placement to receipt (MATRECTRANS receipt date)
  • LT_stated = vendor-quoted lead time (tracked separately, not used in LTPF directly)

Worked example: 1,840 PO lines placed with AI lead-time predictions last quarter. Aggregate absolute percent error of prediction versus actual: 14.7%. LTPF = 1 − 0.147 = 0.853. By segment: domestic vendors LTPF = 0.91 (well-calibrated); offshore vendors LTPF = 0.72 (poorly calibrated, vendor mix shifted post-tariff). Expedited POs excluded from natural-flow measurement.

Edge cases: Expedited POs distort the metric. Separate natural-flow and expedited cohorts. Foreign vendors and high-disruption segments tracked separately. New vendors with insufficient history excluded until at least 20 POs accumulate.

Data sources: Predicted lead time from lead-time AI platform (must be captured at placement, not regenerated later). Vendor-stated lead time from PO terms or vendor master. Actual lead time from PO placement date to MATRECTRANS receipt. Vendor segmentation from vendor master (geography, category) and item master.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Quarterly
Category
E - Supplier Reality & Lead Time Truth
← Substitution Fitness Yield (SFY)Vendor Consolidation Fragility Index (VCFI) →

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