Definition
The spread between reliability-engineer-judged criticality and AI-cost-optimized safety stock level for the same SKU, measured as the SKU-weighted gap between the two judgments.
The Measurement Gap
The reliability team classifies a turbine blade cooling passage insert as criticality-1. Irreplaceable, single-source, catastrophic production impact if unavailable. The AI cost optimizer, looking at carrying cost, demand frequency, and lead time, recommends reducing safety stock from 4 units to 1. The optimizer is not wrong on its own terms: the carrying cost is high, the demand is rare, and the model correctly identified it as slow-moving. But the optimizer does not have the reliability team's judgment that this part cannot be substituted, cannot be expedited, and has a 26-week lead time from the only qualified manufacturer. The two systems are measuring different things (cost efficiency versus operational risk) and there is no metric that tells you when they diverge. Inventory accuracy does not distinguish between well-stocked and critically-under-stocked. Fill rate averages the easy wins with the catastrophic misses. ABC classification captures movement-based demand but not reliability-based criticality. Carrying cost percentage rewards lower stock without weighing the production-impact distribution of a stockout. The AI optimizer is doing exactly what it was designed to do; the reliability engineer is doing exactly what they were trained to do; and the gap between their judgments is invisible to every standard MRO supply chain metric.
Decision Supported:
Which SKUs require reliability-engineer review before AI cost optimization is applied, and whether the criticality classification should be a hard input constraint to the optimizer rather than a soft consideration.
Formula
Plain language: Take the reliability team's criticality rating for each SKU and translate it to a stocking-policy implication (quintile rank). Take the AI optimizer's recommendation and rank it the same way. Measure the gap between the two ranks across all SKUs.
Notation:
CODI = Sigma_skus |rank_criticality(s) - rank_ai_stock(s)| / N_skus
Variables:
| Variable | Definition | Unit |
|---|---|---|
| rank_criticality(s) | Quintile rank from reliability criticality classification for SKU s | 1–5 |
| rank_ai_stock(s) | Quintile rank of AI-recommended safety stock level for SKU s | 1–5 |
| N_skus | Count of SKUs in scope | Count |
| s | Individual SKU identifier | ID |
Proposed formula, pending validation pilots.