Receipt Discrepancy Source Attribution (RDSA)

GMI-MROSC-015MRO Supply ChainInfrastructure-DependentPattern O-8: Sensor Disambiguation

Definition

The fraction of receipt discrepancies where the AI correctly attributes root cause (distinguishing item-master drift, vendor packing variance, scanning error, and counterfeit/substitution) rather than misrouting investigation effort to the wrong team.

The Measurement Gap

Legacy MRO supply chain metrics do not measure whether AI receipt discrepancy attribution is correct. Receipt cycle time measures process speed. Vendor dispute volume measures the consequence of wrong attribution (falsely accusing vendors) but not the attribution error itself. Inventory accuracy measures eventual cleanup. When the AI attributes a discrepancy to vendor packing variance and routes to a vendor query, but the actual root cause was an item-master update five months ago that changed barcode matching, the procurement team burns weeks on a vendor escalation that should never have happened. The same misattribution pattern repeating across dozens of receipts and multiple vendors consumes hundreds of hours of labor and damages vendor relationships. The AI's confidence in its attribution is what the team acts on, but the AI's actual accuracy on attribution types may vary dramatically: 84% correct on scanning errors, 71% on vendor packing, but only 48% on item-master drift. The decision RDSA supports: whether AI triage of receipt discrepancies can be trusted to dispatch to the right team, or whether storeroom personnel should validate AI attribution before acting.

Formula

RDSA = correctly_attributed_discrepancies / total_attributed_discrepancies

Variables:

  • correctly_attributed = resolution-audit root cause matches the AI's attribution
  • attributed_discrepancies = discrepancies where the AI assigned a confidence-above-threshold attribution
  • Attribution taxonomy: vendor packing variance, item-master drift, scanning error, counterfeit/substitution, transit damage, miscount

Worked example: Last quarter, 412 AI-attributed receipt discrepancies. 218 resolved with root cause matching AI attribution; 134 resolved with different root cause; 60 resolved ambiguously (multiple causes or undetermined). Counting unambiguous cases: RDSA = 218 / (218 + 134) = 0.619. By type: vendor-packing attributions correct 71%; item-master-drift attributions correct 48%; scanning-error attributions correct 84%. The item-master attribution accuracy is the weakest signal.

Edge cases: Requires resolution attribution captured as structured data. Most common pilot blocker. Distinguish wrong attribution from ambiguous-cause cases. Report alongside confidence-rejection rate. If AI marks most discrepancies as ambiguous, RDSA is computed on a subset.

Data sources: AI attribution events from receipt AI platform. Resolution outcomes from discrepancy resolution workflow (often requires instrumentation to capture as structured data). Discrepancy taxonomy must be stable across measurement window.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Monthly
Category
F - Inspection, Provenance & Master Data Integrity
← Counterfeit Mode Coverage Score (CMCS)Receipt Inspection Adequacy Ratio (RIAR) →

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