Definition
The percent of AI-recommended part substitutions that meet or exceed the engineering-expected service life of the original part, rather than deferring an underlying fit-for-use failure that surfaces as a premature asset failure attributed to other causes.
The Measurement Gap
During a stockout on a critical pump seal, the AI substitution recommender suggests an alternative with 94% confidence based on dimensional match and material compatibility. The planner accepts under pressure: the asset is down and the original part has a 12-week lead time. The substitute is installed. Eight months later the seal fails prematurely. The failure is attributed to the asset, not to the substitution decision. The maintenance team replaces the seal again, this time with the original part. Nobody connects the premature failure back to the AI recommendation that put the wrong part on the asset. This is not a one-time event. Across a large MRO operation, substitutions happen dozens or hundreds of times per quarter, driven by stockouts, obsolescence transitions, and cost optimization. Each substitution is a bet that the alternate part will perform as well as the original. Fill rate does not distinguish between an issue of the correct part and an issue of a substitute. Inventory accuracy does not track whether the substitute matched the engineering fitness-for-use requirement. Asset reliability metrics see the premature failure but have no attribution path back to the substitution decision. The substitution recommender's confidence score measures dimensional and material similarity, not actual service-life outcome, and the gap between similarity and fitness is where failures hide.
Decision Supported:
Whether to accept AI-recommended substitutions automatically, require engineering review by criticality class, or feed substitution outcomes back to retrain the recommender.
Formula
Plain language: Track every substitution. When the substituted part fails or is removed, compare its service life to what the original part would have delivered. The fraction meeting expected service life is the yield.
Notation:
SFY = parts_meeting_service_life / total_substitutions_with_known_outcome
Variables:
| Variable | Definition | Unit |
|---|---|---|
| parts_meeting_service_life | Substituted parts with service life greater than or equal to 0.9 times original-part expected MTBF | Count |
| total_substitutions_with_known_outcome | Substitution events where the part has reached end of service or failure within the outcome window | Count |
| MTBF | Mean time between failures for the original part (from OEM spec or reliability data) | Hours or days |
| Outcome window | Expected service life of original part plus 30 days | Time |
Proposed formula, pending validation pilots.