Industrial AI planning systems combine predictive models, optimization logic, controls and governance rules. An improvement in the overall policy does not reveal whether a single addition made an independent contribution or whether that contribution justifies its burden. To address this attribution problem we propose an observable evidence standard that separates opportunity, executed‑plan change, operational consequence, equivalence, uncertainty and operating scope.
Using forecast‑to‑plan replay on Favorita, M5 and Walmart data we applied the standard to heterogeneous planning additions. Three limits of single‑indicator evaluation emerged. First, a composite policy can hide a negligible component contribution: one control altered only 15 of 681 executed plans. Second, a change in execution does not guarantee stable benefit: the evaluated multi‑period capability helped in several long‑lead settings but was 7.6% more costly under tight capacity. Third, indirect evidence can prioritize review but cannot complete attribution: independent runs produced different paths in 19 of 599 cases that a rule‑based check predicted would be unchanged.
A lower‑cost diagnostic matched the full‑evaluation winner in only 55.0% of cases, yet selected an option within a $10^{-6}$ cost‑equivalence tolerance in 79.5%; the remaining cases prevented the proxy from authorizing acceptance.
These experiments establish capability‑level contribution evidence; the organization completes the value decision by weighing it against implementation and sustaining burden. The capability evidence record links observations to validation, deployment scope, fallback, monitoring, redesign and retirement decisions.
Review