Definition
No existing KPI measures generative design manufacturability failures specifically.
The Measurement Gap
The decision DMER supports is whether the generative design system's manufacturing constraint model is complete enough to produce designs that work in production and which constraint categories are missing or incomplete. The measurement gap exists because engineering change orders (ECOs) are tracked in PLM systems, but root cause classification rarely distinguishes between design errors, requirement changes, and manufacturing constraint gaps. When an AI-generated design fails in manufacturing, the ECO closes the immediate problem, but nobody aggregates whether AI designs fail more often than human designs or whether the failures cluster around specific constraint types. The invisible failure mode is systematic constraint incompleteness. Generative design systems optimize within their constraint sets. If the constraint set is missing fixture access requirements, tolerance stack-up rules, or process capability limits, the AI will repeatedly produce designs that violate those missing constraints. Each violation generates an ECO, but without DMER, nobody sees the pattern or updates the AI's constraint model to prevent recurrence.
Formula
Of all designs released to production that originated from generative design, count how many required unplanned ECOs due to manufacturing feasibility issues. The formulas are: DMER_AI = ECOs_manufacturing_AI / Designs_released_AI times 100. DMER_Human = ECOs_manufacturing_Human / Designs_released_Human times 100. DMER_Delta = DMER_AI minus DMER_Human.
Variables: ECOs_manufacturing_AI is the count of ECOs triggered by manufacturing feasibility issues on AI-generated designs, per period. Designs_released_AI is the count of AI-generated designs released to production in the period. ECOs_manufacturing_Human and Designs_released_Human are the same counts for human-generated designs.
Sign convention: DMER_Delta above 0 means AI designs fail in manufacturing more often than human designs. Below 0 means AI designs are more manufacturable. Track the absolute DMER_AI regardless of comparison.
Edge cases: not all ECOs on AI-generated designs are caused by the AI's constraint gap. Only count ECOs where the root cause is design not manufacturable as specified. For mixed designs (human-modified AI output), attribute based on whether the non-manufacturable feature was AI-generated or human-modified. Minimum sample: 10 AI-generated design releases per measurement period.
Proposed formula, pending validation pilots.