Skip to main content
PrismaWerk
Back to case studies
QualitySemiconductor

Defect Classification Evaluation Framework

A production-aligned evaluation framework that scored defect classifiers by operational risk instead of only standard ML metrics.

Semiconductor wafer inspection scene representing defect classification and quality evaluation.
Scoring
Risk-Aware
Weighted Model
Cost
Ready
Fab

Client context

A structured evaluation framework developed to assess the real-world performance of defect classification systems.

It integrates statistical model evaluation with operational cost modeling, emphasizing the practical consequences of misclassifications in a high-volume fab environment.

Problem

  • +Operational misalignment: Traditional ML metrics do not reflect the real cost of misclassified defects.
  • +High-stakes errors: Misclassifications can trigger unnecessary tool interventions or allow critical excursions to pass.

Approach

  • +Custom scoring function: Incorporates defect severity and classification confidence.
  • +Comparative Analysis: Benchmarks models based on operational impact (yield loss risk) vs just technical metrics.
  • +Expert Disagreement: Includes weighting for ambiguous ground truth labels.

Measured result

  • +Improved model selection: Selected models with better production reliability than standard accuracy metrics.
  • +Risk-aligned insights: Revealed cost variance between models with similar F1 scores.

Next step

Have a similar operational constraint?

Bring the current baseline, target, and available data. PrismaWerk will help define the first measurable pilot.

Book a strategy callExplore demos