Classification for Binning is the application of ML classification algorithms to sort finished chips into performance bins — predicting whether a die will be fast, typical, or slow based on inline process measurements, enabling early yield prediction and optimized testing strategies.

How Is It Applied?

• Features: Inline metrology (CD, thickness, overlay), process tool data, wafer position.
• Labels: Final electrical test bin assignments (speed grades, pass/fail).
• Models: Random forests, gradient boosting, neural networks trained on historical data.
• Prediction: Predict bin assignment from inline data before final test — enables sort/test optimization.

Why It Matters

• Test Time Reduction: Pre-classify wafers to focus expensive testing on borderline cases.
• Yield Prediction: Predict yield and bin distribution before wafers reach final test.
• Revenue Optimization: Earlier bin prediction enables better production planning and customer allocation.

Classification for Binning is predicting chip performance from process data — using ML to sort dies before they reach the tester.