Yield Predictor Console

A die that ships has survived three separate gates, and total yield is the fraction that makes it through all of them. First it must dodge the random defects scattered across the wafer. Then it must be free of systematic problems — the design-process interactions that affect every wafer the same way. Finally it must meet its parametric specs: fast enough, cool enough, within leakage. Miss any one and the die is lost, which is why the three yields multiply rather than add.

That multiplication is the insight people most often get wrong. Ninety percent in each of three components is not ninety percent overall — it's seventy-three. Individually-respectable numbers compound into a meaningful total loss, so a high overall yield demands that all three be genuinely good, not just on average. This console makes the compounding explicit and shows the total falling out of the product.

The components also fail in different, invisible ways. Random loss shows on a defect map. Systematic loss repeats wafer to wafer at the same locations. But parametric loss is invisible until electrical test — a perfectly clean die that simply runs too slow or too hot for its bin, a victim of process variation against the design margins. Treating all yield loss as ‘defects’ misses this entirely, which is why separating the components matters.

Most valuable of all is the attribution. Knowing whether you're losing dies to defects, design, or binning tells you where improvement actually pays — a cleaner process, better design-for-manufacturing, or tighter process control are three different investments. The loss Pareto here is the roadmap. Drill the random term in the Defect Density console and the parametric term in the Process Variation console.