First Pass Yield: Definition, Formula and Why It Compounds

The formula is simple enough to fit in a sentence. What it does when you chain it across a routing is the part most quality reports never show.

The formula, and what actually gets counted

First pass yield is units passed without rework, divided by units started. A unit that fails, gets reworked, and passes on the retest does not count as a first-pass pass. That distinction is the whole point of the metric: it separates units that were built correctly from units that were eventually made correct.

Two related numbers get confused with it constantly. Throughput yield counts a pass at a single step regardless of whether rework happened first, so it hides the labor and material that went into getting there. Rolled throughput yield (RTY) goes the other direction: it multiplies FPY across every step in a routing, because a unit only has a true first-pass success if it passed every station without rework, not just the last one.

Why a 95% line is an 81% line

This is where the metric earns its keep. A line with four steps, each running a respectable 95% first pass yield, does not produce 95% good units. It produces 81%.

The math is multiplication, not averaging: 0.95 × 0.95 × 0.95 × 0.95 = 0.8145. Every additional step compounds the loss instead of diluting it. Extend the same line to ten steps, still at a "good" 95% each, and rolled throughput yield drops to roughly 60%, meaning close to 4 in 10 units need rework or get scrapped by the time they reach the end, even though every individual station looked fine on its own report.

This is why a plant with acceptable-looking station yields can still have a rework problem nobody can find: no single station is bad enough to flag, but the routing has enough steps that the compounding does the damage anyway.

Why final-inspection yield lies to you

Final inspection measures what leaves the line after every rework loop has already run. A plant can report 99% at final inspection while running a mediocre first pass yield, because rework absorbed the difference before the unit ever reached the inspector. That 99% is real, but it's measuring the wrong thing: it tells you what shipped, not what it cost to get there in labor, scrap, and schedule risk.

Catching defects in process, at the station where they happen, is what actually moves first pass yield, because it removes the rework loop instead of feeding it. Zaleco cut its scrap rate 20% and raised operational availability 30% within two months by moving quality control in-process instead of catching problems downstream. Faurecia's multimodal AI quality control at Yancheng cut scrap cost 75.8% and customer complaints 94% the same way: by catching the failure closer to where it originates instead of at the far end of the routing.

There's a second, less obvious way in-process inspection can hurt first pass yield: false calls. Siemens Rastatt's AOI machines were flagging good boards as defective often enough to cause alarm fatigue, and operators responding to constant false alarms started waving through boards that actually were bad. Adding AI-powered false-call reduction on top of the existing AOI machines, filtering the noise rather than replacing the inspection, improved first pass yield 42% with ROI in 8 months. The lesson generalizes: an inspection system that cries wolf doesn't just waste time, it trains the people watching it to stop trusting real alarms. Both of these are software doing what computer-vision inspection is actually for: adding a check at the step, not just at the door.

Worked example: what a 65% rework cut means for FPY

Dana reduced axle rework 65% using ML-driven root cause analysis from Acerta, landing at a 4% resulting rework rate. That single pair of numbers lets you back out the starting point: if 4% is what's left after a 65% reduction, the original rework rate was 4% ÷ (1 − 0.65) ≈ 11.4%. First pass yield moved from roughly 88.6% to roughly 96%, an almost eight-point swing, and it came from finding the actual cause of the rework rather than reworking faster.

That 65% sits above the Quality Control & Inspection benchmark median for defect and scrap reduction across our corpus: 51% (6%-75.8%, n=14). Faurecia's 75.8% scrap cost cut above is the top of that same range. Reported inspection accuracy across the deployments that measure it lands at a 99% median (97%-99.9%, n=3), which is the ceiling most in-process systems are chasing: catching essentially everything, at the step, before it ever reaches a rework queue.

Where to go next

• The quality control use cases and benchmarks pages break down outcomes across the full 125-deployment corpus.
• Vendors building the inspection layer that moves FPY are ranked on the quality inspection software hub.
• Arkansas Steel improved process yield from 86% to 90%, a smaller single-digit move that still represents a meaningful cut in scrap at that volume.
• Owlet saved $953K a year and validated root cause 8x faster after shortening the loop between a failure and knowing why it happened.