FPYFirst Pass Yield

FPY for a single step is the count of units that pass on the first attempt divided by the count of units that started. Final yield includes units that passed after rework; it can be high even when FPY is low, and the gap is the visible cost of poor quality. Rolled Throughput Yield (RTY) is the product of FPY across every step in the process — RTY = FPY₁ × FPY₂ × … × FPYₙ. A 10-step process with 95% FPY per step has an RTY of 0.95¹⁰ ≈ 60%; the same process at 99% per step has RTY 90%. RTY is brutal to look at and that is exactly why it is the right metric: it exposes the compounding cost of acceptable-looking single-step performance.

• FPY (single step) = units passing first time ÷ units started.
• RTY = FPY₁ × FPY₂ × … × FPYₙ.
• DPU (defects per unit) = total defects ÷ units; FPY ≈ e^(−DPU) for low-defect processes.
• DPMO (defects per million opportunities) — scales to sigma level: DPMO 3.4 → 6σ; DPMO 233 → 5σ; DPMO 6,210 → 4σ; DPMO 66,807 → 3σ.
• Hidden factory = (units final-passed − units first-pass-passed) ÷ units final-passed — the proportion of output that required rework to ship.

Final yield says 'we shipped 99.5% of what we started'. FPY adds 'and 12% needed rework to get there'. The 12% is the hidden factory: extra labour, extra inspection, extra documentation, extra equipment hours, extra opportunity to introduce error during the rework, extra deviation handling — all consumed to recover units that should have passed first time. In a regulated environment the hidden factory is also a compliance load: every rework is a deviation, every deviation needs investigation and disposition, every disposition needs QA sign-off, and the rework path itself usually requires validation. A high final yield with a low FPY is a process operating on QA heroics, not on capability.

• Process capability (Cpk / Ppk) — a Cpk of 1.33 corresponds to ~63 ppm out-of-spec; 1.67 to ~0.6 ppm; 1.00 to ~2,700 ppm. FPY drops in lockstep with capability.
• Material variation — incoming-lot variability propagates straight into output variability; supplier consolidation usually lifts FPY.
• Equipment drift — calibration slipping, tooling wear, sensor noise; SPC catches drift before FPY drops materially.
• Operator variability — training, fatigue, shift handover; standard work and poka-yoke reduce reproducibility loss.
• Environmental conditions — temperature, humidity, particulate, vibration; especially material in sterile / electronics / coating processes.
• Specification tightness — a spec set tighter than capability supports guarantees FPY problems; capability and spec must be designed together (QbD).
• Detection placement — moving inspection upstream (catch the defect at the step that made it, not after three more steps add value) lifts both FPY and the meaningful repair cost.

In pharma, food and supplements the term FPY is less common — the equivalent at the batch level is Right First Time (RFT): the percentage of batches that pass every IPC test and every release test without OOS investigation, rework, retest or QA-disposition deviation. RFT is the Annual Product Review's most-watched lifecycle metric and a leading indicator of the next quarter's deviation backlog. A drop in RFT often precedes a Stage-3 CPV signal by 2-3 quarters because deviations accumulate before they trigger a Ppk excursion. In device manufacturing (21 CFR 820 / ISO 13485) FPY at the unit level remains the cleaner metric because the production unit is a discrete device, not a 100-kg batch.

1. Reporting final yield as FPY — masking the entire hidden factory.
2. Counting rework as 'passed' rather than as a separate rework lot — same hidden-factory blindness.
3. FPY measured at the end-of-line only, so per-step problems are invisible — measure at every step to find the rework concentration.
4. Outliers excluded as 'test errors' without an OOS investigation — selectively boosting FPY by reclassifying inconvenient failures.
5. FPY denominator inconsistent — units started one shift, scaled to a daily / weekly number with no documented rule.
6. Process change celebrated for raising final yield while FPY actually dropped — the change is doing more damage and being rescued by more rework.
7. FPY trended without the corresponding sigma-level — a 99.5% FPY sounds great until you realise it's 5,000 ppm defects, ~3.5σ, and an FDA quality-metrics submission expects better.

• Per-step FPY captured at every operation as units / batches transition; the rework path is its own counted event, never merged into 'passed'.
• RTY computed nightly per production line per product per shift; the hidden-factory percentage is shown alongside.
• Pharma / supplement / food customers see RFT at the batch level with the same engine — every IPC failure, every OOS, every deviation tracks back to the batch and to the underlying step.
• Device customers see FPY at the unit level, with the DHR carrying the per-step pass/fail record.
• Sigma level and DPMO computed automatically and displayed in the same dashboard as the Stage-3 Pp/Ppk trend.
• Drop-detection on FPY triggers a Stage-3 review work item before the Ppk excursion arrives.
• FPY rolled into the OEE Quality element so OEE and FPY can never disagree.