First-pass yield

Of 200 brackets entering one operation under the same acceptance rule, 180 pass first time, 15 need rework and 5 are scrapped immediately. Rework recovers 12, so first-pass yield is 90% and final acceptable yield is 96% with 192 accepted. Both divide by the original 200 units. If each reworked unit is attempted once more, 215 attempts measure activity, while the unit denominator stays 200.

ASQ excludes units scrapped, rerun, retested, returned or diverted to offline repair from first-pass yield. ASQ quality glossary, “First pass yield”

Define the boundary before calculating

For a fully resolved cohort:

FPY = first-pass acceptable units / units entering the defined process × 100%.

State the operation or route, revision, acceptance criteria and cohort. Count unique entering units.

For a practical record, preserve the unit or lot identity, first entry, first disposition, any recovery route and final disposition. Record a failure when it first occurs; do not overwrite it with the later successful outcome. If serial tracking is unavailable, reconcile quantities at the selected batch boundary and state that individual histories are unavailable.

A repeated check after an initial failed acceptance result must not become an apparent first-pass success merely because the latest result passes. Conversely, a prescribed sequence of multiple normal checks is not automatically rework: define the required first-pass route before observing results. Distinguish invalid measurement events and corrected data-entry errors from actual product nonconformance using documented disposition rules.

Keep defects, defective units and recovery separate

A unit can contain multiple defects. Minitab distinguishes a defect in a quality characteristic from a defective unit representing the overall item; its attribute-chart guidance separates charts for defective units from charts for defect counts. Count a unit once in the FPY failure numerator even if inspection identifies three different problems. Keep the defect detail separately for investigation. Minitab attribute control-chart guidance

For planning, report at least first-pass good, recovered good, scrap and unresolved quantities. These categories prevent a high final yield from hiding the capacity consumed by recovery. They also prevent “100% minus scrap rate” from being mislabeled FPY when some units needed correction.

Vorne's OEE guidance similarly counts units needing rework as quality rejects on their first run, making OEE quality similar to FPY. It warns that subsequent rework runs complicate counting. A machine-level OEE result is therefore not automatically the FPY of an entire manufacturing route: inspect the equipment boundary and how repeat runs enter its counts. Vorne OEE FAQ, reworked parts

Original example: first-pass and final output

Assume a closed cohort of 200 fabricated brackets, each entering the same operation once initially. The acceptance rule is unchanged throughout the batch. On the first attempt, 180 pass, 15 require rework and 5 are scrapped. Of the 15 reworked units, 12 subsequently pass and 3 are scrapped. No units remain unresolved.

The quantities reconcile: 180 first-pass good + 12 recovered good + 8 total scrap = 200. The six-percentage-point gap between 96% and 90% represents the 12 recovered units relative to original starts. It is not the percentage that entered rework: three attempted recoveries failed.

If each of the 15 units receives exactly one additional processing attempt, the operation records 215 attempts. Dividing 192 final good units by 215 attempts gives about 89.30%, a different ratio with a different denominator. Calling it the batch FPY would conceal the counting change. This example assumes all recovery stays within the stated cohort and excludes replacement parts.

Rolled throughput yield and unit histories

Minitab describes rolled throughput yield (RTY) as the probability of traversing the complete process without defects, calculated by multiplying step throughput yields that exclude rework and scrap. Its example follows a progressively reduced first-pass population through successive stages. Preserve unrounded counts when reproducing such a calculation. Minitab throughput and rolled throughput yield

For an original no-recovery example, 1000 units enter step A, 950 pass it first time, and only those 950 continue to step B. If 900 pass B first time, the conditional yields are 950/1000 and 900/950. Their product is exactly 900/1000 = 90%. No independence assumption is required for that linked, conditional-count calculation.

Now consider a different process: all 100 units reach both stations after any needed correction. Ten fail A first time and ten fail B first time, so each station reports 90% FPY. If the same ten units fail both, 90 units pass the whole route without correction. If completely different groups fail, only 80 units do. The product 0.9 × 0.9 = 81% does not recover either actual result from these marginal station rates.

Thus a product of marginal station yields is a model requiring suitable assumptions about dependence and populations. To measure actual end-to-end first-pass performance, count units that passed every required stage without recovery, or use consistently conditioned survivor counts. Do not present a modeled RTY as an observed unique-unit count.

Aggregation and reporting limits

Combine comparable cohorts by summing first-pass successes and starts, not by averaging their percentages equally. For example, 9/10 and 800/1000 yield a combined 809/1010 = 80.10%, rather than the unweighted average of 90% and 80%, which is 85%. Keep different products or routes separate when aggregation would hide a material mix change.

A shift's starts and completions may belong to different cohorts. Units still awaiting inspection cannot be silently counted as failures or successes. State the cutoff, pending quantity and whether the report follows a starting cohort to resolution or records completed dispositions. A zero-start denominator gives no yield estimate, not 0% or 100%.

For trend monitoring, Minitab's P chart is designed for proportions of units classified into two categories. Appropriate subgrouping and statistical assumptions still matter; a dashboard target is not a statistical control limit. Minitab P chart overview

FPY does not establish inspection effectiveness, escaped-defect risk, delivered reliability or economic loss by itself. Pair it with defect modes, recovery time, scrap and the relevant acceptance evidence. The cited sources are undated online definitions reviewed on 2026-11-01; no publication edition is inferred from their website footer years.

Related reference

Production planning guides