First Pass Yield and How to Improve It

What First Pass Yield Measures

First pass yield is the proportion of units that pass all the tests the first time, without a rework or a retest. It differs from the final yield, which counts the units that pass after being reworked, and the difference is the hidden factory: the rework, the retest and the scrap that the final yield conceals. A line with a final yield of ninety eight per cent and a first pass yield of eighty per cent is doing a great deal of rework, and the rework costs money, consumes capacity and introduces the risk of a latent defect. First pass yield is therefore the number that describes the process rather than the outcome.

How It Is Measured

The measurement requires that each unit’s result at each station is recorded and that a failure is attributed to the station where it was found. A unit that fails at two stations should be counted once, at the first station that found it, otherwise the yields multiply to a figure that no station recognises. The measurement also requires that a reworked unit is distinguished from a first pass unit, which means the rework has to be recorded rather than done quietly at the bench. Where the data is incomplete, the yield figure is an average of the visible and the invisible, and the improvement effort has no target.

The Defect Pareto

The improvement starts with the defects, grouped by type and by station, ordered by frequency. The Pareto usually shows that a small number of defect types account for most of the loss, and the effort should be directed at the largest rather than distributed evenly. The types should be described precisely enough to be actionable: insufficient solder on a specific package, a specific placement offset, a specific test failure. A category as broad as solder defect cannot be acted on, while insufficient paste on a fine pitch device can be traced to a stencil, a printer or a paste condition.

Attributing the Cause

Each significant defect type should be traced to its cause, and the trace uses the process data: the machine, the program, the paste lot, the stencil, the profile and the operator. A defect that appears on one machine and not another identifies the machine; one that appears with one paste lot identifies the material; one that appears at a particular time of day identifies a condition. The trace should be systematic and it should use the data that the line already generates, because a hypothesis tested against the records is faster than an experiment. Where the data does not exist, that is itself a finding about the process’s control.

The Improvements That Raise It

The improvements that raise the first pass yield are usually structural rather than heroic. A stencil whose aperture is corrected, a support pin that is moved, a squeegee that is replaced on a schedule, a profile that is re-measured, a feeder that is maintained and a component that is substituted for one with better placement behaviour all remove a defect type permanently. Mistake proofing, such as scanning a reel before it is loaded, removes a whole class of defects rather than a single instance. The improvement should be verified by the yield’s trend after the change, and the change should be recorded so that the gain is not lost when the personnel change.

The Relationship to the Test Strategy

A higher first pass yield changes the test strategy. Where the process is stable and the defect rate is low, the inspection can be sampled rather than complete, and the savings from the reduced inspection can be reinvested. Where the yield is low, adding inspection is the wrong answer, because inspection finds defects rather than preventing them, and the cost of the inspection is added to the cost of the rework. The correct sequence is to improve the process and then to reduce the inspection, and the yield data is the evidence that allows the reduction to be justified.

Recording and the Trend

The yield should be recorded over time and by product, because the trend is more informative than an average. A yield that is falling identifies a change, and a yield that improves after a change confirms the change’s effect. The records should include the rework and the scrap, since those are the costs that the first pass yield represents. Where the yield is reported to the customer or the management, the definition should be stated with the figure, because a first pass yield, a final yield and a rolled throughput yield are different numbers and are often compared as if they were the same.

Yield from the Customer’s View

The customer’s experience of quality is not the line’s first pass yield but the rate of failure after the product is in use, and the two are related in a way that is easy to forget. A unit that is reworked to pass a test may have been subjected to additional heat, handling and stress, so its reliability is not the same as a unit that passed first time. A high rework rate therefore produces a population with a mixed history, some of which is more likely to fail. Improving the first pass yield improves the field reliability as well as the cost, which is the argument that justifies the investment.

FAQ

What is first pass yield? The proportion of units that pass all tests the first time, without rework or retest.

How does it differ from final yield? The final yield includes the reworked units, so it hides the rework, the retest and the risk of a latent defect.

Where should improvement start? With the defect Pareto, directed at the largest types, described precisely enough to be actionable.

How is a defect traced to its cause? Using the process data: the machine, the program, the material lot and the profile.

Should inspection be increased to raise the yield? No. Inspection finds defects; the process change prevents them, and the inspection can then be reduced.

Conclusion

First pass yield describes the process and exposes the hidden factory, so measure it honestly, act on the Pareto and confirm the trend. Improve the process before the inspection. Yield and process control belong to quality management, the line it measures is described in SMT PCB assembly, and the tests it uses are part of PCBA testing. Yield targets for a new product are set during prototype PCB assembly in 2026.