AQL Sampling Plans for Final PCB Inspection

A sampling plan is a statistical agreement about how much defect is tolerable in a lot, and it is often applied to electronics as if it were a checklist. The AQL value defines the quality level that will be accepted most of the time, and understanding what the plan does and does not guarantee is the difference between using it well and using it to certify a bad lot.

What an AQL Plan Does and Does Not Do

The plan takes a sample from a lot, counts the defects found and compares the count with an acceptance number. If the count is at or below it, the lot is accepted; above it, the lot is rejected. What the plan guarantees is a defined risk of accepting a lot at a stated quality level, not an absence of defects in the accepted lot.

That distinction matters because acceptance is not verification. A lot accepted at an AQL of 1.0 percent can contain appreciable defect content, and it will do so by design. Where a defect is safety-related or would be difficult to detect in service, sampling is the wrong tool and a different approach is needed.

Choosing the Inspection Level

The general inspection levels in the widely used standards use roman numerals, with level two as the default and levels one and three for reduced or increased discrimination. Within the general levels the same AQL produces different sample sizes, so the choice of level is as important as the choice of AQL.

The special levels are used where a specific characteristic needs more inspection than the general plan provides, for example a critical dimension on a connector. Moving to a special level for one characteristic rather than raising the level for the whole product keeps the inspection effort proportionate.

Classifying Defects by Severity

Defects are normally classified as critical, major and minor, with different acceptance levels for each. A critical defect is one that could cause harm or a regulatory failure and is usually given a zero acceptance number, while a major defect affects function and a minor defect affects appearance. The classification is a product decision that has to be made once and applied consistently.

Misclassification is a common weakness. If inspectors disagree about whether a scratch is major or minor, the acceptance number applied to it is arbitrary, and the same lot can be accepted or rejected depending on who inspected it. Written criteria with photographs resolve most of this before the plan is used.

<img src="https://www.gopcba.com/wp-content/uploads/2026/06/Juniper-Networks.png" alt="Inspector working from an AQL sampling plan at a final inspection bench” />

Sample Size and Acceptance Numbers

The sample size is derived from the lot size and the inspection level, and the acceptance number from the sample size and the AQL. For a lot of 1200 units at level two, the sample is often around 80 units, with an acceptance number of two or three at an AQL of 1.0 percent for major defects. Reading the number from the table rather than choosing it is what keeps the plan statistically valid.

It also follows that small lots produce small samples with very low acceptance numbers. A lot of 50 units may require a sample of eight with zero acceptance, which means a single defect rejects the lot. That outcome is a property of the plan rather than an accident, and it should be understood before it surprises a production team.

Sampling a Product That Is Mostly One Unit

Electronics inspection often involves characteristics rather than identical units: a board may have hundreds of joints, and the sample is usually boards rather than joints. That means the defect rate measured is a rate per board, which can be dominated by a single systematic defect or by many small ones, and the two have very different implications.

Where a systematic defect exists, sampling will detect it with high probability even at a low rate per board, which is why sampling works well for process problems. Where the defects are random and rare, it takes a large sample to find them, and the sample size may not be economically available.

Switching Rules and Tightened Inspection

The standards define switching rules that move a supplier between normal, tightened and reduced inspection depending on the recent history. Two consecutive rejections normally trigger tightened inspection, which uses a larger sample or a lower acceptance number, and a sustained record of acceptance allows reduced inspection.

Applying the switching rules is one of the most valuable parts of the system and one of the most frequently skipped. A supplier that has been on tightened inspection for months without a formal review, or one that has never moved to reduced inspection despite a clean record, indicates that the plan is being applied mechanically rather than managed.

Sampling Versus 100 Percent Automated Inspection

Where AOI or X-ray covers a characteristic completely, sampling becomes a check on the inspection process rather than on the product. That is a different question, and it is answered by a periodic verification that the machine still detects a known defect rather than by sampling the boards.

The two approaches are complementary. Automated inspection screens every unit for the defects it is programmed to find, while sampling audits the process and covers characteristics the automation cannot see. Confusing them leads either to sampling for something already fully inspected or to assuming that automation has replaced the need for an audit.

Records and Traceability of Sampling

A sampling record should show the lot, the sample size, the number and type of defects found, the acceptance number applied and the disposition. Without the sample size the defect count is meaningless, and without the classification the acceptance number cannot be checked.

Where a lot is rejected, the record should also show what was done with it: sorted, reworked, returned or scrapped. The disposition is the part that closes the loop, because a rejected lot that is quietly sorted and shipped leaves no evidence of the process problem that produced it.

Where Sampling Fails

Sampling fails when the lot is not homogeneous. If a defect is confined to one panel position, one feeder or one shift, a random sample may miss it entirely, and the acceptance decision says nothing about the population. Where a process change has occurred within the lot, the lot should be split rather than sampled as one.

It also fails when the inspection itself is unreliable. A sample inspected by a method with poor repeatability produces a number that looks like a statistical result but reflects the measurement system, which is the case for judging appearance criteria that two inspectors apply differently. Establishing agreement first makes the sampling result meaningful.

Related reading: our fabrication notes, board quality and design release notes cover the same ground.

FAQ

What does an AQL value actually mean? It is the quality level that the plan accepts most of the time. It is not a guarantee that accepted lots contain no defects above that level.

Which inspection level should be used by default? General inspection level two is the usual default. Special levels are applied to specific characteristics that need tighter discrimination.

Why does a small lot sometimes reject on a single defect? Because the sample size is small and the acceptance number derived from the table is often zero. That is the plan behaving as designed, not an error.