IllustrationAQL sampling inspects a random sample from each lot instead of every unit. Lot size and inspection level give a code letter, the code letter sets the sample size, and the AQL you choose sets how many defective units the sample may contain before the lot is rejected. It decides acceptance and does not measure lot quality.
Most hardware teams meet AQL sampling at the first production lot. The factory or a third-party inspector proposes a plan, the numbers look standard, and nobody asks what the plan can and cannot see. This guide explains how the tables work, how buyers set AQL values by defect class, how likely a bad lot is to pass, and which defects in an electronic device a final AQL check will never find.
This is an engineering summary. The sampling tables come from ANSI/ASQ Z1.4 and ISO 2859-1, which are sold standards. We quote only plan values published on the open web by inspection companies and one quality engineer's blog, each linked below. None of these is a standards body, and we do not reproduce the full tables.
How does AQL sampling work?
AQL sampling is a form of acceptance sampling. The NIST/SEMATECH e-Handbook says acceptance sampling is used when testing is destructive, when 100% inspection costs too much, or when 100% inspection takes too long. Its main purpose is to decide whether a lot is likely to be acceptable. NIST adds that acceptance sampling is not meant to estimate the quality of the lot (as of November 2026).
The NIST handbook defines the AQL as a percent defective that is the baseline requirement for the quality of the producer's product. QIMA, an inspection company, calls it the acceptable (or acceptance) quality limit and says AQL levels are set separately for each defect type: critical, major and minor (both as of November 2026). The idea is the same under either name: the buyer and the factory agree on a baseline, and each lot is accepted or rejected against it.
Where the tables come from
NIST traces the tables to attribute sampling tables built for the US Army and Navy during World War II, combined in 1950 as MIL-STD-105A. MIL-STD-105D was issued in 1963, adopted by ANSI as Z1.4 in 1971 and, with minor changes, by ISO as ISO 2859 in 1974. NIST lists 105E, issued in 1989, as the latest revision, and says the three standards keep being revised while the basic tables remain the same (as of November 2026). QIMA says its own tables are based on ANSI/ASQ Z1.4-2008. When a contract names a standard, name the edition too.
The plan in one sequence
NIST lists the steps for MIL-STD-105E: decide the AQL, decide the inspection level, determine the lot size, read the sample size code letter, choose the type of sampling, read the plan, then start with normal inspection and follow the switching rules. A full scheme has a normal, a tightened and a reduced plan, plus rules for moving between them (as of November 2026).
How do you read an AQL table based on ANSI Z1.4?
Every AQL plan reads two tables. The first turns lot size and inspection level into a code letter. The second turns the code letter and the AQL into a sample size, an accept number and a reject number. NIST notes that the standard gives a sample size code letter, and the letter then leads to the sample size (as of November 2026).
| Term | What it is | Who sets it |
|---|---|---|
| Lot size | The number of units in the batch offered for inspection. | Fixed by the order or the shipment. |
| Inspection level | Sets how sample size grows with lot size. There are three general levels (I, II, III) and four special levels (S-1 to S-4). | The buyer, usually with the inspector. |
| Sample size code letter | A letter read from the first table, from lot size and inspection level. | The table. No one chooses it. |
| Sample size | How many units are pulled at random and checked. | The second table, from the code letter. |
| AQL | The defect level that is the baseline requirement for the producer. Set one value per defect class. | The buyer and the factory, agreed in writing before production. |
| Accept (Ac) and reject (Re) numbers | The most defective units the sample may contain and still pass, and the count that fails the lot. | The second table, from the code letter and the AQL. |
Worked examples from published plans
These three plans are copied from guides by QIMA, an inspection company, and FamiSourcing, a China sourcing firm that also sells quality inspection. All are single sampling at normal inspection. We did not read them from the standard itself. The FamiSourcing example is a lot of 900 T-shirts; the plan values depend only on lot size, inspection level and AQL, so they read the same for a lot of 900 devices.
| Lot size | Level | Code letter | Sample | AQL | Accept | Reject | Source, as of November 2026 |
|---|---|---|---|---|---|---|---|
| 900 units | General II | J | 80 | 2.5 (major) | 5 | 6 | FamiSourcing |
| 900 units | General II | J | 80 | 4.0 (minor) | 7 | 8 | FamiSourcing |
| 4,000 units | General II | L | 200 | 2.5 | 10 | 11 | QIMA |
Read the first row this way. The inspector pulls 80 units at random from the lot of 900. If 5 or fewer have a major defect, the lot passes on majors. If 6 or more do, it fails. Minor defects are counted separately against their own numbers, and the lot must pass on every class.
Which acceptable quality limit should you set for each defect class?
QIMA defines the three classes this way. Critical defects are safety or compliance failures. Major defects affect function or appearance enough to likely cause a return. Minor defects are small cosmetic issues that do not affect function (as of November 2026).
On values, QIMA says AQL 2.5 is the most common standard for general consumer goods, AQL 1.0 is used for higher-risk categories, AQL 4.0 is used when minor cosmetic defects are acceptable, and the AQL for critical defects is typically 0. FamiSourcing's worked example uses 0 for critical, 2.5 for major and 4.0 for minor, and allows no critical defect in the sample (both as of November 2026). QIMA also warns that most buyers copy a setting they have seen elsewhere instead of setting it from their own risk.
For an electronic device, the hard part is the defect list. Write it before the first lot. Typical candidates for critical are anything that touches battery safety, exposed live parts, or a missing regulatory label. Typical majors are a unit that does not power on, a button that does not work, or a missing charger. Typical minors are small scratches or a slightly off-center print. These are examples to adapt, and your list should come from your own product risks.
Which inspection level fits a small electronics batch?
A QualityInspection.org blog post by Renaud Anjoran, an ASQ Certified Quality Engineer, calls General Level II the most widely used level and the default. Level I suits suppliers with a strong record, and the article warns that settling on it by default to spend less on inspections is very risky. Level III suits suppliers with recent problems or high-value products. It quotes ISO 2859-1 that special levels may be used where relatively small sample sizes are necessary and larger sampling risks can be tolerated, and says they are typically used for tests that take a long time or destroy the samples (as of November 2026).
The same article gives the sample sizes for two lot sizes at each level. The figures show that the AQL stays fixed while the level changes how many units are pulled.
| Inspection level | Sample, lot of 5,000 | Sample, lot of 40,000 |
|---|---|---|
| S-1 | 5 | 8 |
| S-2 | 8 | 13 |
| S-3 | 20 | 32 |
| S-4 | 32 | 80 |
| General I | 80 | 200 |
| General II | 200 | 500 |
| General III | 315 | 800 |
Source: QualityInspection.org, How the AQL inspection levels in ISO 2859-1 affect sampling size, as of November 2026.
The sample does not grow in step with the lot. At Level II, 200 units is 4% of a lot of 5,000, while 500 units is 1.25% of a lot of 40,000. The FamiSourcing plan pulls 80 of 900, close to 9%. Small pilot lots therefore carry a high inspection share. The article notes that Level III can also suit small quantities, where the inspection takes only one day whatever level is chosen.
How likely is a bad lot to pass an AQL check?
Every plan has an operating characteristic (OC) curve. NIST defines it as the probability of accepting the lot plotted against the lot's percent defective. It also defines two risks: the producer's risk of rejecting a lot at the AQL, and the consumer's risk of accepting a lot at a much worse level the consumer would not accept (as of November 2026).
We calculated the acceptance chance for the two AQL 2.5 plans above, using the exact hypergeometric formula for the lot sizes given. The math is ours, the plans are from the linked guides.
| Share of the lot that is defective | Lot of 900, sample 80, accept 5 | Lot of 4,000, sample 200, accept 10 |
|---|---|---|
| About 2.5% (at the AQL) | 98.8 in 100 (23 bad units) | 99.0 in 100 (100 bad units) |
| 5% | 79.7 in 100 (45 bad units) | 58.3 in 100 (200 bad units) |
| 8% | 36.5 in 100 (72 bad units) | 6.4 in 100 (320 bad units) |
| 10% | 16.5 in 100 (90 bad units) | 0.7 in 100 (400 bad units) |
Two points stand out. A lot at twice the AQL still passes most of the time under both plans. And the smaller plan protects the buyer less: a lot with 8% major defects passes about one time in three. If that risk is too high for your product, raise the inspection level, lower the AQL, or test every unit for the defects that matter most.
What does AQL sampling miss on electronic devices?
A final AQL inspection checks a sample of packed goods against a checklist. It is good at what an inspector can see and try in a few minutes per unit. Many electronics failures are outside that window. The table is our engineering view of where each kind of defect is caught.
| Defect | Caught by a final AQL check? | Where to catch it |
|---|---|---|
| Cosmetic marks, wrong label, missing accessory, damaged box | Yes. This is what a final visual AQL check is built for. | Define each one as major or minor in the inspection checklist. |
| Unit does not power on or fails a basic function | Only if the checklist includes a function test on the sampled units. | A functional test on every unit at the end of the line, with the results logged. |
| Wrong firmware version or wrong configuration | Only if the inspector reads the version on each sampled unit. | The test station reads and logs the firmware version and device ID of every unit. |
| RF output, sensor accuracy, audio quality | Rarely. These need fixtures and instruments the inspector usually does not carry. | Production test limits set during development, measured on the line. |
| Intermittent faults, such as a reset that shows up once a day | No. A short check on a sample cannot see a fault that appears hours later. | Burn-in or soak testing on a defined share of units, set by your own failure data. |
| Latent solder or assembly defects | No. They pass at the factory and fail in the field. | Process controls at assembly and reliability tests during development. |
| Battery life and battery safety | Only basic checks, such as swelling or charge state. | Incoming checks on cells and a charge and discharge test at the line. |
The pattern is simple. Anything that can kill or mislabel a unit should be tested on every unit at the line, where the cost per test is low. The final AQL check then confirms the packed product. For the line tests themselves, see in-circuit test vs functional test. Before the first lot, a first article inspection confirms that the production build matches the drawings, which no sampling plan does.
When a lot fails
NIST describes a common procedure when sampling and testing are non-destructive: inspect rejected lots 100% and replace every defective unit with a good one. QIMA says its inspectors document each defect with photos, the buyer gets a corrective action report for the supplier, and a re-inspection confirms the lot after corrections (both as of November 2026). Agree in the purchase order who pays for sorting and re-inspection.
What should you ask your hardware team?
Bring these questions to the review of the quality plan, before the first production lot is built.
- Which standard and which edition does our inspection plan cite, and do we own a copy of it?
- What inspection level and what AQL do we use for critical, major and minor defects, and where are they written into the purchase order?
- Which defects in our checklist are critical, and does one critical defect in the sample reject the lot?
- Does the final inspection include a power-on and function test on every sampled unit, or only a visual check?
- Which checks run on every unit at the line, and where are those logs stored?
- Does the inspector read the firmware version of each sampled unit?
- What happens to a rejected lot: full sorting, rework, re-inspection, or all three?
- Who pays for a re-inspection after a failed lot?
For the checks that run from incoming parts to packing, see quality checks from parts to packing.
Questions about AQL sampling
What is the difference between AQL and the defect rate of my lot?
The AQL is the baseline quality you agreed with the factory. The NIST handbook says the main purpose of acceptance sampling is to decide whether the lot is likely to be acceptable. It is not meant to estimate the lot's quality (NIST/SEMATECH e-Handbook, as of November 2026). A lot that passes at AQL 2.5 can still contain more than 2.5% defective units.
Is ANSI/ASQ Z1.4 the same as ISO 2859-1?
They come from the same root. NIST says MIL-STD-105D was adopted by ANSI as Z1.4 in 1971 and, with minor changes, by ISO as ISO 2859 in 1974. It adds that the three standards keep being revised but the basic tables remain the same (source, as of November 2026). We have not compared the current editions line by line, so confirm with the edition your contract cites.
What AQL should I use for an electronic device?
No source we read gives one value for electronics. QIMA says AQL 2.5 is the most common setting for general consumer goods, AQL 1.0 is used for higher-risk categories, and the AQL for critical defects is typically 0 (QIMA, as of November 2026). Choose from your own risk and return costs, and write the values into the purchase order.
Can I skip AQL inspection if every unit is tested on the line?
Line tests and a final AQL check look at different things. A line test measures function on every unit. A final AQL check looks at the packed product the way a customer receives it: labels, accessories, cosmetics and packing. Keeping both covers both kinds of defect, and the line test logs are part of the evidence a lot is ready.
How can Duobay help with a quality plan for my device?
Duobay develops custom devices with a Shenzhen engineering team and builds samples and small batches of 10–1,000 units. The quality plan for your product sets which checks apply and how many units are sampled. We do not publish prices, and initial calls are free. Tell us about your device.
If you want a quality plan and a defect list written for your device before the first lot, tell us about your device.
Sources. All pages accessed November 6, 2026. The QIMA, QualityInspection.org and FamiSourcing pages are industry guides that summarize sold standards. Confirm plan values against the edition of ANSI/ASQ Z1.4 or ISO 2859-1 your contract cites.
- NIST/SEMATECH e-Handbook of Statistical Methods, 6.2.1 What is Acceptance Sampling? (link)
- NIST/SEMATECH e-Handbook of Statistical Methods, 6.2.2 What kinds of Lot Acceptance Sampling Plans are there? (link)
- NIST/SEMATECH e-Handbook of Statistical Methods, 6.2.3.1 Choosing a Sampling Plan: MIL Standard 105D (link)
- QIMA, Acceptable Quality Limit (AQL), tables based on ANSI/ASQ Z1.4-2008 (link)
- QualityInspection.org, How the AQL Inspection Levels in ISO 2859-1 Affect Sampling Size (published June 16, 2021, updated January 2026) (link)
- FamiSourcing, AQL Sampling 101: Meaning, Tables, Levels for Inspection (link)
Planning a device like this? Tell us what it needs to do. A Duobay engineer replies within 1 business day. Send a project brief →