AQL Inspection for Garments: Sampling Table, Defect Classes and Contract Terms

How does AQL inspection work?

The method rests on four concepts: lot, sample, acceptance number (Ac) and rejection number (Re). The products to be inspected are defined as a lot, the sample size (n) is looked up in the table, that many pieces are drawn at random and inspected, and the defective pieces are counted. If the count is equal to or below the acceptance number the lot is accepted; if it is equal to or above the rejection number it is rejected. In single sampling plans under normal and tightened inspection the rejection number is the acceptance number plus one, so there is no grey zone. Under reduced inspection there can be a gap between the two numbers; if the result falls in the gap the lot is accepted, but normal inspection is restored from the next lot (MIL-STD-105E, paragraph 4.10.1.4).

According to the standard, a lot should consist, as far as possible, of products of the same type, grade, size and composition, made under the same conditions and at the same time (MIL-STD-105E, paragraph 4.3). In garments it is the buyer who decides whether a lot is defined per order or per style and colour; write it down in advance.

What AQL means, and what it does not

The standard defines AQL, for a continuing series of lots, as the limit of a satisfactory process average (MIL-STD-105E, paragraph 3.1). Three consequences follow:

• AQL is not a right to ship defective products. The standard states explicitly that choosing an AQL does not give the supplier the right to knowingly deliver defective units (MIL-STD-105E, paragraph 4.4.2).
• AQL alone does not tell you a lot's chance of acceptance. For that you need the plan's operating characteristic (OC) curve (MIL-STD-105E, note to paragraph 3.1). An example from such a curve follows below.
• AQL plans are designed for a continuing series of lots. MIL-STD-105E defines AQL for "a continuing series of lots" (paragraph 3.1). The introduction to ISO 2859-1:2026 likewise says the standard is designed for a continuing series of lots from the same producer and the same process, and that where the producer or process differs it should be applied to each separately.

The standard: ISO 2859-1 and the 2026 edition

AQL plans are in ISO 2859-1; the US counterpart is ANSI/ASQ Z1.4 (Wave PLM). The abbreviation stands for "acceptance quality limit" in the ISO title and for "acceptable quality level" in the US military standard MIL-STD-105E (1989); both expansions are still in use.

The third edition of the standard was published in January 2026; the 1999 edition, its 2001 corrigendum (Cor.1) and its 2011 amendment (Amd.1) were withdrawn on the same date (EVS standard record). As of the date of this article, that is the edition in force. The foreword in the publisher preview lists the main changes as: a new procedure for switching from normal or reduced inspection to skip-lot inspection, producer and product conditions for skip-lot inspection, guidance on the effectiveness of skip-lot inspection and on random selection of lots, and the removal of the per-plan OC curves in favour of a method for constructing them, placed in Annex E (ISO 2859-1:2026, foreword).

The foreword does not list any change to the sample-size and accept/reject tables. The table values in this article were read from Table I and Table II-A of MIL-STD-105E and match two industry examples (see below). Even so, confirm a value against the ISO 2859-1:2026 tables before you write it into a contract.

How do you read an AQL table?

The table is read in two steps; the third step is drawing and counting the sample.

1. Find the code letter. Cross the lot size with the inspection level in the sample-size code letter table. At general level II, a lot of 1,201–3,200 pieces gives code letter K.
2. Find the plan. The cell where the code-letter row meets the AQL column gives the sample size (n), the acceptance number (Ac) and the rejection number (Re). Code letter K means n = 125; the AQL 2.5 column reads 7 / 8 and the AQL 4.0 column reads 10 / 11.
3. Draw and count the sample. Draw n pieces at random, count the defective pieces by defect class and compare the counts with the acceptance and rejection numbers.

Sample plan table for general inspection level II, single sampling, normal inspection:

The values were read from MIL-STD-105E Table I and Table II-A. QIMA states that in a lot of 4,000 pieces (code letter L, AQL 2.5) 200 pieces are inspected, with acceptance up to 10 defective and rejection at 11; Wave PLM gives 125 pieces for a lot of 1,201–3,200. If the sample size is equal to or larger than the lot size, every piece is inspected (note to Table II-A). For lots above 35,000 pieces and for other AQL values such as 1.0, 1.5 and 6.5, consult the full table and its arrow rules.

Worked example: a 2,400-piece T-shirt lot

The lot is defined as one lot per style and colour; general level II, major AQL 2.5 and minor AQL 4.0 are used.

1. 2,400 pieces falls in the 1,201–3,200 range: code letter K.
2. For K, n = 125. Major: accept 7, reject 8. Minor: accept 10, reject 11. For critical defects a single defect stops the lot (QIMA).
3. The inspector draws 125 pieces at random. Result: 0 critical, 5 major and 9 minor defective pieces. Major 5 ≤ 7, minor 9 ≤ 10, critical 0: the lot is accepted. Had 8 major defective pieces turned up, the lot would have failed whatever the minor count; each class is judged against its own acceptance number.

The sample must be chosen so that it represents the sizes, colours and cartons in the lot. The standard says that, where needed, the sample should be taken in proportion to the size of the sub-lots and drawn at random within each sub-lot (MIL-STD-105E, paragraph 4.5.1). A sample consists of units selected without regard to their quality (paragraph 3.20); picking from the top cartons or from pieces that look good distorts the result.

Inspection level: I, II, III

The inspection level sets the relationship between lot size and sample size. Level II is normally used; level I is chosen when less discrimination is enough and level III when more is needed. The special levels S-1 to S-4 are for cases where a small sample is necessary and large sampling risks can be tolerated (MIL-STD-105E, paragraph 4.9.1). For a lot of 2,000 pieces at AQL 2.5 the table looks like this:

As the level goes up, the sample gets larger, the plan catches defective lots more often, and inspection costs rise.

Critical, major and minor defects: classification

AQL values are set per defect class, so classifying defects correctly matters as much as the AQL itself. The standard defines the classes as follows (MIL-STD-105E, paragraphs 3.5, 3.14 and 3.16; the summary in our wording):

Four rules on classification:

• The buyer prepares the defect list, and it goes into the contract. The same defect can fall into a different class for a different buyer. For the information a garment label must carry, see mandatory information on garment labels.
• State how defects are counted. The standard defines two ways of counting: percent defective (defective pieces) and defects per hundred units (MIL-STD-105E, paragraph 4.2). Under the defective-piece method, a piece with more than one defect is counted once, in the class of its most serious defect (paragraphs 3.6, 3.15 and 3.17). The two methods give different results.
• Classes are judged separately. Each class has its own AQL and acceptance number, and the switching rules apply independently by class (paragraphs 4.4.3 and 4.6).
• Sampling is not enough for the critical class. A plan with an acceptance number of 0 does not guarantee that a critical defect will be caught: if 1% of a 2,000-piece lot (20 pieces) is critically defective, the probability that none appears in a sample of 125 is about 28% (0.99 to the power of 125; our calculation). Our recommendation at Tekstil Rehberi is to manage risks such as needle fragments, small parts and restricted substances not by sampling alone but through 100% inspection (QIMA describes passing garments through a metal detector for needle fragments), process controls and laboratory testing.

How much assurance does AQL give?

The probability of acceptance shows how often a plan rejects a defective lot. The table below shows how often plan K (125 pieces, major AQL 2.5, accept 7 / reject 8) accepts a 2,000-piece lot at different real defect rates.

The values are our own calculation: a binomial distribution assuming a random sample. For a finite lot of 2,000 pieces the hypergeometric distribution differs by at most one percentage point, and there may be small differences from the OC tables in ISO 2859-1. Because ISO 2859-1:2026 removed the per-plan OC curves, they have to be produced with the method in Annex E.

Three conclusions follow from the table:

• AQL 2.5 does not mean "at most 2.5% defects". A lot with 5% defective pieces is accepted roughly seven times in ten. The probability of acceptance falls to 50% at a defect rate of about 6% and to 10% at about 9%.
• The bar for the factory is clear. A lot whose real major defect rate is about 3.2% or lower passes this plan in 95 of 100 inspections. The goal is not "passing the AQL" but keeping the defect rate well below it through inline checks.
• The inspection level changes the assurance. For the same lot with 5% defective pieces, the probability of acceptance is about 76% at level I, 71% at level II and 58% at level III; at 8% defective it is about 43%, 21% and 7% respectively.

Where does AQL inspection take place in garment production?

AQL sampling is not a single type of inspection. It can be used at different stages, but the accept/reject decision is mostly made at the final stage. For the whole production flow, see garment manufacturing process.

Inspection inside the factory is more frequent. A Turkish vocational training module (MEGEP) describes measurement as carried out at three stages: before production (raw materials, materials), during production and after production (finished goods). SPS Commerce likewise says that relying on a final inspection alone is no longer enough and that defects should be caught early and cheaply. Our recommendation at Tekstil Rehberi is to combine frequent or 100% inspection on the line with the final AQL inspection: AQL is a tool for the acceptance decision, not for preventing defects.

If final inspection is carried out when 80% of the order is packed, the remaining 20% falls outside the sampling. Decide in the contract in advance how that part will be handled (a second inspection, or a commitment before shipment).

What is checked at final inspection?

Steps adapted from QIMA's description of garment inspection:

1. Measurements. For each size in the order, the measurement points are compared with the measurement chart in the tech pack. Tolerance varies by measurement point: QIMA gives the example that a deviation of a quarter inch (about 0.6 cm) in chest width can be acceptable on a shirt, while the same deviation in sleeve length can count as a defect. Write the tolerance per measurement point. How many pieces are measured depends on the buyer's procedure; the sources give no number.
2. Fit. The piece is put on a mannequin of the right size and compared with reference images.
3. Trims. Pull tests on zips, repeated opening and closing tests on buttons and snaps, and stretch tests on elastic and tape are carried out on a set number of pieces.
4. Appearance and workmanship. Colour is compared with the approved sample and Pantone; print and embroidery position, pattern matching at seams on striped fabric, a metal detector for needle fragments and colour fastness to rubbing are checked. Woven garments are also tested for seam slippage.
5. Labels. Fibre composition, country of origin, producer or importer details, size and care instructions are verified; barcodes are scanned.
6. Packing. The correct quantity and size ratio per carton, carton material and closure, and shipping marks are checked.
7. Records. Defect counts are reported by class together with photographs, and the accept or reject decision is written down.

Performance properties such as shrinkage, colour fastness and chemical compliance are verified not by AQL inspection in the field but by laboratory testing; for test services, see the testing laboratories category in our directory.

What happens if the lot fails?

An AQL plan only says accept or reject; what follows is written in the contract. The buyer generally has three options: reject the lot and ask the factory for rework or replacement, bargain for a discount and accept the lot as it is, or have the factory carry out 100% sorting (Wave PLM). The choice depends on the seriousness of the defects, the shipping date and the factory's corrective plan.

Settle in advance which plan will be used to re-inspect a corrected lot, who bears the cost, and by what date the result will be given. ISO 2859-1 has a separate clause on resubmitted lots; read its conditions from the text of the standard.

Switching rules and skip-lot inspection

In a series of lots, results change the severity of inspection. The rules in MIL-STD-105E (paragraphs 4.7 and 4.8) are:

• Normal to tightened inspection: if 2 of the last 5 or fewer consecutive lots are rejected on original inspection.
• Tightened to normal inspection: if 5 consecutive lots are accepted on original inspection.
• Normal to reduced inspection: if the last 10 lots have been under normal inspection and all accepted, the total number of defects in the samples is below the limit in the table, production is steady and reduced inspection is wanted.
• Discontinuation of inspection: if the number of rejected lots under tightened inspection reaches five, inspection is discontinued until corrective action has been taken.

ISO 2859-1:2026 gathers the switching rules in clause 9 and adds skip-lot inspection in its 2026 edition: for a producer who consistently delivers good quality, the sample size may be reduced or the inspection of randomly selected lots skipped, at the discretion of the responsible authority (ISO 2859-1:2026, introduction). Confirm the numerical thresholds and the qualifications for skip-lot inspection from clause 9 of the standard; the numbers above belong to MIL-STD-105E.

What to write into the contract and purchase order

The MEGEP training module says that acceptable defect quantities in production and delivery should be set out as a written document in the agreement between producer and customer. Wave PLM also reports that brands that write AQL values into the order and the cost sheet have fewer disputes about a "pass" decision (an observation by a software vendor). Checklist:

1. AQL values by class (for example critical 0, major 2.5, minor 4.0) and the inspection level.
2. The lot definition (order, style-colour) and how the sample will be drawn: at random, in proportion to the size and carton distribution.
3. The defect list and its classes; the measurement chart and tolerances per measurement point.
4. The counting method: defective pieces or total defects.
5. When inspection takes place (for example when 80% of the order is packed), who carries it out (the buyer, a third party) and who pays.
6. The reference: the approved pre-production (PP) sample and visual standards. For the types of sample, see the sampling process.
7. What happens on rejection: time and cost of rework or sorting, the plan for re-inspection, and any right of appeal or second inspection.
8. Separate measures for critical risks: a test report, metal checks, trim documents.

For how contract manufacturing is set up (CMT versus full package), see the contract manufacturing guide; for general contract clauses, see the contract manufacturing agreement. Take legal advice for the contract itself; this article is not legal advice.

What to agree with a contract factory (CMT)

• Settle the scope in the quote. Are AQL final inspection and in-house inspection included in the price, and who bears the rework cost of a defective piece? These items also appear in the quote comparison in contract T-shirt sewing prices.
• Ask the factory to run its own final inspection on your plan. An in-house inspection with the same code letter, the same AQL and the same defect list shows your result in advance. Have results recorded by lot number; the record helps in a dispute (a Tekstil Rehberi recommendation).
• Tie the target to the defect rate. If your plan is K with AQL 2.5, a factory that keeps the major defect rate well below 3% raises its chance of passing (see the table above).

Choosing AQL values: notes for brands and buyers

The AQL value is chosen according to the product and the risk; there is no standard "right" value. Wave PLM's recommendations (not a standard, one vendor's suggestion):

A tighter AQL means fewer defective pieces get through, but also more rejections and higher inspection cost.

Turkish AQL terms on inspection documents

If you buy from a Turkish factory, inspection reports and quotations may use these terms. The table maps them to the English terms used in this guide.

Common mistakes

• Taking the AQL value for the percentage of defects allowed in the lot.
• Drawing the sample from pieces that look good or from the top cartons.
• Not writing the defect list and its classes into the contract, and not stating the counting method.
• Treating a plan designed for a continuing series of lots as adequate assurance for a single lot.
• Managing critical risks by sampling alone.
• Leaving measurements, labels and packing out of the inspection, or confusing laboratory tests with AQL inspection.
• Inspecting only at the end.

Summary: the steps of AQL inspection

Define the lot and the defect list, choose the inspection level and the AQL values by class, find the code letter, sample size and accept/reject numbers in the table, draw the sample at random and count, compare the result with the acceptance number class by class, and write down in advance what happens on rejection. Keep the probability-of-acceptance table in mind: AQL is a balance of risk, not a guarantee of a defect percentage. Companies offering quality control services are listed in the quality control category, and workshops that do pressing, packing and contract sewing in the CMT sewing category.