AQL chart for garment inspection: 2.5 and 4.0 explained

An AQL chart turns two facts, the size of your lot and the AQL you agreed, into three numbers: how many garments to inspect, how many defects you can accept, and the count at which the lot fails. For most apparel at General Inspection Level II, a lot of 3,201 to 10,000 pieces means inspecting 200 garments; at AQL 2.5 the lot passes with 10 major defects or fewer and fails at 11. At AQL 4.0 the same 200 pieces may carry 14 minor defects and fail at 15. Jump to the AQL 2.5 table.

What does AQL 2.5 actually mean?

AQL stands for Acceptance Quality Limit. It is the index that selects a sampling plan from the tables in ISO 2859-1, which uses the same tables as the American ANSI/ASQ Z1.4. The US National Institute of Standards and Technology describes the AQL as the baseline quality requirement for the producer’s product: a percent defective the process should be able to stay under.

So “AQL 2.5” is not a promise that 2.5% of your garments will be faulty, and it is not a guarantee that only 2.5% are. It is a rule for a decision. You draw a random sample of a fixed size, count what you find, and compare the count with an accept number and a reject number. Apparel orders usually carry two limits at once: 2.5 for major defects and 4.0 for minor defects, with critical defects set at zero. The quality control page explains how we apply those limits on an order. This article is about reading the chart itself.

Step 1: how do you find the sample size code letter?

The first table converts lot size into a code letter. The lot is the quantity presented for inspection, normally the whole order for one style from one factory. Pick the row for your lot size, then the column for your inspection level. Level II is the default unless the buyer specifies otherwise.

Read from Table I of MIL-STD-105E, the military standard these tables come from. The rows from 91 to 150,000 pieces were checked row by row on 23 September 2026, and the rows for smaller and larger lots on 1 October 2026. The letter is not a sample size yet. It is a pointer into the second table.

Step 2: the AQL 2.5 table, with sample size and accept and reject numbers

Print it: the AQL 2.5 table as a one-page PDF (A4), with AQL 1.5 and 4.0 alongside, free to print and share.

The second table, single sampling plans for normal inspection, gives each letter a sample size and, under each AQL column, an Ac (accept) and Re (reject) number. If the defects found are equal to or below Ac, the lot is accepted. If they reach Re, it is rejected. In the normal single sampling table there is no grey zone: Re is always Ac plus one.

Figures read from Table II-A of MIL-STD-105E, the master table of single sampling plans for normal inspection. The rows from 91 to 35,000 pieces were checked on 23 September 2026, and the other rows on 1 October 2026. Where the printed table has an arrow, the cell shows the plan the arrow points to and that plan’s sample size: “0 / 1 on 8” means inspect 8 pieces, accept none and reject at 1, and if that sample is as large as your lot, inspect every piece. We have reproduced the table, not calculated it. Always confirm against the table your inspector or your retailer’s manual uses.

A worked example: 2,400 hoodies at AQL 2.5 and 4.0

Say one style of hoodie, 2,400 pieces, made in one factory, is ready for final inspection at General Level II with AQL 2.5 major and 4.0 minor.

• Code letter: 2,400 falls in 1,201 to 3,200, so Level II gives K.
• Sample: K means 125 pieces, drawn at random across colours, sizes and cartons.
• Majors at AQL 2.5: accept with 7 or fewer, reject at 8.
• Minors at AQL 4.0: accept with 10 or fewer, reject at 11.
• Criticals: one is enough to reject.

The two counts are judged separately. Six majors and eleven minors fail the lot on minors, even though the majors passed. Six majors and ten minors pass both. A lot can fail on a single critical defect, such as a broken needle point, whatever the other two counts say.

Notice what the chart does not do. It does not scale neatly with quantity. A lot of 1,201 pieces and a lot of 3,200 pieces both get a sample of 125. That is deliberate: the tables protect the buyer more on larger lots, where the sample is a smaller share but the evidence is stronger.

Why AQL 2.5 is not a 2.5% defect allowance

A claim still repeated on inspection and sourcing websites is that AQL 2.5 lets 2.5% of the order be defective. The standard behind the chart says otherwise. MIL-STD-105E, a US military standard dated 10 May 1989, is the last revision of the standard that, as MIL-STD-105D, was adopted as ANSI Z1.4 in 1971 and, with minor changes, as ISO 2859 in 1974; NIST notes that the basic tables of the three remain the same. It is no longer a current standard: Notice 1 cancelled it on 27 February 1995 and referred future purchases to ANSI/ASQC Z1.4-1993. We quote it because it is free to read, but a purchase order or quality manual should cite the current standard, ISO 2859-1:2026 or ANSI/ASQ Z1.4. Its clause 3.1 defines the AQL, when a continuous series of lots is considered, as the limit of a satisfactory process average. Clause 4.4.2 adds that selecting or using an AQL does not give the supplier the right to supply any defective unit.

The accept number shows the same thing. At General Level II, a lot of 3,201 to 10,000 garments is code letter L: 200 are inspected, and at AQL 2.5 the lot passes with up to 10 major defects and fails at 11. Ten in 200 is 5% of the sample, twice the figure on the label. At Level III the same lot is letter M: 315 inspected, accept 14, reject 15, so up to 4.4% of the sample.

Clause 3.1 also warns that the AQL alone does not give the chance of accepting any single lot; for that you read the operating characteristic curve of the plan. MIL-STD-105E prints those values, and for this lot size they show how often a worse lot still passes.

Figures from Tables II-A, VI-A, X-L-1 and X-M-1 of MIL-STD-105E, checked 23 September 2026. The last three rows are the share of garments in the whole lot with a defect of the class being counted; the standard gives the same values for defects per 100 garments. Clause 3.1 explains the pattern: a plan is built to accept the great majority of lots when the process average is no worse than the AQL, which is why, under Level II, a lot more than twice as bad as the label, at 5.33% defective, still passes half the time. Level III costs 115 more garments to inspect and lowers each of the last three figures.

What this changes on a single order

Clause 1.2 of MIL-STD-105E says the plans are meant mainly for a continuing series of lots, and that anyone inspecting an isolated lot should consult the operating characteristic curves to find a plan with the protection they want. Clause 4.11 sets out how: choose a limiting quality, the defect level you want accepted only rarely, and a consumer’s risk, commonly 10% or 5%, then take the plan from Tables VI and VII. NIST describes the buyer’s side of the same trade-off as the lot tolerance percent defective (LTPD), a defect level the consumer finds unacceptable and wants accepted only with a low probability. ISO 2859-1:2026 lists tables of consumer’s risk quality (CRQ) in its contents; ask your inspector to use them when an order is a one-off.

• First or one-off order: decide the share of defective garments you are not willing to see pass more than 1 time in 10, and check that the level and AQL in your purchase order deliver it. If 7.7% at Level II is too high, choose Level III or a tighter AQL before production, not after a disappointing report.
• Defects found in the sample: ask for every defective garment the inspector finds to be repaired or replaced, even when the lot passes. Clause 4.4.2 supports that position.
• Repeat orders: because the AQL describes a process average, keep the defect counts from every inspection and compare them order by order, using the same defect classification and inspection checklist each time.

What do the arrows mean, and what about small lots?

Some cells in the printed table hold an arrow instead of numbers. That happens where the sample for that letter is too small to support the chosen AQL. The instruction printed with the table is to use the first sampling plan the arrow points to, and to use that plan’s sample size as well as its numbers. If the sample size then equals or exceeds the lot, inspect every piece.

This is why small orders feel expensive to inspect. A lot of 120 pieces at Level II needs 20 garments checked, a sixth of the order. For runs made through a small-run route, a buyer can accept that proportion, or check every piece of a first order and move to sampling once the factory has earned it.

Level I, II or III: which inspection level should you use?

ISO 2859-1 offers three general levels. Level II is the default. Level I inspects fewer pieces and discriminates less; Level III inspects more and discriminates more. On a lot of 3,201 to 10,000 pieces, the choice changes the sample from 80 (Level I, letter J) to 200 (Level II, letter L) to 315 (Level III, letter M).

The table also has four special levels, S-1 to S-4, with much smaller samples. Clause 4.9.1 of MIL-STD-105E says they may be used where relatively small samples are necessary and large sampling risks can or must be tolerated. In apparel they suit checks that cost a garment or a lot of time per piece, such as a full measurement check against every point on the spec or a destructive seam test, where a small sample is agreed in advance.

A careful buyer changes the level for a reason and writes it into the order: Level III on a first order with a new factory, or a tighter AQL of 1.5 on majors for a premium product.

Tightened, reduced and skip-lot inspection

AQL tables were designed for a continuing series of lots from one producer, not a single shipment. The introduction to ISO 2859-1 says the scheme protects the buyer by switching to tightened inspection when quality deteriorates, and rewards consistent quality by switching to a smaller sample. MIL-STD-105E sets the trigger in clause 4.7.1: tightened inspection starts when 2 out of 2, 3, 4 or 5 consecutive lots are rejected on original inspection.

• Normal: where every relationship starts, and the table above.
• Tightened: stricter acceptance criteria, used after repeated failures until quality recovers.
• Reduced: smaller samples after a run of accepted lots and steady production, at the discretion of the buyer or its quality team.
• Skip-lot: inspecting only randomly chosen lots once a producer is qualified. This is new in the 2026 edition.

For a brand placing two or three orders a year, the switching rules rarely come into play. They matter once you run repeat programmes with the same factory, and they are a good reason to keep inspection records in one place.

Did ISO 2859-1:2026 change the AQL chart?

ISO published a third edition, ISO 2859-1:2026, which cancels and replaces the 1999 edition. Its foreword lists the main changes: a new procedure for switching to skip-lot inspection, guidance on qualifying producers and products for it, guidance on randomly selecting lots, and the removal of the printed operating characteristic curves in favour of methods to construct them. The code letters and the acceptance tables are not in that list of changes.

So a chart printed from the 1999 edition still gives the same answers for normal single sampling. What changes is the paperwork: new contracts and quality manuals should cite the 2026 edition, and anyone offering skip-lot inspection should be able to show how they qualified the factory for it.

Five mistakes people make reading the chart

• Treating two factories as one lot. The standard applies to lots from one producer using one process; different producers are sampled separately.
• Counting defects when you meant defective pieces. The standard supports both nonconforming items and nonconformities per 100 items. Agree which one the report will count: a shirt with three faults is one defective piece but three defects.
• Sampling from the top cartons. The sample must be random across the lot, or the statistics mean nothing.
• Inspecting before the goods exist. Clause 4.5.2 of MIL-STD-105E lets the sample be drawn after every unit in the lot has been assembled, or during assembly only if the lot size is fixed before the first sample unit is drawn.
• Agreeing an AQL without defining a major. The chart counts; it does not classify. Classification lives in your tech pack and defect list.

What to do next

For Level I or III, or for an AQL of 0.65, 1.0 or 6.5, our AQL sample size calculator reads the sample size and the accept and reject numbers from the same table for any lot size.

Put the inspection standard, level and AQLs in the purchase order before production, next to the sealed sample and the measurement tolerances in your tech pack. If you are producing hoodies, t-shirts or anything else in Bangladesh and want the inspection plan agreed before a single piece is cut, see how we work or send us your quantities and style list and we will set out the sample sizes that apply.