Nearly every pre-shipment inspection report from China references AQL. Your inspector writes “AQL 2.5, Level II, Ac 10 / Re 11” and you are supposed to know whether that means your goods passed. Most buyers do not — they nod, pay the balance, and find out what AQL meant when the container arrives.
This guide explains AQL sampling the way it actually works on the inspection floor: the standard behind it, the three defect classes, how the sample size is picked, and how to read the accept/reject numbers. By the end you will be able to write an inspection standard into your purchase order and read the report that comes back.
What AQL is — and what it is not
AQL stands for Acceptable Quality Limit (also called Acceptable Quality Level). It is a statistical system defined in the international standard ISO 2859-1, published in North America as ANSI/ASQ Z1.4 — both descended from the older US military standard MIL-STD-105E. It is the global benchmark for acceptance sampling: instead of checking every unit in a production run, which would be prohibitively expensive, inspectors draw a random sample and use the results to decide whether the whole lot is accepted or rejected.
Two misunderstandings cause most of the trouble:
- AQL is not “the maximum percentage of defects allowed in your lot.” Formally, it is the worst tolerable process average — the quality level a supplier’s ongoing production must achieve for lots to be accepted routinely. A lot at exactly the AQL level has roughly a 95% chance of being accepted.
- AQL 2.5 does not mean “2.5% defective is fine.” It is a statistical threshold tied to a specific sample size and a specific accept/reject count. A 2.5% defect rate in a small sample can still fail, and a lot can pass at AQL 2.5 while containing defects — that is the nature of sampling. If you need zero defects of a certain kind, you set that class to AQL 0 (critical defects) or require 100% inspection for that check.
The three defect classes
Before any numbers, you classify what can go wrong. Every inspection plan uses three classes:
- Critical defects — AQL 0. A defect that makes the product unsafe or illegal, or that causes complete failure. A live wire exposed on a switch panel, a battery that overheats, a missing earth connection. Tolerance is zero: one critical defect found in the sample rejects the lot.
- Major defects — usually AQL 2.5. A defect that makes the product unusable, unsaleable, or significantly below specification — a customer would return it. A non-functioning button, a wrong color, a missing component, a cracked housing. AQL 2.5 for major defects is the most common value in consumer goods, and the garment industry standard is Critical 0 / Major 2.5 / Minor 4.0.
- Minor defects — usually AQL 4.0. A cosmetic or small deviation that does not affect function: a tiny scratch on the back panel, a slightly off-center label, a minor finish variation. Customers usually accept these, but you still cap them.
Your defect list is the most important part of the plan, and it is the part buyers most often skip. Sit down with the product and write out what counts as critical, major, and minor — with photos. Send that list to the factory and to the inspector before production starts. An inspector cannot enforce a standard you never defined, and “good quality” is not a standard.
How the sample size gets picked
The standard uses two inputs: your lot size (total units in the order) and an inspection level. The default is General Inspection Level II — it is the level used in the vast majority of commercial inspections worldwide. Level I uses smaller samples (cheaper, less discrimination) and Level III uses larger ones (more discrimination, higher cost). Special levels S-1 to S-4 exist for destructive or expensive tests, where you can only afford to test a handful of units.
The chain runs in one direction: lot size plus inspection level gives a code letter, the code letter gives the sample size, and the sample size plus your AQL gives the accept (Ac) and reject (Re) numbers.
At General Inspection Level II, the code letters work out like this (single sampling, normal inspection):
- Lot 2–8: code A, sample 2
- Lot 9–15: code B, sample 3
- Lot 16–25: code C, sample 5
- Lot 26–50: code D, sample 8
- Lot 51–90: code E, sample 13
- Lot 91–150: code F, sample 20
- Lot 151–280: code G, sample 32
- Lot 281–500: code H, sample 50
- Lot 501–1,200: code J, sample 80
- Lot 1,201–3,200: code K, sample 125
- Lot 3,201–10,000: code L, sample 200
- Lot 10,001–35,000: code M, sample 315
- Lot 35,001–150,000: code N, sample 500
- Lot 150,001–500,000: code P, sample 800
- Lot 500,001 and over: code Q, sample 1,250
Notice what this means in practice: whether your order is 3,500 units or 9,500 units, the inspector pulls the same 200 units. The sample does not scale with your order — it scales with the lot-size band. That is why the lot size you declare matters, and why splitting one order into two “lots” to get a smaller sample is a trick that only hurts you.
Worked example: a 5,000-unit order
Take a typical first order: 5,000 smart switch panels. You write into the purchase order: ISO 2859-1, General Inspection Level II, single sampling, normal inspection, Critical 0 / Major 2.5 / Minor 4.0.
The chain runs:
- Lot size 5,000 at Level II gives code letter L.
- Code L means a sample of 200 units, drawn at random from the finished, packed goods.
- The master table gives the accept/reject numbers: at AQL 2.5 (major), Ac 10 / Re 11 — ten or fewer major defects passes, eleven or more fails. At AQL 4.0 (minor), Ac 14 / Re 15. At AQL 0 (critical), Ac 0 / Re 1 — any single critical defect rejects the lot.
The inspector pulls 200 units at random, tests function on each, checks appearance, verifies quantity and packaging, and counts defects by class. Say the report comes back with 0 critical, 8 major, 12 minor. The lot passes: 8 is at or below Ac 10 for major, 12 is at or below Ac 14 for minor. If it had come back with 11 major defects, the lot fails — even though 11 out of 200 is only 5.5% of the sample and 0.22% of the order. That is the deal you signed when you wrote AQL 2.5 into the order.
One more detail that surprises buyers: the inspector draws the sample from finished goods only. If production is not 100% complete and packed, the inspection cannot be a proper pre-shipment inspection — the sample would not represent the lot. Never let a factory rush an inspection on half-finished goods.
Normal, tightened, and reduced inspection
The standard has three severities, and the switching rules between them are part of the system:
- Normal inspection is where you start, and where you stay as long as lots keep passing.
- Tightened inspection kicks in after repeated failures — same code letter, larger sample or stricter Ac/Re. It stays in force until a run of consecutive lots passes.
- Reduced inspection is the reward for a long run of passes: smaller samples, lower cost. It is only used when the supplier has proven consistent quality.
For a first order with a new factory, normal inspection is the right choice. Do not accept reduced inspection on a first order, and do not let a factory talk you out of inspection entirely because “we already checked everything ourselves.”
Five mistakes that make AQL useless
- Writing AQL into the order but never defining the defect list. The inspector then classifies defects by their own judgment, and your idea of “major” may not match theirs.
- Accepting the factory’s in-house inspection as the pre-shipment inspection. The factory inspecting its own goods is not independent verification. Hire a third-party firm (SGS, Bureau Veritas, Intertek, QIMA, or a smaller specialist) — a basic pre-shipment inspection typically costs a few hundred dollars, far less than a failed container.
- Paying the balance before the inspection report arrives. The whole point of inspection-gated payment is leverage: the 70% balance is released only after the report passes. Once the money is sent, rework becomes a negotiation you have already lost.
- Inspecting too early. If the goods are not fully produced and packed, the sample is not representative. The report will say so in small print; read it.
- Treating “pass” as “perfect.” A passed lot at AQL 2.5 can still contain defective units — up to the accept number in the sample implies a small percentage across the lot. If your product cannot tolerate any functional defect reaching the customer, add a 100% function test for the critical functions on top of AQL sampling, and write that into the order too.
What to write in your purchase order
Copy this line, fill in your numbers, and put it in every purchase order:
Pre-shipment inspection per ISO 2859-1 (ANSI/ASQ Z1.4), General Inspection Level II, single sampling, normal inspection. Critical AQL 0, Major AQL 2.5, Minor AQL 4.0. Inspection by [third-party firm] on 100% finished and packed goods. Balance payment released only after a passed inspection report. Defect classification per attached checklist [attach your list with photos].
Then attach the defect checklist. That one paragraph, enforced, prevents more losses than any other sentence in the order.
Sources
- IndexFox, “AQL Sampling Explained | How to Read an Inspection Report (ISO 2859-1)” — https://www.indexfox.com/insights/sourcing-guides/aql-sampling-explained/
- AQM BD, “AQL Calculator | Sample Size & ISO 2859-1 Tool” — https://aqmbd.com/quality-control/aql-calculator.php
- SDF Ltd, “AQL Calculator — Free ISO 2859-1 Sample Size Tool” — https://sdfltd.com/tools/aql-calculator/
- NR Tech Studio, “AQL Calculator: Sampling Plans per ANSI/ASQ Z1.4” — https://www.nrtechstudio.com/tools/aql-calculator/
- KeepPerfectBags, “AQL Sampling Table: Lot Size to Sample Size and Ac/Re” — https://www.keepperfectbags.com/tools/aql-sampling-table/