Short answer first: AQL sampling answers two questions — how many units to pull from your shipment, and how many defects of each type are tolerable before the lot is rejected. You control exactly two numbers: the inspection level (how many to sample) and the AQL thresholds (how much to tolerate). For most consumer goods imported from China, the workable default is Level II sampling with AQL 2.5 for major defects, 4.0 for minor, and zero tolerance for critical — and the settings around that default are where your protection actually lives.
This guide walks through the system behind those numbers: where they come from, how to read the table, when to tighten them, and the two supplier tactics that quietly defeat an inspection that looks perfectly rigorous on paper.
Why You Sample at All Instead of Inspecting Everything
Because 100% inspection is an illusion you should not pay for twice.
A real full inspection of 5,000 units means a full day or more per inspector, paid, on-site — and when factories “100% inspect” themselves before you arrive, the sorting is done by the people whose paycheck depends on the goods passing. Independent full sorting by your inspectors is the only 100% inspection with meaning, and for most products it is uneconomical.
Statistical sampling exists precisely for this problem: a well-designed random sample gives you a defensible estimate of the whole lot’s quality at a fraction of the cost. AQL is how the math is standardized so both you and the factory argue from the same chart instead of from feelings.
The Standard Behind the Numbers: ISO 2859-1
AQL tables are not vendor folklore. They come from ISO 2859-1 (the American twin is ANSI/ASQ Z1.4) — an internationally agreed sampling system with one job: turn “how big is the lot?” into “how many units do I open, and how many defects still count as a pass?”
The flow has three steps.
Step 1 — Lot size determines a sample code letter. Find your lot size in the table and pick an inspection level. Normal general inspections use Level II. Level I saves time on established suppliers; Level III buys you more units opened when risk is elevated.
Step 2 — The code letter fixes your sample size. A lot of 1,201–3,200 units at Level II gives code K — a sample of 125 units, regardless of whether the lot is 1,300 or 3,200. Notice the curve: sample size grows sub-linearly with lot size. Doubling your order does not double your inspection time.
Step 3 — Your AQL thresholds turn the sample into accept/reject numbers. Read across the AQL column you chose. Where a number falls on an arrow, you follow it to the first usable row — the tables are built to keep sample sizes and tolerances statistically consistent.
Worked example — a 2,000-unit lot (code K, n = 125):
| Defect class | AQL set | With n = 125 | The lot is rejected if |
|---|---|---|---|
| Critical | 0 | accept 0, reject 1 | you find even 1 |
| Major | 2.5 | accept ≤ 7, reject ≥ 8 | 8 majors appear |
| Minor | 4.0 | accept ≤ 10, reject ≥ 11 | 11 minors appear |
Seventeen units out of 125 — roughly 13% of the sample — can carry a defect and the container still ships, as long as none are critical and the mix stays inside the lines. That is what “AQL 2.5” actually means. You are buying statistical control, not perfection. The moment you understand this, you also understand why the next section matters more than the table.
Critical, Major, Minor: The Definition That Decides Everything
The entire AQL system hangs on which class a defect is assigned to — and that assignment is exactly where disputes hide.
Critical defects are safety, legal or un-sellable problems: exposed wiring, a banned substance, broken glass in food packaging, missing compliance marks. The standard answer is AQL 0: one is a failed shipment.
Major defects break use or sellability without being dangerous: a zipper that jams, a colorway visibly off-spec against the approved sample, a hole where there should be stitching.
Minor defects are cosmetic flaws a normal buyer might tolerate: a stray thread, a faint scuff, slight print misalignment.
Here is the manipulation: a supplier who reclassifies problems downward converts a rejected shipment into a passed one without changing a single physical defect. Seven “major” holes become seven “minor” thread ends, and at AQL 4.0 the same seven units that rejected the lot now let it sail. This is why your purchase order must define the defect classes — and why the inspector who reports to you, not to the factory, is the one who classifies. (The factory-approved sample is the reference standard for “off-spec” calls; more on keeping the sample and the bulk honest in our sample vs. bulk guide.)
When 2.5 Is Too Generous: Tightening the Default
The consumer-goods default is a floor, not a commandment. Tighten toward AQL 1.0 or 1.5 — or move to Level III sampling, or full sorting — when any of these are true:
Functional failure is expensive. Electronics with firmware, appliances, anything with a power source. A field failure rate on imported consumer electronics typically runs 2–5%; your tolerance math should reflect what a return actually costs you, not the chart’s comfort zone.
A defect can hurt someone. Children’s products, food-contact materials, anything with a battery, anything with a blade. Critical stays 0, and majors on safety-relevant features effectively should.
It is the first order from a new supplier. Level II on a partner with zero history is sampling the wrong thing — you have not yet earned statistical trust. Inspect harder on order one, and you will inspect cheaper on orders two through ten.
Your margin cannot absorb the rework. If a minor-heavy lot is unsellable in your market even though it technically passes, the 4.0 minor threshold is protecting the factory, not you.
Conversely, a seasoned supplier with three consecutive clean results can justify dropping to Level I on stable SKUs. That is the direction sampling should loosen — never the direction the factory suggests.
The Two Supplier Tactics That Defeat a Perfect Inspection
Every AQL chart assumes the sample is random. These two tactics break that assumption while leaving the paperwork spotless.
Pre-sorting. The factory quietly pulls defective units out of the cartons in the days before your inspector arrives — or “inspects and repacks” every carton on the route to the warehouse. When your inspector opens the lot, it genuinely is better than reality, the sample passes, and the defects ship in the cartons nobody opened. Counter: unannounced timing (you set the date, not the factory), the inspector selects which cartons to open themselves, and any carton that fails gets expanded inspection on the spot.
Re-packing inside the window. During multi-day inspections, the factory “fixes” issues as they are found — so what you sampled at 9am on Monday is not what ships on Friday. Counter: final quantities confirmed by the inspector’s own re-count after the last carton is sealed, or a random second-sample pulled from sealed containers on loading day.
Neither tactic looks wrong on a report. That is precisely why the report format deserves its own read — How to Read an Inspection Report covers what a clean report should contain and what an empty-looking one hides.
Writing It Into the Purchase Order
Put these five terms in the PO — this is the difference between a policy and a preference:
Inspection follows ISO 2859-1 (ANSI Z1.4), Level II, single sampling, normal inspection.
AQL: 2.5 major / 4.0 minor / 0 critical — tightened on new suppliers and safety features, with defect classes defined by reference to the approved sample.
Sample selection is performed solely by the buyer’s inspector, cartons chosen on arrival, not offered.
No shipment before written pass results. A failed lot triggers rework and re-inspection at supplier cost.
Payment milestones are tied to the inspection outcome (deposit already paid, balance on pass) — leverage is only real if it arrives after the verdict.
What This Means in Practice
AQL is the least glamorous part of quality control and the part suppliers test most. You do not need to memorize the table — you need to set the two numbers, define the three classes, choose who pulls the sample, and put all of it in writing before production starts.
If you would rather someone else run this ritual on every order: our quality control service performs unannounced inspections to these exact standards, selects the cartons ourselves, and reports in the format described above — inspection reports that arrive before your balance does, never after your container has.
And if your next shipment is from a supplier you have not tested before: that is exactly the moment to tighten, not to trust the default.
WhatsApp: +852 46401579 · Email: info@ezysourcing.com — free initial consultation, honest answer on whether your current QC terms are load-bearing.
Written by the founder of EzySourcing, a Hong Kong-based sourcing firm working exclusively for buyers — after more than a decade advising foreign importers on sourcing and doing business in China.