AQL Sampling, Explained: How QC Actually Inspects Your Vietnam Order (and the Setting That Quietly Sets Your Risk)

In our experience qualifying suppliers across six product categories, the buyers who get burned by a quality claim almost never got burned by the factory. They got burned by their own inspection setup: a wrong level, a loose AQL, or a report they never read closely enough to notice the sample was 50 units on a 5,000-piece order. AQL is not a grade the factory earns. It is a contract you write, in numbers, before a single carton is opened. This guide walks the mechanics the way an inspector actually runs them on the floor in Ho Chi Minh City or Bac Ninh, and flags where a buyer silently hands away protection.

What AQL sampling actually is (and what it is not)

Start with the name, because the name is where most of the confusion lives. AQL stands for Acceptance Quality Limit. It is the maximum share of defective units that an inspection plan will still accept as satisfactory. Set AQL 2.5 for major defects and you are telling the inspector, in effect, a lot running at about 2.5% majors should usually pass. The two governing standards are ISO 2859-1, whose third edition was published on 22 January 2026 and carries the title “Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection,” and its long-standing US equivalent ANSI/ASQ Z1.4. They use the same code letters and the same tables, so a Vietnam factory quoting “ISO 2859-1” and a US buyer quoting “Z1.4” are speaking the same language.

Here is the part buyers get wrong most often. AQL sampling is not 100% inspection. The inspector pulls a random sample from the finished, packed lot and judges the entire shipment on that sample. On a 5,000-unit order the sample under the standard default is 200 pieces, not 5,000, which trades a small, known statistical risk for an inspection you can finish in a day. It is also not a fixed percentage. People assume “we inspect 10%,” but the sample size is pulled from a table by lot size, not by a flat cut, so doubling the lot grows the sample far less than double. Treat AQL as full coverage, or as a percentage, and you have misread the tool before using it.

How the inspector gets from your lot to a pass or fail

The mechanics are fixed, which is the good news, because it means you can audit them. Every AQL inspection walks the same four steps.

First, the lot size. This is the number of units in the batch being offered for inspection, and it sets everything downstream. Second, the inspection level, which is a dial for how hard you look. ISO 2859-1 and Z1.4 give three general levels, I, II, and III; Level II is the default used for most consumer goods unless the buyer specifies otherwise. Third, the standard crosses lot size against inspection level to land on a sample size code letter, a single letter from A to R. A lot of 3,201 to 10,000 units at Level II gives code letter L, which sets a sample of 200 pieces. Fourth, the inspector reads across the AQL you chose to find the acceptance number and rejection number. At code letter L and AQL 2.5, the accept number is 21 and the reject number is 22: 21 or fewer major defects in the 200-piece sample and the lot passes, 22 or more and it fails.

That last step is the whole decision. The inspector is not scoring the lot on feel. They are counting defects against a number that you set weeks earlier when you signed off the inspection level and the AQL. The standard even has a purpose baked into it: pass or fail the lot in a way that pressures the supplier to keep its real process average at or below the AQL you asked for. Get the setup right and the math protects you. Get it loose and the same math waves marginal lots through.

The three settings that quietly move your risk

This is where know-how separates buyers who sleep well from buyers who eat charge-backs. Three choices, all made before inspection, change how much defective product you accept, and none of them announces itself in the pass/fail line of the report.

The inspection level. Level II is the default for a reason, but a buyer who quietly accepts Level I to save inspection time is checking a smaller sample and catching fewer problems on the same lot. A buyer who bumps a first order with a new supplier to Level III is pulling a larger sample and tightening the screen. The level is a knob on statistical power, and it is invisible in the headline result. The report says “PASS” either way.

The AQL tier per defect class. Real inspections do not use one AQL. They split defects into three classes and set a separate AQL for each. The near-universal setup for consumer goods is AQL 0 for critical defects, AQL 2.5 for major defects, and AQL 4.0 for minor defects, a convention published by inspection firms like QIMA và Tetra Inspection. Loosen majors to AQL 4.0 to make a tight order pass more easily, and you have formally agreed to accept more defective goods. That is a business decision, not a technicality, and it should be made on purpose.

Which defects count as major versus minor. The tables are objective, but the classification feeding them is a judgement call written in your defect standard. A loose seam booked as “minor” instead of “major” moves under a looser AQL and a higher acceptance number. If you have not written and agreed a defect classification with the factory, the inspector or the supplier is writing it for you.

Common AQL settings, and what each one accepts

The table below shows how the three standard defect classes map to AQLs and what that means in plain terms. These are the values used across most consumer-goods programs; your own risk tolerance can tighten them.

How the sample and accept number scale with lot size, all at the default Level II and AQL 2.5 for majors:

Accept/reject figures follow the ISO 2859-1 / Z1.4 single sampling plan for normal inspection, cross-checked against Tetra Inspection’s published table. Notice the sample barely quadruples while the lot grows more than thirtyfold. That is the code-letter logic at work, and it is exactly why “we inspect a fixed percentage” is the wrong mental model.

Vietnam vs the fixed-percentage habit: how buyers actually get protected

For buyers, the verdict is simple. Vietnam factories already run ISO 2859-1 inspections every day, so the winning move is not to invent your own sampling scheme; it is to specify the standard one tightly, name your inspection level and your three AQLs in the purchase order, and hold the timing. Do that and a Vietnamese supplier meets you with a process built for it.

Timing: the setting hidden inside the calendar

One more variable decides what your inspection is worth, and it is not on the tables at all. It is when you run it. The AQL sample is pulled at random from the finished lot, so a final random inspection is only meaningful once production is genuinely done. The industry convention is 100% of units produced and at least 80% packed before a final random inspection, per QC firms such as Testcoo. Inspect too early and the sample is not representative of what ships.

The practitioner move on a first order or a tight deadline is to add a during-production inspection at roughly 20% to 80% completion, so a systemic defect surfaces while the line can still fix it. The pattern we see most often on a blown Vietnam order is not a bad factory. It is a buyer who ran one end-of-line inspection, hit a defect the plant had been repeating since unit 200, and had no room left to rework before the vessel date. Timing is a risk setting as real as the AQL, and it costs almost nothing to get right.

You can pin the classification side of this down before you place the order. Use the Công cụ tính chi phí nhập khẩu của VietConnect to price in inspection and any rework buffer alongside duty and freight, so a tighter level or a second inspection is a line item you planned for rather than a surprise. And filter the danh bạ nhà cung cấp đã được xác minh for factories that already hold the quality certifications and run formal AQL inspections, so the sampling discipline is native to the plant rather than something you are teaching mid-order.

Các câu hỏi thường gặp

Q: Does AQL inspection mean every unit in my Vietnam order gets checked?

No. AQL inspection checks a random sample drawn from the finished lot, not the full quantity. On a 5,000-unit order at the standard Level II, the sample is 200 pieces. The whole lot is accepted or rejected on the defects found in that sample, which is why the inspection level and AQL you set matter so much.

Q: What AQL should I use for a consumer-goods order from Vietnam?

Most consumer-goods programs use AQL 0 for critical defects, AQL 2.5 for major defects, and AQL 4.0 for minor defects, at General Inspection Level II. Tighten majors to AQL 1.5, or move to Level III, on a first order with a new supplier or on any product where a defect is expensive to discover in market.

Q: How is the sample size decided, and is it a percentage of my order?

It is not a percentage. Under ISO 2859-1 and ANSI/ASQ Z1.4, the sample size comes from a table: your lot size and inspection level give a code letter, and the code letter gives the sample size. A lot of 3,201 to 10,000 units at Level II uses code letter L and a 200-piece sample, whether the order is 3,500 or 9,000 units.

Q: What is the difference between inspection Level I, II, and III?

They set how hard the inspection looks. Level II is the default and covers most orders. Level I pulls a smaller sample, catching fewer defects but costing less time. Level III pulls a larger sample and tightens the screen. The level is chosen before inspection and does not show up in the plain pass/fail line, so agree it deliberately.

Q: When should a final inspection happen on a Vietnam order?

A final random inspection should run once 100% of units are produced and at least 80% are packed, so the random sample truly represents the shipment. On tight timelines or first orders, add a during-production inspection at roughly 20% to 80% completion, so a repeating defect is caught while the line can still correct it before the goods are finished.