AQL guide: acceptance sampling explained

For a lot of 5,000 pieces at inspection level II, the inspector checks 200 pieces under AQL (ISO 2859-1 / ANSI Z1.4). The lot is rejected at 1 critical defect, 11 major defects (AQL 2.5) or 15 minor defects (AQL 4.0).

What is AQL?

AQL stands for “Acceptable Quality Level”. It is a random sampling method defined by ISO 2859-1 (formerly MIL-STD-105E) and by its American equivalent, ANSI/ASQ Z1.4.

The AQL standard enables a good overview of the quality level without inspecting 100% of the production . Even though it involves a complex statistics'sampling method, it basically provides an optimized level of confidence versus a defined number of checked pieces. AQL tables also allow you to determine the maximum number of unsatisfactory pieces for three different types of defects, which are sorted by the severity of the consequences they have. Commonly, these defects can be:

• Critical: The product is unsafe or hazardous for the end user or/and contravenes mandatory regulations.
• Major: The product's usage is affected, and/or the defects are perceivable enough to affect the product's marketability or sellability.
• Minor: Here, the sellability is not directly compromised but that kind of defects should be mentioned, counted and limited.

How to read the AQL tables: a worked example

It starts with two already filled in tables, A and B. Table A determines a letter based on two factors:

• Production quantity
• Inspection level which will be applied: at level I fewer pieces will be checked, in level III, more pieces will be checked.

Let’s say we have a production (*Lot or Batch Size) of 5,000 pieces. You can see here that on the standard level II, the matching letter is L.

Once you know the letter, you can use Table B. You can see there, in the left column that for the letter L the corresponding sampling size is 200 pieces.

Ac – Acceptance number   Re – Rejection number

Then we look at the figures in this row. They correspond to the different percentages for Critical, Major, and Minor Defects. In this example, let’s say the client is buying high-value products packaging. He sets the limit at 0 Critical Defects, 1.5% of Major Defects, and 4% of Minor Defects.

Based on that, one can determine that for a sample size of 200 pieces with an AQL of 1.5:

• Accept (Ac = 7): up to 7 defective pieces is an "acceptable number".
• Reject (Re = 8): 8 defective pieces or more and the sample fails inspection, so the whole lot is rejected.

In other words, to comply with AQL 1.5, no more than 7 units from that sample size may fail inspection.

General (I, II, III) and special (S-1 to S-4) inspection levels

The inspection level sets the sample size for a given lot size. ISO 2859-1 and ANSI/ASQ Z1.4 define seven:

• General levels I, II and III: used for visual inspection and routine checks. Level II is the default; level I reduces the sample (a supplier with a good track record), level III increases it (after a failed inspection or for a sensitive product).
• Special levels S-1, S-2, S-3 and S-4: very small samples, for when checking many pieces is impossible or too costly and a higher sampling risk is acceptable. They are generally kept for measurements and for long or destructive tests (carton drop test, product teardown), not for visual checks.

For a lot of 5,000 pieces:

Do not confuse these levels with normal, tightened or reduced inspection: those are the standard’s switching rules, which tighten or relax the acceptance criteria over a series of lots depending on previous results.

ISO 2859-1, ANSI/ASQ Z1.4, MIL-STD-105E: one table

The AQL tables were first published by the US military as MIL-STD-105, whose last version, MIL-STD-105E, was cancelled in 1995. They live on under two names:

• ISO 2859-1, the international standard, adopted in China as GB/T 2828.1;
• ANSI/ASQ Z1.4, its American equivalent, often quoted by North American buyers.

For single sampling under normal inspection, which covers almost every pre-shipment inspection, the code letters, sample sizes and accept and reject numbers are the same. Whether your customer or purchase order says “ANSI Z1.4” or “ISO 2859-1”, the inspector reads the same table.

Supplier’s risk vs buyer’s risk

From the above example, you may have noticed that 1.5% of 200 is not 7. This is where the statistics come into play. Because we are only checking a small part of the production, it is still possible that in this particular inspection batch we find 6 defects while the actual proportion of defects in the total production (total lot size) does not exceed 1.5%.

For suppliers, there is a potential risk of having their production rejected despite meeting the level of quality requested. Statistically, however, the probability of finding 8 or more defective pieces out of 200 inspected, is rather low, if the total percentage of defective pieces is below 1.5%.

On the other hand, there is also a risk that an AQL based inspection will report the sample as “passed”. But actually, the proportion of defects in the inspected batch really is proportionally smaller than what it is in the entire production. This is what we call “the buyer’s risk”. It should not be underestimated.

The AQL standard favors suppliers as it takes into account their risk more extensively than the buyer’s risk. Therefore, when an AQL result is failed, clients have all the legitimacy to ask their suppliers to rework the production and pay for reinspection.

Choosing your inspection level and AQL values

Buyers should make full use of the tools they have at their disposal, to enhance the inspections' accuracy. First of all, the levels (I, II and III) should vary according to the trust of the buyer in his manufacturer.

The larger the sample size is, the lower the margin of error. The inspection result should reflect more accurately the overall quality of the production, the upper we climb on the level ladder. ( Level III being the more accurate )

Therefore, if, after various orders, you find that your supplier’s quality has consistently been good, you may consider lowering the sampling size to Level I. On the contrary, if the last report showed that the production failed to meet the requirements, it makes sense to increase the check to Level III.

As for the severity, for regular consumer products, the standard tolerates 0 Critical Defects, 2.5% of Major Defects, and 4% of Minor Defects. The buyer may choose a different level depending on the industry, product’s value, and the type of end users. It is essential to communicate about these AQL levels with the factory prior to placing an order, as well as to refer to them in the Purchase Order or contract.

AQL explained on video

Two three-minute videos (in French): how AQL sampling works, then what to do when a lot fails inspection. More on our YouTube channel.

Full AQL tables (ISO 2859-1)

AQL sampling is mostly used in pre-shipment inspection. To see how sampling fits into a complete approach, read our guide to quality control in China.