ISO 2859-1 General Level II, read properly
Lot size sets the sample, the sample sets Ac and Re, and the AQL you quote is not the number of defects you are allowed. Working through one lot end to end.
ISO 2859-1 is quoted more often than it is read. It is worth walking one lot through the tables from start to finish, because almost every misunderstanding about AQL comes from skipping a step.
Step one. Lot size gives you a code letter
Take a lot of 5,000 pieces. Under General Inspection Level II — the default unless you have agreed otherwise — a lot between 3,201 and 10,000 falls in one band, and a lot of 3,200 sits in the band below it. That band gives you a code letter, and the code letter is the only thing that matters going forward. The lot size does nothing else.
Step two. The code letter gives you a sample size
For this lot the sample is 200 pieces. Notice what has not happened: nobody has mentioned a percentage. The sample is not 5% of the lot, and doubling the lot does not double the sample. The relationship is stepped, not proportional, which is why a 3,500-piece lot and a 9,000-piece lot can pull the same 200 pieces.
Step three. The AQL gives you Ac and Re
Now the AQL enters, and only now. At AQL 2.5 on a sample of 200, the acceptance number is 10 and the rejection number is 11. At AQL 4.0 it is 14 and 15. Those are the numbers the inspector actually works to.
What AQL is not
AQL 2.5 does not mean 2.5% defects are acceptable. It is a producer’s risk parameter: the quality level at which the sampling plan will accept lots most of the time. A lot running at exactly 2.5% defective has a high chance of passing, but the plan is designed around that probability, not around a permission.
This distinction matters commercially. A factory told it is “allowed 2.5%” will manage towards 2.5%. A factory told the plan accepts at 10 majors in 200 knows exactly where it stands on this lot, and so do you.
Severity classes are separate plans
Critical, major and minor are not one budget. They are usually three separate accept/reject decisions running on the same sample, commonly with critical at zero, majors at AQL 2.5 and minors at AQL 4.0. A lot can pass on minors and fail on majors, and one critical defect ends it regardless of everything else.
Normal, tightened, reduced
The last piece people skip. The plan is meant to move. Consecutive rejections switch you to tightened inspection, a run of acceptances can earn reduced. If you have been running normal severity on the same supplier for three years regardless of performance, you are not really running 2859-1 — you are running a fixed sample size and calling it a standard.