What acceptance sampling actually proves.

Why anybody samples at all.

Checking every unit of a container is impossible for most goods and destroys some of them. Sampling exists because a properly chosen subset supports a decision about the whole, and because the alternative is either checking nothing or checking everything at a cost nobody will pay.

The method is old, it is standardised, and its logic is that you accept a defined probability of being wrong in exchange for a decision you can afford. Both sides agreeing to that in advance is what makes the result binding rather than arguable.

The three things you set.

All three are commercial decisions rather than technical ones, and all three belong in the contract before production starts.

Inspection level

How much of the consignment gets opened. A higher level costs more inspector time and gives a tighter decision. It is chosen against the value of the goods and the consequence of a bad lot reaching your customer.

Acceptance quality limit

The proportion of defective units you are willing to accept in a lot. Set separately for critical, major and minor defects, because those are different kinds of problem.

Defect classification

What counts as critical, major or minor, written for your product. This is the part everybody skips and the part that decides every argument. A loose joint on a chair and a scratch on its underside are not the same defect.

What it cannot do.

It cannot tell you that every unit is good. That is not a weakness of the method, it is the method. A buyer who understands this treats a passed inspection as what it is, and a buyer who does not treats the first defective unit as evidence of fraud.

It also cannot judge anything the inspector was not told to look at. An inspection against a vague specification produces a report that says the goods conform to the purchase order, which is worth nothing. The sampling plan is only as good as the defect list underneath it.

And it says nothing about properties that need a laboratory. Moisture, residues, composition and microbiology are measured rather than inspected, and they need a sample sent somewhere with instruments.

Setting it up for an Indonesian order.

Write the defect classification with the producer, not at them. A workshop in Jepara knows which faults are common in their process and which are rare, and a classification written in a European office often polices the wrong things.

Agree the level and the limits in the order, alongside what happens on a failure: rework and re-inspect at whose cost, a discount, or cancellation. Deciding that while a failed container sits in a yard is the worst possible time.

Then attach the money. A balance falling due on a passed inspection is the single most useful term a buyer can hold, and it is the reason the whole apparatus works.

Questions buyers ask.

Read next.

• Inspecting before it sails

  The five moments worth inspecting, what a report has to contain, and the mistake that makes an inspection worthless.

• What a sample actually proves

  The narrow question a sample answers, and why the sealed duplicate you never open is worth more than the one you evaluate.

• Writing a specification suppliers can price

  The document that decides what arrives. What belongs in it, what has to be measurable, and the words that cause disputes.

• Pre-shipment inspection

  An independent check before the container is sealed. What it covers, what it cannot, and why it costs a fraction of the problem it prevents.

• Writing a defect list

  Critical, major and minor, defined for your product. The document everybody skips and the one that decides disputes.

• Choosing who inspects

  Global firm, local specialist or your own agent. Six questions that separate a real inspection from photographs.

• Choosing a laboratory

  Accreditation, scope, location and turnaround. Which decide whether a result is usable.