An inspection report with a zero on it feels like good news. The zero is the least informative number on the page, and the sample size decides how little it means.
An inspection report arrives with a line that says "AQL 2.5, 125 pieces inspected, 0 defects found, PASS". Buyers read the zero and relax. The zero is the least informative number on the page.
AQL 2.5 is a sampling scheme. It maps your lot size to a sample size, and the sample size to an acceptance number and a rejection number. Nothing about it says "2.5% defective is acceptable" in the way buyers usually assume. It says: draw this many, and if the defects you find are at or below this count, accept the lot.
The practical consequence: a report showing four defects on a 1,000-piece order is a pass. Eight is a fail. Same product, same inspector, two defects apart — and the difference decides whether you ship or rework.
Sampling can only rule things out. If you inspect 125 pieces and find nothing, you have not shown the lot is clean — you have shown that a lot with a defect rate above roughly 2.4% would probably have produced at least one bad piece in your sample. Above 2.4% is unlikely. Below 2.4% remains entirely possible.
Run the same logic at smaller sample sizes and the result is uncomfortable. Twenty pieces, zero defects, and the lot could still be running at 15% defective. This is not a flaw in the inspector's work. It is what sampling does, and it is why the sample size on the report matters more than the defect count.
| Field | Why it matters |
|---|---|
| Sample size | Should match the AQL table for your lot size. A 30-piece sample on a 5,000-piece order is not AQL 2.5 whatever the header says. |
| Accept / reject numbers | These tell you the decision rule that was applied. If they are absent, a judgment call replaced a standard. |
| Defect classification | Critical, major and minor are counted separately and have different acceptance numbers. A report that lumps them into one total has hidden the most important distinction. |
| Photographs of each defect | The count is arguable. A photograph of the specific unit with a measurement is not. |
| What was not checked | Function, packaging, labelling, barcode scanning and carton drop tests are separate from AQL work. Many reports are silent about them, and silence is not a pass. |
Tell you the defect rate of the whole lot. Predict how the product performs in use. Catch a defect that only appears after shipping, or one caused by a material substitution that looks identical. And it cannot fix a specification that was never written down — an inspector measures against a standard, and if the standard is vague the report will be vague in exactly the same places.
The useful way to hold an inspection report is as evidence about a sample, dated, with photographs. That is a real thing to have. It is not a certificate of quality, and the reports that present themselves as one are usually the ones with the least detail in them.
Every inspection report states the sample size, the acceptance numbers and what was not checked.
Send a supplier to check