If a supplier tells you every kit passes 100 percent final inspection, ask them what their AQL is. The two statements are usually not compatible, and the answer tells you how their quality system actually works.
What AQL actually means
AQL stands for Acceptable Quality Limit. It is the defect rate at which a lot is still considered acceptable by the sampling plan you agreed to use.
It is not a target. It is not a promise that your defect rate will be that number. It is the line the plan uses to decide whether to accept or reject a lot based on a sample, and it is deliberately a statistical statement rather than an absolute one.
The two standards most people work from are ANSI/ASQ Z1.4 and ISO 2859-1. Both do the same job. You bring a lot size, an inspection level and an AQL value, and the table gives you a sample size and an accept or reject number. Find seven defects when the plan says accept on five, reject the lot.
How a sampling plan actually gets chosen
Four inputs decide the plan, and each one is a real decision rather than a default:
- Lot size. A run of four hundred kits and a run of forty thousand do not get the same sample.
- Inspection level. General level II is the usual starting point. Level III pulls a larger sample and is used when the consequence of an escape is high. Special levels pull smaller samples and are used for destructive or expensive tests.
- AQL value. Critical, major and minor defects normally carry different AQLs on the same plan, because a missing lancet and a crooked label are not the same event.
- Severity state. Normal, tightened or reduced. A plan that keeps rejecting moves to tightened. A plan with a long clean history can move to reduced. If a supplier has never moved off normal in either direction, their plan is decorative.
That last point is the one worth pressing. Switching rules are the part of the standard that makes it a control system rather than a spreadsheet. A supplier who cannot tell you when they last went to tightened inspection is not running the plan, they are running the sample size.
Why 100 percent inspection is usually not what is happening
Full inspection of every unit is real, and there are places it belongs. It is also expensive, slow, and less reliable than people assume, because inspector fatigue on repetitive visual checks is well documented. A tired human looking at the four thousandth identical kit is not a better instrument than a properly sized sample.
So when a supplier claims 100 percent final inspection on a high volume kit program, one of a few things is usually true. They are inspecting one specific attribute on every unit and sampling the rest. They are describing an automated check, such as a scale verification or a vision system, and calling it inspection. Or the claim is aspirational.
None of those are scandals. The first two are good practice. The problem is only the language, and the reason the language matters is that it is a quality claim in a regulated context. If your supplier writes it in marketing copy, it will eventually be read back to them by an auditor or a client, and it will not survive.
The accurate version of the claim is narrower and stronger: inspection happens at defined points, at a level set by the procedure, the AQL and the client requirement, and that happens every time. That is a statement about process discipline rather than about counting every unit, and it is the one you actually want to be true.
Where inspection belongs in a kit build
One final check at the end is the weakest possible design, because it finds problems at the most expensive moment. Layered checkpoints are cheaper and catch different failure modes:
- Incoming, on receipt of components. Catches the supplier defect before you build it into four thousand kits.
- In process, during assembly. Catches the operator or setup error while the run can still be corrected.
- Line clearance between runs. Catches the mixed component and the stray label, which is the defect class regulators care about most.
- Final, against the packing configuration. Catches the assembly that is complete but wrong.
- Release, against the batch record. Catches the paperwork gap, which is the defect that fails an audit even when every kit is perfect.
The specific defect classification matters as much as the count. A kit with a smudged outer label and a kit with the wrong preservation medium are not the same finding, and a plan that treats them the same is not protecting the patient.
What to ask a supplier
- What is the AQL for critical, major and minor defects on my program, and where is it written down?
- Which inspection level, and why that one for my lot sizes?
- Show me a rejected lot from the last year and what happened next.
- When did you last move to tightened inspection, and what triggered it?
- How are defects classified, and who approves the classification?
- What is checked at incoming, in process, line clearance, final and release?
A supplier who answers those specifically is running a quality system. A supplier who answers with a percentage is quoting a brochure.
Related reading: Audit your specimen kit supplier and What QMSR means for at-home test kit companies.
Ask for the plan, not the percentage. The plan tells you what happens on a bad day, and the percentage only tells you what someone wrote in a deck.
