Quality

AQL Inspection for Phone Cases: Sampling, Defect Classes and What to Reject

8 November 2026

AQL is widely quoted and widely misread. It is not a target defect rate, and it does not mean "2.5% of your order may be faulty". It is a rule for deciding, from a sample, whether a whole batch is acceptable. Understanding the rule turns an argument at the end of production into a plan agreed at the start.

What the standard actually says

The reference is ISO 2859-1, the same tables that appear as GB/T 2828.1. For a given batch size and inspection level, the standard fixes how many pieces to pull and how many defects are allowed before the batch is rejected. Two levels are commonly used in consumer goods: 2.5 for major defects and 4.0 for minor ones.

Read the row for your batch, pull that many pieces at random from across the whole lot rather than from the top carton, and count defects. Confirm the exact figures against the published tables before you write them into a purchase order; the point here is the shape of the rule, not a substitute for the standard.

Three classes, and why the split matters

Every inspection plan has to define what counts as a defect before anything is counted, otherwise the same scratch is major to the buyer and minor to the factory. Our defect glossary names each fault and where the cosmetic line actually sits.

• Critical. Unsafe or unusable. Sharp flash on an edge, a case that will not seat on the handset, a camera opening that blocks the lens, a cracked shell. One critical defect rejects the batch, regardless of the AQL number.
• Major. Functional or clearly visible. Assembly gap at a seam, a button that does not return, print misregistration beyond the agreed tolerance, visible colour difference against the approved sample.
• Minor. Cosmetic and not noticed in normal use. Faint surface marks inside the shell, slight variation in packaging print, dust under a protective film.

The checks most inspections skip

A generic third-party inspection checklist was written for consumer electronics, and it will pass a case that the customer rejects. These six are specific to this product:

1. Fitment on a real handset. Not a dummy model. Push the case on and take it off ten times; record whether the rim whitens at the stress points.
2. Button actuation. Every button, every sample. A mushy volume key is a major defect that no visual checklist catches.
3. Camera opening concentricity. Measure the offset between the opening centre and the lens centre; a step that shows as a shadow in photographs is a return reason.
4. Print adhesion. A cross-hatch or tape pull on a printed area. This is the only way to catch a batch printed with insufficient primer.
5. Colour against the golden sample. Under a defined light source, ideally with a spectrophotometer reporting Delta E. Eyeballing under the factory's own lighting is not a measurement; the method is set out in colour approval.
6. Lining and inlay adhesion. Microfibre linings that lift at the corner are invisible at inspection and obvious after a month in a bag.

The golden sample itself has to exist before production starts, and how it is approved matters as much as the sample; the sample approval process covers the paperwork that makes it hold up later.

What happens when a batch fails

A rejection is not the end of the order, and the plan should say what comes next before it is needed. Four options are normal, and they cost very different amounts.

1. Sort and re-inspect. The factory screens 100% of the lot and removes the defective units. Acceptable when the defect is visible and the batch is small. Insist on a re-inspection of the sorted lot at the same AQL level, because sorting by eye misses what a checklist would catch.
2. Accept with a price adjustment. Reasonable for a minor cosmetic deviation that will not cause returns. Get the adjustment agreed in writing before the goods ship, not after.
3. Rework. Possible for print and lining defects. Not possible for a dimensional fault in the shell, where the only fix is a tooling change.
4. Remake. Reserved for critical defects. This is where the delivery date is genuinely lost.

The clause that saves the argument is a definition of who decides. Nominate in the purchase order whether the buyer's inspector, a third party, or an agreed third-party laboratory has the final call on colour and fitment. Without that, a failed inspection becomes a negotiation about opinions.

When to inspect

Two points cover most of the risk. A during-production check at roughly 30 to 50% of output catches a systematic fault while it is still cheap to correct. A final random inspection, after 100% of units are packed and at least 80% are cartoned, is the one that decides whether you release payment.

For a first order with a new supplier, an inspection is worth the fee, and it is cheaper than discovering the problem through a factory audit after two failed shipments. For repeat orders, a supplier that can send its own AQL report with photographs on every batch is usually the better partner; treat occasional unannounced audits as the check on that self-reporting.

Two of the assemblies this checklist was written against: Aegis Clear Dual-Layer Magnetic Camera-Control Button for iPhone 18 Pro Max and Aegis Clear Dual-Layer Magnetic for Galaxy S26 Ultra.