An inspection report is only as good as the defect classification list behind it — the agreed definitions of what counts as critical, major and minor. Without one, pass and fail are negotiation outcomes, not measurements. This article defines the three classes with practical examples and shows how to write a classification standard your factory signs and your inspectors apply. It is written for sellers and buyers running AQL-based inspections.
Why classification comes before sampling
AQL arithmetic counts defect classes separately: critical defects against a zero tolerance, major defects against one accept/reject number, minor defects against another. That means the same physical flaw can pass one shipment and fail another depending entirely on which bucket the inspector places it in — and buckets are a matter of definition. If the definition lives in the inspector's head, or in the factory's head, your inspection standard is a mood. The classification list is the document that makes pass/fail reproducible, and it applies at every inspection gate in the sequence described in our inspection types guide.
The three classes, defined
| Class | Working definition | Typical examples | Handling |
|---|---|---|---|
| Critical | A safety, regulatory or legal failure — or a defect that makes the product unsellable as a matter of law or basic safety | Sharp edges or points, exposed wiring, a missing mandatory warning, a needle or metal fragment in a textile, mold contamination | Zero tolerance: one critical defect fails the lot |
| Major | A defect a reasonable customer would notice and act on — return, complaint, refund; affects function, fit or prominent appearance | Product does not power on, a zipper fails after a few cycles, wrong color against the approved sample, a visible stain on a front panel, missing accessory | Counted against the major AQL — commonly 2.5 |
| Minor | A workmanship or cosmetic deviation most customers would accept at the price point | An untrimmed thread end inside a seam, a faint scuff on a non-visible surface, slight glue residue, a small carton dent on outer packaging | Counted against the minor AQL — commonly 4.0 |
Category examples: where the lines actually fall
Classification is category-specific. The same physical deviation can be minor on one product and major on another, which is why borrowed lists fail and category-specific ones work:
| Deviation | Apparel | Consumer electronics | Home goods |
|---|---|---|---|
| Small surface scratch | Minor if hidden; major on a visible panel of a premium garment | Minor on a housing rear; major on a front face or display lens | Depends on finish class — judged against the golden sample |
| Slight color deviation | Major against the approved swatch; minor if within an agreed tolerance band | Minor for housing tone; major if parts of one product mismatch each other | A set with mismatched pieces is typically major |
| Function degradation | Almost always major — seam strength, zip cycles | Major at minimum; anything affecting safe operation escalates toward critical | Major for anything load-bearing or food-contact adjacent |
Writing the standard: golden sample plus defect library
- Sign the golden sample. The approved reference sample, signed and dated by both sides, stored where both sides can reach it — every later judgment refers to it.
- Set measurement tolerances. For dimensions, weights and colors: numbers with ranges, not adjectives. "Approximately" is not a tolerance.
- Build a defect photo library. Photograph real examples of each expected defect type and assign each a class. Photographs end most classification arguments before they start.
- Get factory sign-off. The classification list is countersigned by the factory before production. A standard the factory never agreed to is a dispute calendar.
- Cite it in every inspection. Each report references the list version and golden sample used, so results stay comparable across shipments and suppliers.
Programs that already run signed sample rounds — private label development above all — will recognize the discipline: the golden sample is not a courtesy copy, it is the legal reference your later inspections enforce against.
Borderline calls and disputes
Even good lists leave edge cases, and three mechanisms keep them small. First, escalation rules: agree that location changes class — a scratch on a visible front face counts at least one class higher than the same scratch on a hidden surface. Second, boundary samples: physical examples of the worst acceptable minor and the least major, kept beside the golden sample, so inspectors compare against objects rather than vocabulary. Third, a tiebreaker path: when buyer and factory genuinely disagree on a borderline classification, a third-party inspection or laboratory judgment resolves it — which is always cheaper than a standoff with goods stuck at the factory and a vessel date approaching.
Classification is also a margin instrument
The classes drive money in both directions. Over-strict classification — treating every cosmetic deviation as major — produces failed shipments, manufactured re-inspections, and factories that quietly pad quotes against expected disputes. Under-strict classification moves the cost to your customers: defects you called minor come back as one-star reviews and returns. The defensible standard sits at your price point and channel reality — what your actual customer notices, tolerates and returns.
Our AQL guide covers how the counts become accept/reject decisions; the gate system on the quality and compliance page shows where the classification list enters the workflow — agreed at pre-production, applied at every inspection thereafter.
Frequently asked questions
Who decides the classification — buyer, factory or inspector?+
The buyer sets the standard, the factory signs it, and the inspector applies it. An inspector classifying without a written, signed list is improvising — and improvisation does not reproduce across shipments or across inspection companies.
Can a defect change class between orders?+
The list itself should be stable; escalation rules handle context such as location, visibility and product variant. If the same deviation keeps being reclassified, the list needs a new photograph and a decision — not another round of ad-hoc judgment.
What is the most common classification mistake?+
No defect photo library. Written definitions drift in translation and across inspectors; photographs of real examples from your own production keep both sides calibrated to the same boundary.
How does classification interact with AQL numbers?+
Directly: counts are tallied per class, and each class is judged against its own accept/reject numbers, with critical judged against zero. A lot can pass on majors and fail on minors — or fail outright on a single critical defect.
Should classification differ between retail channels?+
It can. A premium DTC channel justifies stricter cosmetic classes than a discount channel for the same product. Keep the lists separate and labeled, so inspectors apply the right standard per order.
