Cover: AI-generated conceptual illustration for WEFTSIGNAL, not documentary photography.
The classification is agreed before anyone inspects
Critical, major and minor decide whether a shipment is accepted, reworked or rejected. They look like descriptions of severity; in practice they are a contract. The class a defect falls into is not discovered at the inspection table. It was settled earlier, when buyer and supplier wrote down what each class means for this order. Inspect first and classify later, and every borderline call becomes a negotiation under time pressure.
Define the classes by consequence, not by adjective
Define each class by what the defect does to the product or the wearer rather than by how bad it looks. A class built on consequence can be applied by different people and reach the same answer; a class built on an adjective cannot, because “noticeable” depends on the light, the viewer and the day. Writing consequence into the definition forces the harder question — consequence to whom, and at what point — to be answered in the document.
A common framework runs in three tiers. A defect that compromises safety or makes the product unusable sits at the top. One a customer would reasonably reject on opening the package sits in the middle. One affecting neither function nor saleability but deviating from the specification sits below that. The framework is only as good as the examples attached to it.
Why safety defects earn their own class
Where a defect can injure a wearer — a detached component on an infant style, a sharp edge, a cord in the wrong place — the consequence differs categorically from a cosmetic flaw, and treating it as one more point on a scale understates it. Keeping that tier separate also lets it carry its own rule, typically no permitted count in the shipment. Such a rule must be written for the product and market concerned, because obligations differ by jurisdiction and age group.
Handling the borderline case
A list covering only obvious examples has left the expensive work undone. The borderline case is where disagreement lives, and the document should name it in advance. Two devices help. The first is a tie-breaker: when a defect could sit in two classes, which governs, and who decides if two readers still differ. The second is a physical reference — a sealed sample carrying the defect, or a photograph at a stated viewing distance — so visibility stops being opinion.
Some borderline calls are matters of judgement and will stay that way. That is an argument for naming who holds the judgement, not for pretending it never arises.
The viewing conditions are part of the definition
Visibility depends on light, distance, angle and background. Change any of them and the same defect changes class. A definition calling a flaw major because it is visible has delegated the decision to the room. Stating illumination, viewing distance and time allowed per garment makes the definition reproducible in two rooms, and removes the easiest defence available to a supplier whose goods were rejected.
Where classification meets sampling
Classification and acceptance sampling are separate decisions often collapsed into one. Classification says what a defect is; the sampling scheme says how many of a class a sample may contain before a lot is not accepted. ISO 2859-1 sets out sampling schemes indexed by acceptance quality limit for lot-by-lot inspection by attributes, and attributes are the classified counts a defect list produces. Get the classification wrong and the arithmetic is applied to the wrong quantity.
An illustrative example: the same flaw, two lists
Consider, illustratively, an order whose classification list names a dropped stitch but says nothing about its location. In one reading any dropped stitch is a defect; in another, one inside a seam allowance is not visible and therefore not a defect. The garment is identical in both readings. The count is not, and every figure built on that count inherits the ambiguity.
What a classification list should contain
The list travels with the order and carries the following:
- Classes by consequence: what each tier means for safety, function and saleability, in the buyer’s own words.
- Named examples: defects common to this product and material, each assigned rather than inferred.
- Borderline rule: a tie-breaker for the contested case, and who decides if it still divides.
- Viewing conditions: illumination, distance, angle and time per garment, so severity is reproducible.
- Physical references: sealed samples or photographs carrying the defect at those conditions.
- Zero-tolerance items: anything governed by another rule, with its basis recorded.
- Version control: who may amend the list, and what happens to lots already inspected under it.
Treating the list as a document under control
A classification list is a controlled document whatever else it is: version, approver, date, and it travels with the order it governs. Where a manufacturer runs a certified quality management system, ISO 9001 is the published reference for how such documents and their changes are managed. When a list changes mid-programme, the useful question is which lots were inspected against which version — answerable only if the version was recorded at the time. It sits beside the other instruments an inspection plan assembles, and the same agreement logic governs the colour reference in our fabric colour approval explainer.
Source and further reading
Sources consulted 16 October 2026: ISO — ISO 2859-1, Sampling procedures for inspection by attributes, Part 1; ISO — ISO 9001, Quality management systems: Requirements; CottonWorks — Garment Quality Control. Standards are cited by designation and official catalogue page only, and no acceptance table or defect rate is reproduced here. The three-tier framework, the borderline devices and the worked example are WEFTSIGNAL editorial material, and ISO 9001 is cited for the management of controlled documents rather than for defect classification. Continue exploring quality control coverage and our editorial standards.
