A Defect By Any Other Name Ruins Your Data
Ask three checkers to name the same flaw and you may get broken stitch, needle cut, and damage. Three names, one defect, and now the data says the factory has three small problems instead of one significant one. The defect taxonomy, the controlled list of defect names and their definitions, is unglamorous infrastructure, and half of systematic quality improvement depends on it.
Why names decide what you can fix
Improvement runs on Pareto: rank defects by frequency and cost, attack the top. The ranking is only as good as the naming. Synonyms split one big problem into invisible fragments; vague catch-alls (stitching problem) merge distinct causes into an unfixable blob; and inconsistent major/minor classification corrupts both internal metrics and AQL outcomes, because your checkers and the buyer's inspector are playing by different rulebooks.
A working taxonomy has three properties. Each name maps to one physical phenomenon with a definition and ideally a photo. The list is bounded and versioned: additions are deliberate, not typed ad hoc at the endline table. And each entry carries its default classification and, where useful, its likely operation of origin, so recording a defect simultaneously routes the signal.
One list, every gate
The taxonomy earns its keep by being universal: the same names at inline patrol, endline count, finishing check, and final inspection, and the same names in the dyehouse and fabric inspection for their domains. That is what makes numbers joinable: endline trends predict final results only if both speak one language; rework minutes attribute to operations only if the defect that caused them is named consistently. A factory with four gates and four vocabularies has four opinions instead of one dataset.
Governance, lightly
Someone must own the list, a quality manager with the authority to merge synonyms and retire dead entries, and changes should be versioned so historical data stays interpretable. The touch is light: a quarterly review, driven by the other-category, which is the taxonomy's pressure gauge. When other grows past a few percent, the list is missing a name the floor needs; when it stays small, the vocabulary fits the reality.
The payoff arrives quickly
Taxonomy work sounds bureaucratic and pays within weeks: the first honest Pareto usually surprises everyone, because a defect that had been living under three names steps forward at its true size. The fix that follows was always available. What was missing was the name.