Defects that are the product, and the ones that are not

The situation

21 July, 09:40. A third-party inspector is at Ulvan with the first shipment. He pulls a 200-piece sample, works through it in four hours, and logs 47 major defects. A major defect is a fault a customer would notice and complain about. Tarnwick's plan accepts 10. The lot is failed and the report goes to London the same evening.

Forty-one of the forty-seven are the whiskers, the sanded seat, the ground hems and the destroyed patch above the knee. They are the product. They are on the tech pack, they are on the sealed garment, and the brand paid Sarband a dollar thirty-two a pair to put them there.

The problem is not the inspector

He was right to log them, given what he was given. His list is the general apparel defect list every inspector carries: abrasion, thin place, hole, colour variation within one garment. Every one of those words describes something real on a TK-905. Nothing in his brief tells him how to separate an abrasion that was bought from an abrasion that happened.

And the fix that gets reached for first is worse than the problem. Somebody rings and says the garment is meant to look worn. For the rest of the inspection, every abrasion, every thin place and every colour variation is now a feature. A skipped stitch inside a sanded panel is invisible to the process. That matters more here than anywhere else, because abrasion opens seams up. A skip that a raw jean would hide is a hole on a washed one.

Both readings come from the same absence. On a garment that is deliberately damaged, the defect list has to be style-specific, or it cannot be applied at all.

The annex, in three columns

One page, attached to the PO, and written before the first shipment rather than after it. Three columns, because there are three genuinely different kinds of judgement on this garment.

Column one: deliberate, with limits. Every effect is named, positioned from a seam, and given a tolerance. The effect is not "as sealed garment". It is a dimension like any other:

• Whiskers: six lines each hip, 12 to 17 cm long, starting within 1.5 cm of the fly seam.
• Seat and thigh sanding: inside the boundary drawn on the sealed garment, not beyond it.
• Hem grind: the full circumference, no interruption longer than 2.0 cm.
• Permanganate cast: front thigh only, no overspray onto the back panel.
• Destruction: one patch, 5.0 by 1.5 cm, centred 14.0 cm above the knee point, tolerance 2.0 cm, bar-tacked border intact.

Column two: always a defect, whatever the look. These are the ones the meant to look worn instruction destroys:

• Abrasion outside a drawn zone, at any intensity.
• A hole without a bar-tacked border, anywhere.
• A skipped or broken stitch inside an abraded zone.
• Grit or stone fragments in seams, pockets or turn-ups.
• A rivet burr, a sharp shank end, a loose rivet.
• Backstaining on pocket bags or the topstitch below the agreed grade.
• Colour transfer on rubbing below the agreed grade.
• Shade outside the sealed band.

Column three: checked on a record, not on a garment. This column has no equivalent on an unwashed product. It is the reason a denim final inspection is not only a garment inspection:

• Residual chlorine and residual permanganate, tested on the garment with potassium iodide starch paper, one reading per load, signed.
• Enzyme deactivation confirmed at the end of each recipe.
• The load weight sheet, so you can see whether any load ran single-size or under weight.
• The shade comparison for each block against the sealed standard, dated.

The chemistry underneath the first two is course 3.6's third lesson. What belongs here is the inspection consequence: these are defects that have not appeared yet. A jean with chlorine left in it passes every visual check on the day. Five weeks later, in a warehouse, it fails a tensile test — the pull test that measures the force cloth takes before it tears. You cannot see that on a garment. So you inspect the record, or you do not inspect it at all.

The recount

Against the annex, the same 200 pieces:

The sampling arithmetic itself belongs to course 6.2: how the sample size and the accept number are arrived at, and what a switching rule does to them. None of that changes here. What changes is what counts as a defect before you start counting. That question sits upstream of every sampling decision, and it is not part of any sampling standard.

The measurement that depends on when you took it

There is a second disagreement in the same shipment, and it has the same shape. Ulvan's own quality office measured the first loads on 2 July and reported the waist at plus 0.60. Tarnwick's technologist measured on 6 July and reported plus 0.84, against a tolerance of 0.8. One reading passes, one fails, and both people are honest.

Ulvan measured twenty minutes out of the dryer. A jean is at its shortest hot and dry, and it recovers as it takes moisture back in. On TK-905 the difference is 0.24 cm at the waist. So the annex carries a measurement protocol as well as a defect list: measured flat and relaxed, at least four hours after drying, in a conditioned room. Without it, the two offices will produce different numbers for ever, and neither has any evidence the other is wrong.

Shade is packed, not inspected

The last thing that is different here is what you do with variation you have already accepted.

A sealed band is a range, and every load inside it is passed. But a customer does not see the band. They see two garments on one rail. Across 53 loads the order will drift from one limit toward the other. If cartons are packed in production order, a store receiving three cartons receives three positions in the band.

So the loads are sorted into blocks — light, middle, dark. A carton is filled from one block, and a store allocation is filled from one block. Done at the hanging rail before packing, it costs about 0.28 minutes a pair, or USD 241.08 across this order. Done nowhere, it costs a rail in a shop that looks like two deliveries. That is the complaint a brand's retail team raises rather than its quality team, so it arrives late and from a direction nobody was watching.

Act as a quality manager who has run final inspections on washed denim for fifteen years. Write me a one-page inspection annex for a style with deliberate damage, so that an inspector who has never seen my tech pack can apply it. Style facts: buyer [BUYER], PO [NUMBER], style [STYLE], [QTY] pairs, sealed garment dated [DATE]. The wash effects specified are: [LIST EVERY ONE — WHISKERS, SANDING, GRINDING, DESTRUCTION, PERMANGANATE SPRAY, TINT, 3D CREASES — WITH ANY POSITION OR SIZE THE TECH PACK GIVES]. The recipe includes [SAY WHETHER IT USES CHLORINE BLEACH, PERMANGANATE OR ENZYMES]. The buyer's fastness and physical minimums are [PASTE THEM]. Do the following. First, produce column one: every deliberate effect, positioned FROM A NAMED SEAM OR LANDMARK, with a size and a tolerance, so it is a dimension and not an impression. Say where you had to invent a limit because the tech pack does not give one, and mark those for the buyer to confirm. Second, produce column two: what is a defect whatever the garment is meant to look like, including the ones abrasion makes worse rather than hides. Third, produce column three: the checks that are made on a RECORD rather than on a garment, because the defect has not appeared yet — name the record, who signs it and how often. Fourth, write the measurement protocol: when after drying, in what state, in what atmosphere, and how many pieces. Fifth, write the shade instruction: how loads are blocked, how a carton is filled and how a store allocation is filled. Sixth, tell me who this annex has to be sent to by name, and when, and why sending it after a failed inspection is much weaker than sending it before. Keep it to one page and write it as instructions to an inspector, not as prose.