Inspecting what you cannot see

The category's own problem

Every material in KV-3620 can be checked on the finished parka except the one that costs USD 31.08. The shell can be handled. The tape can be peeled. The zip can be run. The measurements can be taken. The down is inside a closed chamber, and there is no way to look at it that leaves the garment saleable.

Course 6.2 owns AQL sampling and the final inspection, and everything it says applies here unchanged. AQL is the acceptable quality limit: the defect rate the sampling plan is built to pass. What this lesson adds is the part of the garment a final inspection structurally cannot reach, and what you do about it.

The check everybody runs, and its blind spot

The standard check that does not damage the garment is to weigh it. It works because the total is the fill plus everything else:

Read the bottom right. A garment scale cannot see a fill error smaller than 16.5 g, because anything smaller is inside the combined tolerance. And 16.5 g is 7.9% of the fill, while the tolerance the contract turns on is 5%.

So the check everybody runs is blind exactly where the commercial term bites. It will catch a catastrophe. It will not catch the thing you wrote the tolerance for.

The destructive test, priced

The only real measurement of fill weight is to open a garment, remove the down, condition it and weigh it. Corrievreckan's specification calls for 5 garments per 1,000, which on this order is 37 parkas.

• Garments destroyed: 37 at Bo Ha's cost of USD 76.70 is USD 2,837.90.
• Laboratory: 37 tests at USD 32.00 is USD 1,184.00.
• USD 4,021.90 in total, USD 0.54 a parka — 1.7% of the down bill.

That is the whole decision, and it is a comparison rather than a judgement. The test is worth running if it catches a fill error bigger than 1.7%. The dosing bias in lesson 2 was 4.0%. The moisture drift was 3.7%. The failure below is 6.2%. A destructive sample is one of the few tests in apparel that clears its own cost several times over on ordinary, unremarkable errors.

Get it into the contract at order confirmation, and name who pays for the destroyed garments. Argued about in August it becomes a dispute. Agreed in February it is a line item.

The mistake nobody made

12 August, final inspection at Bo Ha. Every parka in the 200-piece sample weighed inside 622 g plus or minus 16.5 g. The lot passed on visual defects at AQL 2.5. The shipment was released.

On 20 August, Corrievreckan's laboratory opened three parkas and reported the fill at 197 g against a specified 210 g. That is 6.2% short.

Here is the chain, and every link in it is correct.

1. The shell fabric was delivered at the top of its weight tolerance: plus 4.2 g a garment. In tolerance.
2. The lining, likewise: plus 3.5 g. In tolerance.
3. A zip tape and a hood laminate were substituted after a supplier change, both approved: plus 5.3 g.
4. Make-up therefore rose from 412 g to 425 g. Nobody measured it, because make-up weight is not a specified quantity. It is whatever is left over.
5. Bo Ha's QC manager had introduced a sensible shortcut in June. Instead of checking the dose at the machine, check the finished parka against 622 g. It is faster, it damages nothing, and it is the only whole-garment check available.
6. The filling room hit 622 g. To do that with a 425 g make-up, the dose has to be 197 g. 425 plus 197 is 622.

Nobody cheated. Nobody was outside a tolerance. The down not delivered is 13 g a parka — USD 1.92 a garment, 96.2 kg, USD 14,238 across the order — and a garment that carries a hangtag claiming a fill weight it does not have.

The check controlled the wrong variable. Total weight is the sum of two quantities and only one of them was specified. So holding the sum constant pushed every drift in the other one straight into the fill. Put that beside lesson 2 and the pattern is worth naming: the same symptom, a parka short on down, arrived twice from two entirely unrelated causes, and neither was visible to the check the trade actually runs.

The fix is cheap and it is upstream:

• Weigh the assembled shell and lining before filling, once a shift. That is the make-up weight, and it turns a leftover into a measured quantity.
• Check the dose at the machine, not the garment at the end. The scale on a filling machine can log every dose. A day's log shows you the spread, and a spread shows you a bias on the first day rather than in the laboratory report.
• Keep the garment weight check. It is still the fastest way to find a catastrophe. Just stop treating it as a fill check.

Three more checks this category needs and the others do not

Down leakage. A chamber holds down only if the cloth does. It is tested on the fabric before the order, with a rubbing test that counts what comes through. It should be repeated on the made garment, because the needle holes are not in the fabric test.

Needle detection, and why it is stricter here. In a woven shirt a broken needle fragment is found, cut out, and the garment is repaired. In a filled chamber it cannot be retrieved, so the garment is scrapped at USD 76.70. That changes the economics of needle control completely. The detector belongs after filling as well as after finishing, and in a down room a broken-needle procedure has a scrap cost attached to every failure to follow it.

Measuring a garment that changes shape. A woven shirt has one chest measurement. A lofted parka's chest depends on how it was laid, whether it was pressed, and how long it has been out of the carton. Two inspectors measuring the same parka 15 mm apart is ordinary, and the tolerance is plus or minus 10 mm. So the disagreement is bigger than the tolerance, and neither inspector is wrong.

The specification therefore has to carry the method, not just the number: lay flat, do not press, let the garment recover, measure at the seam, at a stated point. Course 6.2 owns the measurement audit. What this category adds is that the method is part of the measurement, and a point-of-measure list without one is not enforceable on a garment that lofts.

Act as a quality manager for filled outerwear. Build me an inspection plan for a garment whose fill cannot be inspected without destroying it. Facts: style [STYLE], quantity [QTY], fill [WEIGHT] g with a tolerance of [PERCENT], make-up weight [WEIGHT] g with a tolerance of [GRAMS], factory cost of one garment [AMOUNT], laboratory fill test [AMOUNT] each, down price [AMOUNT] per kilogram. Do the following. First, compute the finished garment weight and its combined tolerance, and tell me in grams and as a percentage of the fill what the smallest fill error is that a garment scale CANNOT detect. Say plainly whether that blind spot is bigger or smaller than the contractual fill tolerance. Second, price a destructive fill test at [N] garments per thousand, including the destroyed garments, and express it per garment and as a percentage of the fill cost, then tell me the size of fill error above which the test pays for itself. Third, list the upstream checks that would catch the same errors without destroying anything, and say what each one assumes. Fourth, list the checks this category needs that a general final inspection does not run, and where each one belongs in the process. Fifth, tell me what a point-of-measure list must say about METHOD for a garment that lofts, and give me the wording. Sixth, name the three things that must be agreed in the contract at order confirmation rather than argued about at inspection. Do not invent a standard or a clause number; where a test method is needed, say what the test does and leave me to name the standard.

What to do on Monday

1. Stop calling the garment weight check a fill check. Know its blind spot in grams and say it out loud.
2. Agree the destructive sample, and who pays for the garments, at order confirmation.
3. Measure the make-up weight, so that total weight stops being one known quantity plus one unknown.
4. Write the measuring method into the point-of-measure list for anything that lofts.