Seam sealing is a machine, not a step

The line on the tech pack

KV-3620's construction page carries one instruction that reads like a detail: "All shell seams within the waterproof envelope to be sealed with 22 mm tape."

It is not a detail. It is a second production process, with its own machines, its own operators, its own consumable, its own minimum order quantity and its own failure modes. On this order it also has its own capacity ceiling. Course 2.4 owns the laminate and course 16.6 owns the claim on the hangtag. What is yours is the operation that stands between them.

The machine, and the speed it actually runs at

A hot-air sealing machine blows heated air into the fold between the tape and the seam, then presses the two together between rollers. Three things are set: nozzle temperature, air volume and roller speed. They are set together, because the bond is an amount of heat delivered per centimetre. Change the speed and you change the temperature the seam sees.

Bo Ha's machines are rated at 8.0 m/min. Ask the industrial engineer what they run at on this garment and the answer is 3.6 m/min.

The gap is not a maintenance problem. Nameplate speed assumes a straight, long, flat seam. A parka's shell seams are short, curved and full of intersections, so the operator spends most of the cycle presenting, aligning, starting and stopping. The number that matters is metres per minute at the garment, measured on this style, and it is roughly half the brochure. Ask for it before you plan anything.

From metres to machine-days

KV-3620 as designed has 14.8 m of taped seam a garment.

SMV is the standard minute value: the work content of the job in minutes. The sewing window for this order is 24 days. The specification as drawn needs 26.2 days of taping on every machine Bo Ha owns. No amount of overtime on the sewing floor helps, because the constraint is not people.

Bo Ha priced a fifth machine at USD 26,000 with a nine-week lead time. Nine weeks was longer than the time remaining. So the only lever left was the specification.

The seam review on 14 May asked one question of every taped seam: is this seam inside the waterproof envelope? Four were not — two pocket bags behind a storm flap, the inner hood seam and the hem facing. Removing them took the taped length from 14.8 m to 11.2 m, and the order from 26.2 machine-days to 22.4, which fits.

Now price what that review saved:

• Tape: 3.6 m a garment at USD 0.19 is USD 0.68.
• Labour: 1.00 minute at USD 0.086 is USD 0.09.
• USD 0.77 a parka, USD 5,698 across the order.

The money is real and it is beside the point. What the review actually bought was 3.8 machine-days that were not for sale at any price. In this category the binding constraint is routinely a machine-hour rather than a dollar. A merchandiser who reads a taped-seam specification as a cost line will not see it coming.

The tape is a material with a minimum

Seam tape is a laminate with a heat-activated adhesive, and the adhesive is matched to the fabric it has to bond to. Change the shell laminate and the approved tape may not bond to the new one. That is the material version of the rule course 16.6 sets out: a specification change re-opens everything that was validated on the old specification.

Tape is also made to order. Corrievreckan's third colourway, Tussock, uses a printed shell. The tape that bonds to the plain laminate does not bond reliably to the print, so Tussock needs a second tape. The Tussock run is 260 parkas.

• Tape needed: 260 multiplied by 11.2 m, plus 8% for wastage and set-up, is 3,145 m.
• The supplier's minimum order quantity is 5,000 m.
• So you buy 5,000 and use 3,145. The surplus of 1,855 m at USD 0.19 is USD 352.45, and it belongs to 260 garments: USD 1.36 a parka, 1.6% of the FOB, on that colourway only.

Nobody made an error. A colour decision made in a design meeting became a materials decision with a minimum attached. The arithmetic only appears if somebody asks whether the tape is the same tape.

How a taped seam fails

Four ways, and they do not look alike:

1. Cold bond. Nozzle temperature low or speed high. The adhesive never wets the fabric. The tape looks fine and lifts at a corner after a few washes.
2. Over-temperature. The opposite setting damages the membrane under the tape. The seam looks perfect, the tape is firmly stuck, and the seam leaks through a membrane that has been cooked.
3. Bridging at an intersection. Where four layers meet, the tape tents over the step and leaves a channel straight through the garment. Almost every leak in a taped garment is at a crossing, because that is the only place the tape is asked to cross a change in thickness.
4. Curves. Tape does not want to lie on a tight radius. Armholes, hoods and the small sizes of any garment are where this shows.

Two of those four pass a visual inspection without any difficulty. That is why a taping line runs its own verification: a peel check and a water-column check on a sample seam, every shift and after every setting change, on the same fabric as the bulk. Course 6.3 owns the methods. What belongs to the floor is the frequency, and the fact that a machine setting drifts.

One more, from the sewing room rather than the taping room. Tape 22 mm wide over a 6 mm seam allowance leaves 8 mm of adhesive on each side of the seam. Over a 9 mm allowance it leaves 6.5 mm, which is 19% less bond width — from a seam allowance still inside most factories' idea of acceptable. The tape's performance is set by an operator who is not in the taping room.

The mistake nobody made

12 August, final inspection. Lot 7,400 parkas, general inspection level II, sample 200, AQL 2.5, accept on 10. AQL is the acceptable quality limit: the defect rate the sampling plan is built to pass. The water test on the taped seams found 3 failures. Three against an accept number of ten. The lot passed, correctly, and course 6.2 owns why that plan is the right plan.

Corrievreckan's QA manager, Malcolm Reith, asked one further question: which garments?

All three were XS.

Both columns are arithmetically correct, and they give the same number of leaking parkas. They are not the same fact. One says there is a small, tolerable background rate spread thinly across a shipment. The other says roughly one XS parka in five leaks, and that is a size the retailer will pull from sale.

The cause was the fourth failure mode above. The XS armhole has the tightest radius in the size range, and at the machine's set speed the roller could not follow it. The defect goes with size, and an inspection sample drawn at random across a lot is not designed to say anything about a subgroup inside it. A size-linked defect is invisible to a sample that ignores size. Nothing is wrong with the sampling plan. It is answering the question it was asked.

The fix was a speed reduction and a slower feed on XS only. It cost 0.4 minutes on 518 garments and nothing on the other 6,882.

Act as an industrial engineer in a technical outerwear factory. I need to know whether this order fits the window, not what it costs. Order facts: style [STYLE], quantity [QTY], size break [LIST], ex-factory [DATE], sewing window [DAYS], working week [DAYS]. Sewing line: [N] operators, [MIN] minutes a day, line efficiency [PERCENT]. SMV, the standard minute value, meaning the work content of the garment in minutes: [PASTE THE OPERATION BULLETIN, OR GIVE THE TOTAL SMV AND THE OPERATIONS THAT ARE NEW TO THIS LINE]. Seam sealing: [M] of taped seam a garment, measured machine speed [M PER MIN] on this style, handling time a garment [MIN], [N] machines available, [HOURS] a shift. Do the following. First, convert the sewing line's earned minutes a day into pieces a day using THIS style's SMV, and give me the steady-state line-days for the order. Second, add a learning ramp based on the operations that are new to the line, in blocks of five days, and give me the realistic line-days. Third, work the taping separately: metres a garment divided by measured speed, plus handling, times quantity, into machine-hours, machine-days and days on the machines available. Fourth, tell me which of the two is the binding constraint and by how many days, and back-calculate the date cutting must start. Fifth, if it does not fit, price four levers — a seam review that removes seams outside the waterproof envelope, an extra machine with its lead time, overtime, and a second line with its own ramp — and say what each one buys in DAYS, not only in money. Sixth, state the capacity I am consuming in earned minutes so I can price it against what the line was booked for. Name every assumption.

What to do on Monday

1. Convert every taped-seam specification into machine-days before you confirm a delivery date: metres a garment, measured speed, machines available.
2. Ask for the measured speed on your style, never the machine's rating.
3. Ask whether every colourway uses the same tape, and get the minimum order quantity before the colour range is signed.
4. Record the size against every seam defect and read the failure rate inside each size, not only across the lot.