Filling the garment: the dose, the drift and the tolerance

The filling room

Bo Ha's filling room is a sealed cell with its own air handling. It holds three down-blowing machines and four operators. A machine is a hopper, an electronic scale and a nozzle. The operator offers up a chamber. The machine weighs a dose and blows it in. Then the chamber is closed with a chain stitch, which is unpicked and re-sewn if the weight is wrong.

Nothing about this is exotic. What is unusual is what the machine is dosing. It is the most expensive material in the garment, one chamber at a time, and once the chamber is shut nobody can see inside it again. Every other material in a parka can be checked on the finished garment. This one cannot.

The fill schedule, not the fill weight

The tech pack says 210 g. The filling room cannot use that. What it needs is a schedule.

Two things fall out of writing it down. The chamber count is 28. So the dose is set 28 times per garment, and 207,200 times across this order. And the doses are not equal, because the chambers are not equal. A sleeve chamber that receives a body dose becomes a stiff tube that will not bend at the elbow.

Grade the schedule, not the total. Fill follows the surface area you have to insulate, not the size step. The chamber count usually stays the same across the size range, so what changes is the dose in each chamber.

The average parka on this order carries 215.8 g, not 210 g. A buyer who orders down as 7,400 multiplied by 210 g buys 1,554.0 kg. That is 42.9 kg short before a single gram is lost in the process — USD 6,349 of down, and a filling room that stops. Now add the 2.5% Bo Ha loses to the filters, the hopper and the reclaim bin. The real purchase is 1,636.8 kg, USD 242,246. The cost sheet's USD 31.08 was the size M's fill. The order's true average is USD 0.86 a parka above it.

The dose drifts in two different ways, and only one of them is visible

The scale on a down-blowing machine holds a tolerance. Bo Ha's are set to plus or minus 0.3 g a dose. Ask what that does to a garment. The answer depends entirely on whether the error is random or systematic.

Random. Each of the 28 doses is a little high or a little low, independently of the others. The errors partly cancel each other out. Independent random errors combine as the square root of the sum of their squares, so the spread on the finished garment is 0.3 multiplied by the square root of 28. That is about 1.6 g, which is 0.8% of the fill. Invisible, harmless, and not worth a meeting.

Systematic. The scale reads 0.3 g light on every dose. Now nothing cancels. The garment is 8.4 g over, which is 4.0% of the fill, and it is over on every parka that machine touches. Across the order that is just over 62 kg of down given away — USD 9,200.

The same 0.3 g. A factor of more than five between the two outcomes. The difference is not the size of the error. It is the structure of the error.

Bo Ha's own habit was to aim at 215 g on the M "to be safe". That is 5 g a garment: 37.0 kg across the order, USD 5,476. The factory pays it, out of a factory margin of USD 7.80. And it stays invisible until somebody weighs the down purchase against the fill schedule.

Underfill is the expensive one, and it is silent

Overfill costs the factory money. Underfill costs the buyer the product.

A chamber that is 20% light does not look 20% light. It looks like a slightly softer patch on a garment that is otherwise correct. On a hanging parka it is not visible at all. It becomes visible on a cold hillside, as a cold band across the lower back. That is the sort of complaint that arrives one garment at a time and is never traced back to a chamber.

This is why fill weight tolerance is a commercial term and not a quality note. Corrievreckan's specification carries it as one: fill weight plus or minus 5%, measured on the garment, and a lot outside it is a lot they may reject. On a 210 g fill that band is plus or minus 10.5 g, so it is 21 g wide. At USD 148.00 a kilogram that band is USD 3.11 of down a parka. It is worth more than the seam tape and more than the baffle mesh. Write a fill tolerance the way you would write a price tolerance, because that is what it is.

The mistake nobody made

On 4 August the filling room re-calibrated all three machines against a certified weight set, signed the log, and ran the day's output at exactly 210.0 g on the M. On 20 August, Corrievreckan's laboratory opened three parkas from the shipment sample and reported the fill at 191.9 g.

Nobody had done anything wrong. Down is a keratin material, so it takes up moisture from the air. That is why every published test method conditions its specimen before it weighs anything — it holds the sample at a set temperature and humidity until the moisture settles. Bo Ha's filling room is air handled for dust, not for humidity. And August in Hai Duong is not February.

The laboratory dried 100.0 g from each lot until the weight stopped falling. The February lot lost 5.1 g. The August lot lost 8.6 g.

• A 210.0 g dose from the February lot delivers 199.3 g of dry down.
• The same 210.0 g dose from the August lot delivers 191.9 g.
• That is 7.4 g less of the material that lofts. It is 3.7% down on the approved sample, from a scale that was correct, an operator who followed the schedule, and a specification that said 210 g.

The fix is not a better scale. It is to stop treating a weight as if it were a quantity of down:

1. Test each lot's moisture on arrival and post the figure in the filling room.
2. Dose against the reference condition the specification names. To deliver February's 199.3 g of dry down out of the August lot, you must weigh 218.0 g, not 210.0 g. That is 8.0 g more, USD 1.18 a parka. It is not a cost overrun. It is the cost of the down you actually specified.
3. Say so in the contract. The wool trade has bought and sold on conditioned weight for a very long time, for exactly this reason. Down is sold on as-received weight, so the moisture is yours to manage.

Act as an outerwear technologist and costing engineer. Build me the fill file for one down garment. Order facts: buyer [BUYER], PO [NUMBER], style [STYLE], quantity [QTY], FOB [PRICE], ship [DATE]. Fill specification: [FILL POWER] fill power, composition [90/10 OR OTHER], [WEIGHT] g in size [SIZE], measured by [METHOD IF STATED]. Construction: [SEWN-THROUGH / KARO-STEP / BOX WALL], baffle wall height [MM], chamber count [N], chamber line length [M], insulated surface area [M2]. Size break: [LIST SIZE AND QUANTITY]. Down prices available to me, per kilogram: [LIST FILL POWER AND PRICE]. Do the following, showing every division. First, convert the specified fill weight and fill power into a total free loft volume, and restate that as the design target. Second, build a ladder of every fill power I listed at that same loft: weight needed, price per garment, finished garment weight. Third, compute the chamber volume from the wall height and the insulated area, and give the ratio of free loft to chamber volume; tell me whether the chamber is over-packed, firm, or under-filled, and what fill weight would restore a firm chamber if the wall height changed by plus or minus 5 mm. Fourth, grade the fill across my size break and give me the total kilograms the order really needs, against the naive figure of quantity multiplied by the size M fill, and name the difference in money. Fifth, add a process loss allowance and state the number you used. Sixth, write me the three questions I should put to the down supplier before I place the order. Do not give me a range where a number is possible, and list every assumption at the end.

What to do on Monday

1. Ask for the fill schedule, not the fill weight. A number with no chamber breakdown behind it cannot be dosed, checked or argued about.
2. Buy down against the graded schedule and the process loss, never against the size M multiplied by the order quantity.
3. Ask which way the filling room drifts, and who is checking. The answer tells you who is paying.
4. Put a moisture test on every lot and write the reference condition into the specification.