Where the inventory went

On-demand manufacturing promises that nothing is made until it is sold. That promise is true about garments. It is worth asking what happens to the cloth.

Made-to-order has a clean and genuinely attractive idea at the centre of it. A brand holds no stock, makes nothing speculatively, and produces each garment only once a customer has paid for it. Nothing is overproduced because nothing is produced early. The waste the industry is most often criticised for — clothes made for nobody — simply never happens.

As a description of the garment, that is accurate.

The difficulty is that a garment is made out of cloth, and cloth is not made one garment at a time.

Fabric is made in lots, and the lot has a size

A dyehouse does not dye a shirt's worth of fabric. It dyes a bath. Export dyehouses are commonly described by their customers as running three to five hundred kilos of fabric per custom colour — a figure that comes from suppliers rather than from a published tariff, and which no mill I could find publishes for itself.

Set that against a garment, and the arithmetic is something any reader can do. A single-jersey T-shirt is cut from roughly a square metre and a half of cloth once the marker and the wastage are counted. At 180 grams per square metre that is about 270 grams — call it a quarter-kilo a shirt.

So one custom colour, ordered from scratch, commits something in the order of eleven hundred to eighteen hundred T-shirts' worth of cloth before anybody has cut anything.

That is not a rule anyone chose. It is the size of a machine.

Who dyes, measured rather than assumed

It matters where in the chain that bath sits, because it decides whether a factory can absorb the commitment or has to pass it on.

I went through the full public census of Bangladeshi export garment factories — 3,320 of them, mapped with the industry's own associations — and counted which processes each one performs in-house.

Nineteen factories in twenty do not dye. They buy cloth that somebody else has already committed to a colour, in a quantity somebody else has already paid for.

The same census answers a second question that is usually asked badly. The average Bangladeshi export factory is often quoted at just under nine hundred workers. The median is 402 — the mean is 2.2 times the median, because a long tail of very large plants pulls it upward. Over half of these factories employ fewer than five hundred people. The typical floor runs seven sewing lines and three hundred machines.

A factory of that size is not the obstacle people imagine when they think about small orders. It is also in no position to fund a warehouse of speculative cloth.

So the inventory did not disappear. It moved.

This is the whole of the argument, and it is not a clever one — it is arithmetic.

If a garment is made only when it is sold, but the cloth it is made from must exist before the order arrives, then somebody is holding cloth against demand that has not happened yet. The stock did not vanish when the business model changed. It moved one step upstream, out of finished garments and into rolls.

And it is worse stock in one specific way. A finished garment that does not sell can at least be discounted to someone who wants that garment. A roll of the wrong colour cannot be discounted to anybody until it has been made into something — which requires deciding what, which is the very decision the model exists to defer.

The relocation has an owner. The only question worth asking is who.

Three answers, and what each one means

"The factory holds it." Then a supplier is financing a customer's working capital, against orders that arrive one piece at a time and might stop tomorrow. Suppliers do sometimes do this — for buyers they have known for years, at a price that reflects it. What they will not do is take that risk for a new account on a purchase order. This is the honest reason a small-quantity price does not come down: the smallness is not the cost, the commitment is.

"We hold it ourselves." This is the coherent answer, and it is what the successful operators actually do. Look at who genuinely manufactures at a minimum of one and a pattern appears that is hard to miss: they are almost all their own brand, or vertically integrated, or both. One well-known Nordic made-to-measure shirt business owns its factories outright. The Chinese mass-customisation houses held up as the industry's proof of concept built their own systems over more than a decade, at a cost reported in the tens of millions of dollars. Several European made-to-measure houses have done this for generations with their own cloth positions.

What you will struggle to find is a contract manufacturer offering made-to-measure at a minimum of one to third parties. That absence is usually read as a gap in the market. I think it is the market stating plainly who is willing to carry the risk.

"Nobody — we buy per order." Then the sums have not been done yet, and finding that out early is a kindness rather than an obstacle.

A note on the machine, because it is usually the first thing asked about

Enquiries of this kind almost always specify the equipment: a single-ply cutter, automated, fed by a digital marker.

Single-ply cutting is not exotic. A cutter working to a printed marker with a hand tool cuts one ply every day of the week — it is what a sample room does, and by the census above four factories in five already cut in-house. An automated single-ply machine is a real and useful thing, and it is a genuinely different machine class from the multi-ply cutter most export floors run: a different knife, a different bed, a fraction of the vacuum. What it buys is speed.

But speed only matters above a volume. A hundred individual garments in a month is hand work. A hundred a day is a serious operation — difficult, but a question of organisation rather than of possibility. When a buyer specifies "highly automated", they have usually inferred the machine from a throughput they have not yet stated. Ask for the monthly volume and the machine question tends to answer itself.

There is one cost of one-at-a-time work that no machine removes. A cutting marker laid for a single garment uses measurably more cloth than one laid for two. In a study of 258 markers for men's shirts on 1.5-metre cloth, a one-garment marker consumed 1.1054 m at 80.4 % efficiency, while a two-garment marker consumed 1.0206 m at 87.1 % — about 8 % more fabric per garment, rising to nearly 10 % against the best marker in the set.

Almost the entire saving arrives at the second garment, not the fiftieth. The same study found the number of sizes on a marker explained only about 3 % of the variation in fabric use. It is not mixing sizes that costs cloth. It is having only one garment on the marker. And that difference is paid by whoever owns the cloth.

What to ask instead of "what is your minimum?"

Minimum order quantity is the wrong first question, and it produces a number that is nearly always wrong in both directions. There is no national minimum anywhere. There is a triangle — quantity, standards, price — and fixing any two moves the third.

For made-to-order specifically there is a better opening question, and it is not technical:

Whose fabric is it, and who pays for it before it is sold?

If the answer is "the factory", then what is being discussed is not an order. It is a partnership, and it should be written as one: a committed range of qualities and colours, over a committed period, against which someone can rationally hold stock. That is a normal commercial arrangement and there is nothing awkward about asking for it.

If the answer is "we do", the conversation becomes a straightforward sourcing problem — which qualities, which colours, how deep — and that is a problem with known methods.

Either way it is a better place to start than a machine specification, and it is a great deal cheaper to discover in an e-mail than after a sample.

— A.K. · TB Textile Sourcing · Slovenia, EU
(European who ran knit factories in Bangladesh for seven years. The factory figures are counted from the public census of 3,320 Bangladeshi export factories; the marker figures are from a published study, read at source.)