Why places become good at one thing

An origin is not good at apparel

Nothing is good at apparel. Knitwear, denim, tailoring, outerwear, lingerie and workwear share a word and very little else. Different machines, different mills, different operations, different minimum quantities, different ways of going wrong. A place that is the obvious choice for one of them can be a poor choice for the next. The single figure that gets quoted about an origin cannot tell you which, because it sums them.

So the useful question about an origin is never "is it competitive". It is "which categories has it built up, and what happens if I ask it for one it has not".

Course 26.6 answers the factory's version of that question. It takes one plant's two years of quotes, splits the win rate by category, and tells the owner which column to stop quoting in. This lesson is the layer underneath: why the columns are where they are, why they move so slowly, and what it costs to sit outside one.

What a cluster is made of

Oskavia's knit district began with one dyehouse. It was built against a single long programme, for a buyer who wanted colour closer to its own market. Everything after that is consequence.

1. An anchor arrives. A dyehouse, a laundry, a tannery, a mill — or simply one buyer's commitment, large enough to justify a building.
2. Machines follow the anchor. They are bought for a category, and once bought they decide which categories are cheap for a decade.
3. Operators accumulate. The town's pool fills with people who have run those machines. So the next factory's training time is short and its learner share is low, which is lesson 3's floor, forming here.
4. Suppliers arrive, because demand is now dense enough. Trims, threads, elastic, packaging, a machine-repair trade, a spare-parts stock, somebody who can re-time a bartacker — the machine that locks a stress point — on a Sunday.
5. The next factory enters that category rather than another. In that category it starts with a shorter training curve, a shorter component lead time and a mechanic ten minutes away.

Step five feeds step three, and the loop closes. That is the whole mechanism. It explains both of the properties that make origin maps frustrating: a specialisation takes years to form, and it does not move when a wage changes.

Norforth's own supplier database is the cheapest measure of it anybody has. Within forty kilometres of that dyehouse it lists 31 knitting factories, 6 trim suppliers, 4 laundries and 2 machine-repair firms. Within forty kilometres of Marendal's capital it lists 2 knitting factories and no laundry. Neither figure is a national statistic. Both were counted from a database Norforth already had.

What it costs to quote outside your cluster

Almiraz makes knits well. In May it quoted Norforth for NRF-1120, a shirt in cotton poplin — a plain, closely woven cloth — 12,000 units. Its sewing was genuinely cheaper than Ferrand's, by 0.38 a shirt. It lost, and the loss had nothing to do with sewing.

A poplin shirt needs sixteen bought components beyond the main cloth: interlining, buttons, thread, collar stays, three labels, a bag, a carton, a hangtag, a hanger, tissue, a size sticker, a back board and a butterfly. Interlining is the stiffening layer bonded inside a collar and cuff. Six of the sixteen have a supplier inside Oskavia's knit district. Ten do not.

The interlining line is worth reading twice. Almiraz needs 4,300 m, and the smallest lot anybody will sell it is 12,000 m at 1.42 a metre. The 7,700 m left over is worth 10,934.00, and Almiraz cannot use it, because it makes no other wovens. If the style repeats, that becomes stock. If it does not, it is 0.9112 on every shirt in the order.

Fusing is the heat-press step that bonds the interlining to the cloth. Almiraz has no press, so it hires one ninety minutes away.

And the component critical path is longer than the cloth's. The slowest local component is 11 days. The slowest imported one is 47. In its own category, Almiraz's components never appear on the critical path at all. One category across, they are the critical path.

None of the 1.6012 is available to Almiraz's management. It could match Ferrand's method exactly and still lose. To buy those ten components locally it would need suppliers who do not exist. And they do not exist because there is not enough woven-shirt demand in the district to feed them. That is the loop from the last section, running against a factory instead of for one.

Specialisations decay, and usually for one reason

The comforting story is that a cluster is lost to a cheaper country. More often it is lost to a machine.

Belrone had real depth in woven shirts for two decades. Then the collar and cuff work in that category moved to automated units at 240,000.00 apiece. Run one at 1,800 shirts a day over five years and three hundred working days, and it costs 0.0889 a shirt. Run it at 400 a day and it costs 0.4000.

Nothing about the small factory got worse. The minimum viable scale moved, and the factories below it were priced out of a category they had been good at for twenty years. Some of the trim suppliers followed the volume out. The mechanics who knew that machinery retired. The operators who could set a collar dispersed into other work. Depth measured in decades thinned in about four years.

Here is the general shape. A cluster is at risk when its category takes a capital step its members cannot fund, when its anchor leaves, or when the skill it rests on is held by people who are not being replaced. All three are visible years ahead, and none of them is a wage.

What starting one costs

This is the honest answer to "what would have to change". It is not encouraging, which is why it is worth stating precisely.

Marendal would like knit capability. The missing piece is not factories — it has factories. It is the dyehouse and finishing line that would take 1.2013 a kilogram of freight, input duty and importer's margin out of every kilogram of cloth its factories buy. On a sweatshirt that is 0.7448, which is larger than most of the differences the whole of lesson 1 was arguing about.

A dyehouse and finishing line on that scale is capital and roughly thirty months to commission, and it needs throughput to survive. On Norforth's own estimate the plant would need about 4,000 tonnes of cloth a year to run economically. Everything knitted in Marendal today is about 1,100 tonnes.

So the mill cannot be built until the factories exist, and the factories cannot win the work until the mill exists. That circle is why origin maps stay recognisable for a decade at a time. Only three things break it, and it is worth knowing which of them you are looking at.

• An anchor buyer who commits volume for long enough to underwrite the plant. That is the mechanism that built Oskavia's district.
• Public capital or a policy that funds the plant ahead of the demand and accepts the loss in the meantime.
• A neighbour, when a mill across a border is close enough to serve the cluster as if it were inside it. That is why a cluster is not a country, and why the two towns either side of a frontier are often one industry.