The register still knows where things were made

We counted Sheffield’s cutlery companies and found the city holds 16% of the national trade against a 0.89% share of British manufacturing overall. The obvious next question was whether that is a Sheffield story or a general one, so we ran the same measurement on other towns with a trade attached to their name.

It is a general one, and in places it is far stronger than Sheffield.

What each town still concentrates in

Read the multiple, not the count. It is the town’s share of that one trade divided by its share of all British manufacturing companies, so 43× means Kidderminster holds forty-three times as much of the carpet trade as its size would otherwise predict. A town with no specialism scores about 1× everywhere.

Nobody planned this. Each row is just what the companies in that town happened to tell Companies House they do, and the pattern that falls out is the nineteenth-century industrial map with most of the volume taken away.

Kidderminster is the strongest concentration we found anywhere, stronger than Sheffield and cutlery. Nine companies, which is not many, but there are only 166 carpet and rug manufacturers in the country and the town holds one in eighteen of them on a thirteen-hundredth of the manufacturing base. Four are in our catalogue: Brintons, Bond Worth, The Grosvenor Wilton Company and Balta Floorcovering UK.

Stoke-on-Trent appears twice because the register splits ceramics into industrial and household, and the town is heavily over-represented in both. The household half is the one most people mean: 31 companies, and we list Emma Bridgewater, Portmeirion, Burleigh Pottery, Adderley Ceramics, Finsbury China, Peregrine Creamware, Repeat Repeat and Topaz China. The industrial half is the supply chain that makes the other half possible, including Potclays, Potterycrafts and Sneyd Ceramics.

Northampton and footwear is the one a reader is most likely to already believe, and it holds up at 5.6×. Fifteen companies at fifteen separate addresses, of which we list six, including Crockett & Jones, Edward Green, Tricker’s and Crown Northampton.

Walsall and leather survives at 5.8× with 29 companies, the second largest count in the table. We list Fairfax Saddles, Frank Baines Saddlery, Thorowgood, Ideal Saddle Company, Patrick Saddlery and Areej London.

The biggest one is the one we can barely show you

Leicester is the outlier in every respect. At 233 companies it has more than seven times the count of any other row, and it holds 10% of a trade with 2,311 companies in it, which is a scale none of the craft towns come close to. The garment trade in Leicester is not a heritage survival. It is the largest single concentration of manufacturing companies in one trade in one town that this measurement found.

We list two of the 233.

That gap is ours, not the register’s, and it is worth stating plainly rather than leaving out of the table. We add a company to the catalogue when we can find a way for somebody to buy from it, and for 231 of these we could not find a shop, a product page or a consumer brand. A catalogue built the way ours is will systematically under-represent contract manufacturing, and Leicester is where that shows up most sharply. Being able to measure the hole is the only reason we can tell you it is there.

Where the method finds nothing

Two towns we expected to produce a result did not, and that matters more for trusting the ten above than any of the ten do.

Macclesfield was the silk town. Its highest multiple on a trade holding at least eight local companies is 2.0×, on a catch-all machinery code. Dundee was jute. Its best is 2.8×, also a catch-all. Neither has a surviving specialism this measurement can see, and the honest conclusion is that in these two places the thing the town was known for is gone rather than diminished.

If the method returned a confident specialism for every town we pointed it at, that would be evidence it was finding patterns in noise. It does not.

The guard that kept two false results out

The first version of this table ranked purely on the multiple, and it was wrong in a way worth showing, because it looked completely plausible.

It reported Dundee as specialising in potato processing at 7.0×. Dundee has one potato company. It reported Northampton’s leading trade as grain milling, on four companies, which outranked the fifteen footwear companies that are the actual story. A single incorporation in a small trade produces an enormous ratio and means nothing at all.

So the tool now requires a trade to hold at least eight companies locally before it will rank it, separately from requiring the trade to have at least eighty nationally. The two floors do different jobs: one stops a trade being judged on a tiny national population, the other stops a town being credited with a specialism it holds by coincidence. Everything excluded is still in the output and is reported as withheld, with the highest withheld multiple printed next to it, so the exclusion can be checked rather than taken on trust. Macclesfield’s highest withheld figure is 13.8× on two companies, which is exactly the kind of number that would have made a better headline and a worse article.

What this does not say

The same four limits as the Sheffield piece, and they apply to every figure above.

It counts companies, not output, employment or revenue: a one-person workshop and a factory of three hundred are both one row. The trade codes are self-filed and unaudited. A registered office is not a works, though we report the distinct-address count for every row and nothing here is one accountant’s client list. And the match is on the postal town, which takes in some addresses outside each boundary.

One more specific to this piece: a town’s absence from the table is not evidence of anything. We ran twelve towns chosen because they have a trade in their reputation, which is the opposite of a representative sample.

Method

Enumerated from Companies House advanced search across 146 manufacturing trade codes, filtered to active companies, counted on 29 September 2026. Trade code descriptions are quoted from the Companies House condensed SIC list. The script is sheffield-concentration.ts in our repository and takes --city, --min and --min-local, so any of this can be re-run or contradicted.