Methodology

Three lanes, three warrants

Every card on this board carries a tag saying which lane it came from, because the lanes carry different warrants and a reader who cannot tell them apart will read our measurement as someone else's filing, or the reverse.

State filings

Closures come from WARN notices: the advance notice an employer must file with a state labor agency before a mass layoff or plant closing. We reproduce the fields as the state published them, including the employer name, and link the state record on every card.

We currently hold California, North Carolina, Ohio and South Carolina. Other states file WARN notices we do not yet hold, so a count on this board is a count of what we hold and never a national total. Each state page carries the date of its most recent notice, so a jurisdiction that has gone quiet is visible as a stale date rather than as an absence.

A WARN notice is a notice. Filings are amended and withdrawn, employee counts are estimates at the time of filing, and an effective date can move. Follow the source link for the record of account.

What we will not say about a filing

We do not tell you that the employer on a WARN notice is a particular company in our index, and we do not link a card to a company profile. Matching a filed employer name to an indexed supplier is done by name, and a name is not an identity: two firms share a distinctive word more often than intuition suggests, and our stored match does not record which rule produced it. Rather than publish a link we cannot stand behind, we publish the name the state filed and let you follow the state record.

Measured concentration

Chokepoints are measured over the manufacturers BOMForge has indexed. For a capability, material or certification, we take the share of indexed suppliers in each state and compute the Herfindahl-Hirschman Index, the sum of squared shares. It runs from near zero, when supply is spread evenly, to 10,000, when a single state holds all of it. Above 2,500 we call it highly concentrated; between 1,500 and 2,500, moderately so. These are the thresholds US antitrust agencies use for market concentration, applied here to geography.

A reading is only published when enough suppliers are indexed to support it. Below that floor the arithmetic describes a handful of firms rather than a market, and publishing it would report a chokepoint wherever our coverage happened to be thin. Keys under the floor are left off the index entirely, and asking for one directly returns a page saying the reading cannot be made.

We publish shares, the index and the band. We do not publish the number of suppliers behind them.

The check that keeps our own errors off this board

A high share in one state can mean the country's supply really is there, or it can mean one source listed a batch of that state's suppliers with the same attribute. The two look identical in the arithmetic, so before publishing we compare the leading state's share of a key against that state's share of everything we index. When the first is several times the second, the reading describes our pipeline rather than American industry, and it is kept off the chokepoint pages.

This is not hypothetical. In September 2026 the six most concentrated materials in the index all read "North Carolina holds 77%", because a single enrichment pass had stamped the same thirty alloys onto roughly seven hundred North Carolina records. When the comparison cannot be made at all, the reading is held back rather than published, because a check that did not run is not a check that passed.

The limit that matters most

A share here is a share of what BOMForge has indexed, not of what exists. Our coverage is uneven across industries and states. A concentration reading tells you how the indexed supply for something is distributed; it is evidence about a market, not a census of one.

The wire

The wire carries headlines from trade and general press covering manufacturing, materials, freight, trade action and defence industry. We show the headline, the publisher and the date, and link out. We do not reproduce article text and we do not summarise a story into a claim of our own.

Refresh

Concentration readings are recomputed nightly. The wire updates through the day. WARN notices are refreshed as the states publish them, which is why a state's most recent notice date is printed next to its count.