The Hidden National Supply Chain

It exists. You just can't query it.

Ask something simple — how many factories are in the United States — and you hit a wall immediately. We don't know. Some counts treat a factory as any manufacturing establishment, including very small facilities with a handful of employees. Others emphasize large plants and quietly drop the long tail. Estimates land between 250,000 and almost 700,000. A single company can report one site as one establishment or split a campus into many establishments by function and reporting convention, and both answers are defensible depending on who's asking.

Now ask the operational version of the question — where is the national supply chain — and it gets worse. The Census Bureau's Business Register is described as the backbone record of U.S. establishments, but establishment-level records aren't available for public use, because federal law protects individual business data. County Business Patterns will tell you how many establishments exist by NAICS code and geography, and it's published as summary statistics, so there's no list there to route an RFQ to. The closest thing most manufacturers and buyers bump into in day-to-day work is the federal contracting identity stack: SAM.gov registration, a UEI, and the small-business search tools populated from it. Those systems exist so the government can buy goods and services. They only cover entities that chose to register to pursue federal awards, and parts of a record can be restricted from public search, so none of it can serve as a complete, routable map of the national supply chain. And don't get me started on NAICS as the way to categorize manufacturers.

So we know the United States has a real manufacturing base, and we have no canonical way for a buying organization to discover it and qualify it with evidence. Buyers know pieces of it — the piece their tools show them, the piece they've personally lived through, the piece sitting in the business cards they've collected. At the end of the day it's the parable of the blind people trying to describe the elephant. A picture is truly worth a thousand words here:

And the elephant is very tangible. It's a thing you can see and touch and smell: VMCs running 6061, lathe cells cutting 17-4, weld fixtures clamped down on tables. It's in industrial parks and barns and garages. It is also boring paperwork. It's the vibrant community on Manufacturing X and LinkedIn and at conferences and trade shows. The evidence all exists. Much like the elephant, putting it together from that many disparate touchpoints is the hard part.

Here's my claim. Most supplier discovery and qualification and re-qualification work in U.S. manufacturing — call it supply chain building — is caused by facility identity and execution evidence that can't travel, not by any lack of supplier capacity. And when I say a supply chain should be "queryable," I mean something specific: a buyer can search, filter, and route work based on verified identity, capability evidence, and performance history. Not marketing fields.

Why directories don't solve it

A directory is a list of names with marketing fields attached. A routable supply chain needs proof attached to work packages. A shop can list "5-axis machining" and "ISO 9001," and the buyer still cannot tell whether that shop will hold a profile callout on the functional surface, run the right datum scheme in the probe routine, keep revision control straight on the traveler, and deliver a cert packet that survives an audit.

Even government-sponsored directories tend to be scoped to specific sectors, voluntary, and explicitly disclaimed as non-endorsements. I don't say that as a criticism. It's a clue. A directory can help you find candidates. It cannot, by itself, carry the liability of being wrong.

Why Google doesn't solve it

Google finds what has been optimized to be found. The best shops in a region are often the least noisy online, because they're busy running travelers, and plenty of them actively do not want inbound RFQ blasts from unfamiliar buyers. Their capability page is two machine moves and one quality manager resignation out of date. Route from search results and you're routing from copy, not from shipment evidence. The buyer learns the predictable lesson — discovery is cheap, qualification is expensive — and then that lesson hardens into policy.

Search engines and directories also feed the noise floor from Paper No. 6. A buyer whose Rolodex is too thin, or who isn't getting responses, goes out and runs a generic search for suppliers, finds them in directories and search engines, and starts blasting demand at a pile of imperfect matches he knows almost nothing about. Which is exactly why we've seen some shops shy away from being discovered through those tools at all.

Why the Rolodex doesn't scale

A Rolodex — a collection of business cards, a contact list, whatever form it takes — is a personal trust graph. It works until it doesn't. It works when the part family is familiar, the material is common, and the buyer can call the owner and get a straight answer about lead time and inspection. It breaks when a program changes, a cert stack changes, or a special process shows up that needs a sub-tier with real traceability.

Maybe the worst failure of this approach is the unknown-unknowns test. A buyer can be perfectly happy with the shop he's used for eleven years and be completely blind to a better one four miles away. Missing chances to scale, to save, to keep improving, and there's no mechanism anywhere that would tell him. How would he find out? He wouldn't. It's a good way to get stuck and never learn how stuck you are.

The Rolodex also dies with turnover. The company keeps buying parts. The individual relationships don't persist in any form the next buyer can route with confidence. So vendor development gets redone inside every firm, over and over, as if each one is the first to discover that the traveler and the cert packet are the real product.

Why trade shows feel necessary

Trade shows are the physical substitute for a missing identity layer. If you can't verify a facility through portable evidence, you go look someone in the eye, ask what machines they run, ask what they do for inspection, and try to infer whether their cert discipline is real. That's expensive, slow, and biased toward whoever can afford booths and travel.

They also select for presentation. The buyer is trying to solve a routing problem for a drawing and a ship date. The trade show solves a marketing problem. Sometimes those overlap.

Trade shows and the business card piles they generate are both roads to the same personal-network answer, and they're worth dwelling on because they're the most common way buyers and suppliers get connected and stay connected. They feel safe. They're classic. It genuinely is great to meet people in person and know who you're working with. They also don't scale. Let's be honest with ourselves about the timeline: in one long lifetime we've gone from inter-office memos and mailing letters and paying a lot of money for long-distance phone calls to short, instant messages as a normal way to do business. We live in an era of speed and connectivity, and even in classic areas of machining the processes are changing rapidly. On top of that, we're in an unprecedented era of change-over in supplier ownership as the boomers retire and shops change hands or close. Rely solely on personal networks, expanded only through in-person events, to develop a supply chain and at best you're missing out on opportunities. At worst you will fail as your small network fails, and you won't realize it until it's too late.

Why verticalization is the default fantasy

When discovery and qualification hurt enough, leadership starts eyeing a mill. If we buy the VMC, we control the schedule. For some companies with deep metrology and process engineering, that's true. For most it just moves the burden. The work package doesn't disappear when you verticalize. You still need an inspection plan, gauge calibration, revision control, a probe routine, outside anodize, outside heat treat, and cert packet discipline that ties all of it back to a PO. A mill without that paperwork and measurement capability isn't supply chain resilience. It's a new way to generate scrap fast.

No man is an island, although we all really, really wish we could be. It's a common fantasy. Some folks do get close to verticalization. But nobody is pulling material from the ground and delivering a finished product to a customer without a high-quality supply chain underneath. Nobody.

What buyers are actually rebuilding, every time

Strip it down and every buying organization is assembling three categories of data about every supplier.

Facility identity: which legal entity is this exactly, where is the floor, which certifications apply to that address, and who is accountable when a cert packet turns out to be wrong.

Execution capability: not "CNC machining," but whether this facility can run this geometry, in this material, to this tolerance block, verified by this inspection method, inside a stable process window.

Performance history: on-time delivery against ship dates, first-pass yield against inspection reports, responsiveness against question lists, and revision control discipline measured by how often the traveler and the inspection package actually match the PO.

None of that is an opinion about a supplier. It's artifacts and rates. Hopefully companies track them internally; many don't apply much rigor to this kind of tracking. And even the ones that want to almost never get to reuse the data outside their own walls, because there's no shared format, no shared verification, and no shared consequence for misrepresentation.

The point, twice

There are really two points here.

The systemic one. The United States does not have a "missing supplier" problem at the national level. There are definitely categories that need work, and there are a lot of categories that are incredibly strong and deep. But like the people with their hands on the elephant, that judgment is built from partial data. What we actually have is a missing map and a missing evidence layer. Some countries have a more routable picture of industrial capacity because their reporting norms and procurement systems are more centralized, and because export-driven industrial policy created stronger incentives to standardize identity and evidence. What we get instead is redundant supplier development performed by thousands of companies in parallel, using email threads, spreadsheets, trade shows, and expensive verticalization experiments. Buyers doing the same work over and over and over.

If coordination is computation, then supplier qualification is a data problem, and what cannot be represented cannot be routed. The fix isn't another directory. The fix is a portable way to anchor facility identity and attach execution evidence to work packages, so a buying organization can route a drawing to a traveler with a ship date and expect the cert packet to close cleanly.

The behavioral one, which is where these papers usually end up. In Paper 1 we talked about the desire to control through yet another standard. In Paper 3, local optimizations producing national failures. In Paper 5, a lack of understanding leading to ambiguity, which leads to slowness and expense. In Paper 6, urgency creating slowness by raising the noise floor. Here it's information hoarding: evidence stays trapped inside firms because there's no shared verification and no safe way to share it. I don't expect these behaviors to change, and I don't think we should expect them to. The right solution has to account for human behavior and complement it instead of fighting it.

Implications

Redundant supplier development keeps running in parallel inside thousands of firms. Discovery stays cheap and qualification stays expensive. And verticalization doesn't fix it — it just shifts the supply chain burden somewhere else, where it's still necessary, only in a different form.

Next up is Paper No. 8, The Trust Spiral. Another one that starts as a system-wide issue and turns out to be about behavior.