How many sites are enough to size a fragmented facility market?

Someone on the sales side asks for a TAM on ready-mix plants, or scrap yards, or aggregate pits, and the honest answer starts with a confession: there's no directory for this. SIC and NAICS codes lump half these operators under "other," trade association rosters only cover dues-paying members, and the state permit database stops at the state line. The real question underneath "how many sites are enough" is a coverage question, not a completeness question. It's enough sites, found the same way across the whole area, to describe the shape of the market with confidence, well short of counting every last gravel operation in the country.

That distinction matters because fragmented facility markets don't behave like markets with a clean list. A directory-backed market lets you sample 200 out of 5,000 known accounts and run a confidence interval. A facility market with no directory has an unknown denominator. You don't know if 400 found sites is 30% of the market or 90% of it, and that uncertainty doesn't go away by finding more sites one state at a time. It goes away by covering ground systematically, at a scale where the gaps left over are small and roughly even, instead of concentrated in whichever regions nobody happened to drive through.

Counting versus estimating

Two different jobs get bundled under "market sizing," and mixing them up is where most facility counts go wrong.

The first job is a census: find and locate every site matching a signature, within a defined area, at a point in time. This is what a geocoded facility count actually is: an inventory assembled site by site. Its quality depends on how the area was covered at the time of the sweep.

The second job is sizing: turning that count, or a count plus a sample, into a revenue or unit estimate. This is where throughput assumptions, average site capacity, and utilization rates come in, and where a planning team starts applying per-site economics to get from "we found 1,840 terminals" to a dollar figure.

The mistake is treating a partial count from job one as if it already answers job two. If you've found 1,840 terminals by pulling together state permits, a trade directory, and a couple of field visits, you don't know whether that's close to complete or missing a third of the market concentrated in states where nobody looked. A facility count built from pulling together whatever lists exist has coverage bias baked in, and no amount of careful math in step two fixes a gap in step one.

When partial coverage becomes the whole population

Here's the practical test: if the sites you haven't found are systematically different from the ones you have, in region, in size, in how recently they were built, your count isn't a sample of the market. It's a biased subset standing in for the market, and every per-site assumption you multiply through inherits that bias.

A permit-based list skews toward operators in states with strict permitting and active enforcement. A trade association roster skews toward members large enough to pay dues. An industry report skews toward whoever the report's authors could reach by phone. None of these are wrong to use, but none of them are a population count either, and a TAM built on top of one should say so rather than present a confidence interval the underlying data can't support.

The fix is covering the ground directly, at country scale, scanning for the physical signature itself instead of relying on whoever filed paperwork about it. That's a different kind of count: one where the gaps are a function of imagery coverage and detection thresholds, which are knowable and consistent, rather than a function of which states happen to have a permitting office that answers emails.

If a facility count is going to carry weight in a planning deck, it needs to be built that way from the start. Facility Discovery runs that kind of country-scale sweep for yards, plants, terminals and pits matching a target signature, turning a market that's never had an address list into one your team can size and work. Get in touch to see what a sweep of your target signature would turn up.