Most real estate platforms begin with properties that are already for sale. iVerify begins with the ground. It learns what the country's existing truck parking, IOS yards, distribution centers, fleet terminals and travel centers actually have in common - road, freight infrastructure, carriers, industry, traffic - and then reads every U.S. market for the same conditions, to identify where another one of that kind of facility is most likely to succeed. Then it narrows the search to the parcels most capable of supporting that use, even if they are not on the market. The best development sites are rarely listed for sale.
Three steps: find the right market below, identify the best locations in the market-to-location screen, and evaluate the parcel and pursue site control in the site report, which carries the tax maps and ownership.
Data vintage - loading.
Pick the facility you are screening for. Every U.S. market is ranked for that use by the model trained on that use. Tap any market for its detail and a way into the location screen.
The filters narrow which markets are shown. They do not change any market's score - nothing on this page does. The score is the model's.
| # | Market | Strong sites | Best site | Sampled | Power units | |
|---|---|---|---|---|---|---|
| loading… | ||||||
The market unit is the CBSA (metro/micro statistical area). Nothing on this page is a formula, and there is nothing on it to tune. Each use has its own model, trained on the real facilities of that kind that exist in the United States today, and a market's rank is that model reading the ground.
For each use, iVerify assembles every known U.S. facility of that kind and an equal-footing set of controls, measures the same conditions at all of them - road geometry and traffic, freight infrastructure, carrier fleets, industrial establishments, rail, water, population and the rest - and fits a gradient-boosted model to tell the two apart. The score is 100 × the model's probability that this is the kind of place that use is built. It is not a rating out of 100 and it is not an opinion; it is a likelihood, and it is only ever as good as the evidence behind it, which is why the evidence is printed on every chip.
A market's score is the 90th percentile of the model's scores across sampled locations inside it - not its best single spot. This is deliberate and it is a business decision, not a statistical one. A developer builds a program, not a building. A market with one strong location and little behind it offers a single site and no follow-on; a market with depth offers a pipeline. Ranking on the best single site would place the shallow market above the deep one, so the screen ranks on depth instead. Each market's detail shows its sample size, its median and its best; a market with too few sampled locations to speak for itself is shown as unranked rather than given a rank it has not earned.
| Use model | Discrimination (AUC) | Trained on | Evidence |
|---|---|---|---|
| loading… | |||
Underlying sources and vintages (read from the load record, not hand-typed):
| Layer | Source | Vintage | Rows |
|---|---|---|---|
| loading… | |||
How to read this screen. This is a market ranking - the first step in the chain, not the site report. The same model, the same scorer and the same facts carry through to the Market-to-Location screen and the site report, so a market's rank here and a site's score there are the same machine talking about the same ground. Parcel-level feasibility, land cost, and flood are resolved at the exact location downstream. Announced projects, where present, appear as a per-market badge with a source link and are never scored.