There is a failure mode in every data-backed property tool that almost nobody writes about, because it looks like a result rather than a problem.
You draw a polygon, set your criteria, and the search returns nothing, or nearly nothing. The screen says, in effect, no properties match. You conclude the market is thin and you move on.
Sometimes that is true. Often it is a statement about the data behind the search, not about the buildings inside the polygon.
Two searches that look identical from the outside
Both of these are real runs from our own production, both finished cleanly, both had zero failed stages.
One was a warehouse sweep over a drawn polygon in St. Petersburg, Florida. Nine agent runs, eight of them reasoning stages: the search brief, the listings sweep, address resolution, a distress pass, a ranker, a best-use pass, and the report. It came back with a longlist, because underneath it there is a parcel roll to enumerate.
The other was a multifamily sweep over a polygon in a city where we have no parcel roll at all. Same product, same nine stages, same clean finish. What it could not do was enumerate a register that does not exist on our side.
From the user's chair, those two runs look like the same search returning different markets. They are not. One is a market answer. The other is a coverage answer wearing a market answer's clothes.
Why tools do not tell you which one you got
Because they usually cannot. A search pipeline knows that a source returned no rows. It does not reliably know whether that means there is nothing there or nobody ever wired this place up. So it reports the only thing it is sure of, zero results, and the interpretation is quietly left to you.
We think that is the wrong default, and we would rather say so openly than discover it with you on a Tuesday afternoon.
How to tell them apart, in any tool
- Ask what the enumeration source is: a parcel roll, an assessor file, a listings feed, a scraped index. If nobody can name it, the empty answer means nothing.
- Ask whether it covers your jurisdiction specifically. Coverage is granular and almost never national. A tool can be statewide in one place, county by county in the next, and absent in the third.
- Test with a known-dense area. Run the same query over a place you are certain has fifty matching properties. If it also comes back thin, you have your answer.
- Look for a stated limit. A vendor who publishes where they are weak is telling you something a vendor who publishes only strengths is not.
Where we actually are
For the United States, our territory sweeps run on parcel and registry data with real depth, and the coverage is uneven: statewide in some states, county by county in others, and absent in a handful. The live count of sources and states is published on our own sources page, which is the figure that moves; we would rather send you there than print a number here that goes stale. We publish the gaps because the places we do not cover matter more to you than the places we do.
Outside the United States, three countries have a live registry connector behind them: Spain, Germany and France. Everywhere else the honest sentence is the same one, and it is not a refusal:
We work there from open sources and the documents you give us. We do not have a government register behind us in that country, so the analysis leans on your material and on open sources rather than on official records.
That is often enough for the job actually in front of you. It is simply not the same thing as a register-backed sweep, and the two should never be quoted at the same price of confidence.
The rule we hold ourselves to
An empty answer is a claim, and a claim needs a source. If a search of ours returns nothing, the useful question is not "is this market dead". It is "which register did you ask, and did it answer".
Both runs are real, finished on 16 September 2026, and measured on our own production database.