Measure/Matching an address to the right building

Method · identity

Matching an address to the right building

Every remote property measurement rests on one silent assumption: that the thing being measured is the thing the customer meant. The best terrain model in the world is worth nothing pointed at the neighbor's roof, and almost nobody says what they do when they cannot tell which roof it is.

We do. This page is how identity is established before anything is measured, what travels with every answer, and what happens when the evidence is not there.

Why this is a hard problem and not a lookup

A geocoder returns a point. A measurement needs a polygon. Those are not the same object and the gap between them is where the errors live.

The point may be a rooftop coordinate, a parcel centroid, a driveway entrance, or a position interpolated along a street segment from the house numbers at each end. On a dense terrace, four buildings sit within fifteen meters of it. On a rural lot, the candidates are a house, a barn, a garage and a neighbor's shed, and the nearest one is frequently the shed.

Locating is not identifying

"The nearest building to this point" is a location. "The building that carries this address" is an identification. Products that quietly present the first as the second are the reason wrong-house measurements are so common and so rarely noticed.

Three strengths of evidence

The evidence that settles which structure an address names comes in grades, and they are not equally strong.

  1. Direct identification. The strongest grade of evidence ties the address to this outline in particular, rather than to a spot somewhere near it.
  2. Corroborated identification. Independent evidence puts the address inside this outline and no other. Strong, though it can be fooled where the underlying records are imprecise.
  3. Neither fired. The nearest outline is all there is. This locates rather than identifies.

Behind those, a proprietary sequence of checks runs in order of strength, stopping the moment it has an identification. Which grade answered ships with every measurement. The sequence itself does not, and this page is not going to publish it.

What ships with every answer

The grade that answered travels with the result. Not a score we invented, not a blended confidence that mixes identity in with everything else — the plain statement of what settled which structure the address names, on that address, on that call.

So a caller never has to assume. A building we identified and a building we merely located are different claims, and they arrive labeled as different claims.

The limits, because they belong here rather than in a footnote

  • Where the underlying records are thin, the evidence is thin too. Rural addresses, new construction and recently renumbered streets are the hard cases, and they are hard in exactly the places where a wrong answer is least likely to be noticed.
  • Not every disagreement is a different building. Some are our outline and somebody else's tracing the same roof and disagreeing about where its center sits. The two failures look the same from outside and only one of them matters.
  • Identity is not a finding. Knowing which structure an address names says nothing about that structure.

What follows from that

Refusal

An address we cannot find is refused outright rather than resolved to something nearby. A refusal is never billed, and the quantity we declined is named in the payload rather than silently omitted — an absence and a refusal are different things and a caller is entitled to know which one they got.

Labeling rather than promoting

An answer reached by proximity alone is returned labeled as such, and it is never sold as a report. Nothing gets quietly upgraded because the alternative was returning nothing.

Lists get a stricter gate than lookups

On a single lookup a human is looking at the imagery and would catch a wrong roof. On a list of five hundred, nobody is. So a property only reaches an outreach list when identity was established and confidence is high and at least 3 m of wall was called — not merely leaning — as falling toward the structure.

Replacement

If the building is wrong, the property is replaced free. That is affordable precisely because the engine measures its own uncertainty and only releases what clears it.

Why the wrong building is worse than a wrong number

A slope that is off by two points is a slightly wrong answer to the right question. A slope measured on the house next door is a confident, well-formatted, entirely wrong answer that no downstream check will catch — because it is internally consistent. It will pass every sanity test you can write, right up until somebody drives there.

That asymmetry is why identity is gated before measurement rather than scored afterwards, and why the identification strength travels with every result. A building we identified and a building we merely located are not the same claim, and a payload that does not distinguish them is hiding the only thing that matters.

What this is not

Identifying a structure is not a finding about that structure. Nothing here says a property has a drainage problem, a wet basement, or any defect. Whether water has ever reached a particular building was not measured and is not claimed.

Getting it as a service

Address-to-building resolution is the first step of every measurement we sell, and it is available on its own for agents and applications that need to know which structure an address names before doing anything else.

Resolve an address and see which signal answered

The free look finds the building, shows it on the model, and says how strongly it was identified. For programmatic access — including the refusals, which are the interesting half — the measurement tools are available by API.

Reports are $25 per property; the platform is $99 a month for 25 lookups or $399 for 200. See pricing.

Common questions

Why is matching an address to a building hard?

Because a geocoder returns a point, not a structure. That point may sit on a driveway, at a parcel centroid, or interpolated along a street segment, and a rural address can have four candidate outlines within thirty meters.

How does Measure decide it has the right building?

Identity has to be established by evidence before anything is measured, and the evidence comes in grades of strength. The two stronger grades identify a building. The weakest — the nearest outline is all there is — only locates one, and it is labeled that way.

How do you know which evidence answered?

Because it ships with the result. Every measurement carries the grade of evidence that identified the building, so an identification and a location are never mistaken for each other. An answer reached by proximity alone is labeled as such and is never sold as a report.

What happens when no evidence fires?

The address is refused rather than matched to the nearest roof. A refusal is never billed and the quantity declined is named in the payload.

Do you publish a national accuracy rate?

No. One national figure would hide the thing that decides whether an answer is safe to use, which is what identified the building on your address. That travels with every result instead, and where nothing identified it the address is refused.