How we define our terms

A Territory is one segment in one metro — for example Commercial HVAC in Phoenix. It is a fixed cell of the index, not a radius or a drawn area.

A Territory is one segment in one metro: Commercial HVAC in Phoenix, Solar in Orlando, Multifamily MEP in New York. It is a cell of a grid whose axes are the segment taxonomy and the metro registry, so a Territory either exists in the index or does not, and its coverage is a fact we can state rather than a result we compute per query.

This is why coverage is quoted per Territory. A metro is not uniformly covered across every kind of work — a jurisdiction may publish rich commercial filings and almost no residential ones — so a claim about "Phoenix" without a segment is an average over cells that differ, and averages of unlike things are the most confident kind of wrong number.

A Territory is not a Branch Radius. It does not move when you move, it is not centered on an address, and two people in different offices asking about the same Territory get the same answer.

A Branch Radius is a point and a distance: an address you enter, plus a mileage, resolved into the contractors whose permit activity is centered inside that circle.

A Branch Radius starts from an address you type — usually a branch, yard, or office — and a distance. It returns the contractors whose center of permit activity in a chosen segment falls inside that circle, ranked by how much they pull there.

The centering matters and is stated on the page: a contractor's activity center is a single point derived from where their permits are, so a contractor centered just outside your radius may still work inside it, and a contractor centered inside it may take jobs well beyond. It is an approximate match by construction, not a per-permit geographic filter.

A Branch Radius is computed per query and is not a saved object. It is the ad-hoc counterpart to a Territory, and the site keeps the two words apart because promising a Territory and delivering a radius is a promise of permanence that a radius cannot keep.

A reliable value is a permit valuation that passed every DQ check we run against it. A valuation that failed any of them is never shown as a number.

Permit valuations come from government portals and arrive with well-known pathologies: a jurisdiction that writes 1 for every filing, a form whose default is 1,000,000, a field that means "fee basis" in one county and "construction cost" in the next. A number that is present is not therefore a number that means anything.

A value is reliable when it is present, is not a known placeholder, and is not an out-of-band outlier for its jurisdiction and work class. Only reliable values are aggregated into a median, a total, or a rank.

When a value is not reliable, the page shows "—" and the reason beneath it. That is not missing data — it is data we hold and are declining to present as if it meant something. Aggregates state how many records they are computed over, so a median over eleven reliable values is never displayed as though it were computed over the whole population.

Field-DQ-cleared means a specific field on a specific record passed that field's data-quality checks — a per-field verdict, never a verdict about the record.

Data quality here is assessed field by field, not record by record. One permit can carry a trustworthy issue date, a trustworthy address, and a valuation that is plainly a placeholder. Marking the whole record "bad" would discard three usable facts to suppress one; marking it "good" would publish the placeholder.

So each field carries its own verdict, and each surface consults the verdict for the field it is about to render. This is why a permit can appear in a count, contribute its address to a map, and still show "—" where its valuation would be.

It follows that "this record is field-DQ-cleared" is not a sentence we can say. The clearance is always of a named field.

A measured value was recorded by the source about this specific filing. A placeholder is a value the source emitted without measuring anything — a form default, a sentinel, or a constant repeated across every record.

The distinction is detectable in the population rather than in the record. A single permit valued at exactly 1,000,000 is unremarkable; ten thousand permits in one jurisdiction valued at exactly 1,000,000 is a form default, and the eleventh is not evidence of a million-dollar job.

Placeholders are identified per jurisdiction and per field, because the same literal can be a real value in one place and a sentinel in another. A 1 is a plausible fee and an implausible construction cost.

Once identified, a placeholder is never shown as a number and never enters an aggregate. It is retained — we know it was emitted, and knowing which jurisdictions emit placeholders is itself a coverage fact — but it is retained as a placeholder, not as a measurement.

An honest null is a "—" shown together with the reason it is a "—". A blank that does not say why it is blank is not an honest null.

There are several distinct reasons a figure may be absent: the source never published it, the source published something that failed DQ, the field does not apply to this kind of work, or coverage for that Territory does not extend to that field. Each of those should lead a reader to a different conclusion, and a bare blank collapses them into one.

So the absence is rendered with its cause attached and visible — not behind a hover, which does not exist on a phone, and not in a footnote at the bottom of the page.

An honest null is also never a zero. Zero is a measurement meaning none; "—" means we are not asserting a number. Substituting one for the other turns an absence of evidence into evidence of absence, in a direction that always flatters the dataset.