Skip to main content
Plumb Intelligence

Methodology

How our numbers are measured

Every statistic on this site is measured against the live database, published as a floor, and re-checked by monitors. This page is the contract behind those numbers — what they mean, how they’re computed, and what we deliberately won’t claim.

The floor rule

Public labels like “1M+ permits”, “65K licenses”, “2.2M+ parcels”, and “910K recorded sales” are floors: each is published only because the measured count exceeds it (last full re-measurement 2026-08-14). Between refreshes the database grows, so drift can only make these labels more true. A monthly claim-drift monitor re-runs every measurement and raises an alarm if any published claim ever overstates the live database — the failure mode the floor rule exists to make impossible.

Published claimFloorMeasured (2026-08-14)
permits total1M+1,075,664
permits 12mo113K119,217
licenses total65K65,263
licenses active59K59,633
parcels all2.2M+2,276,862
parcels maricopa1.75M1,758,610
parcels pima440K448,423
sales total910K910,158
permits maricopa960K965,510
permits pima38K38,412

Sources and freshness

Permits come from 25 Arizona jurisdictions’ own published records — city and county portals, official monthly reports — each ingested on its real publishing cadence and never scraped past a robots.txt or terms that ask otherwise. Freshness is measured, not asserted: a nightly job computes each jurisdiction’s newest permit and the public coverage table shows that date per source. The same page lists the jurisdictions we can’t cover yet, with the reason — a coverage claim is only honest when the gaps are named too.

Contractor linkage

Linking a permit to the statewide AZ ROC roster is a resolution problem, and we publish the measured result (as of 2026-09-04): at least 61% of permits that name a contractor resolve to a verified license, and at least 39% of all permits resolve overall. The gap between those numbers is structural — several jurisdictions publish permits with no contractor field at all, disclosed per source in the coverage table. Where a name is ambiguous, the resolver stops rather than guesses: an honest blank beats a wrong attribution, because a wrong attribution puts someone else’s name on someone’s work.

Valuations and estimates

Declared valuations come from the permit record. When a source publishes none, we may show a labeled estimate (~EST) — the median declared value of comparable permits in the same work class — and the label never comes off. Placeholder values published by portals themselves ($0, $1) are treated as no-data, not as valuations. Sources with no valuation field are marked “n/p” (not published) rather than filled.

Parcels, owners, and the county rule

Parcel and owner-of-record data come from county assessor rolls, joined to permits on the normalized APN — and always county-scoped, because Arizona counties issue colliding APN formats: a Tucson parcel number must never resolve to a Phoenix owner. Owner mailing data is additionally gated in-product. Every join in the codebase runs through one guarded implementation, so the unsafe version cannot be written.

What we deliberately won’t claim

  • No invented statistics — a number appears only if a query produced it, and stat modules hide rather than fall back to a constant when they can’t measure.
  • No “every permit in Arizona” — the coverage page names what we pull and what we can’t, with reasons.
  • No silent estimates — modeled values always carry the ~EST label.
  • No guessed identity links — ambiguous contractor names stay unlinked.

Frequently asked questions

Why do your public numbers look conservative?+

Every public label floors its measured value — "1M+ permits" is published only when the measured count exceeds the round number. Drift between refreshes can therefore only understate, never overstate. A monthly monitor re-measures every claim and alarms if a label ever overstates the live database.

What does your contractor-linkage rate mean?+

Two rates, both measured: at least 61% of permits that NAME a contractor are linked to a verified AZ ROC license, and at least 39% of all permits are linked overall. The gap is structural — many jurisdictions publish permits without naming a contractor at all, and we say so per source rather than inventing links.

What does the ~EST badge on a valuation mean?+

The source published no declared valuation, so we show the median declared value of comparable permits (same work class) as a labeled estimate. Estimates are never mixed silently with declared values, and sources that never publish valuations are marked in the coverage table.

What do the match-confidence badges mean?+

A ✓ VERIFIED badge means the permit record carried the contractor’s license number — an exact ROC join. A ~ LIKELY badge shows the confidence of a name-based match. Ambiguous names are left unlinked rather than guessed: an honest blank beats a wrong attribution.

How fresh is the data?+

Each source refreshes on its own real cadence — most city portals daily, some county reports weekly or monthly. Freshness is measured nightly against our own database and shown per jurisdiction on the coverage page; freshness copy hides rather than falls back to a constant when it can’t be measured.

Can I cite these numbers?+

Yes, with attribution to Plumb Intelligence and the measured-on date. Aggregate figures were last re-measured 2026-08-14; linkage rates 2026-09-04. For press or research use, the founder will pull a fresh measurement on request.