Built on Tenth is independent. No agency referral fees, no paid placements, no client retainers in markets we publish on. This page explains exactly what we collect, how we verify it, and the line we draw between cited primary sources and modeled estimates. If a number isn't cited, it's labeled.
Refresh cadence is honest about staleness — public APIs and scrapes are live; government data sets refresh on their own schedules and we surface that.
| Source | What we use it for | Refresh |
|---|---|---|
| Google Maps + GBP | Map pack rank, review velocity, photo cadence, services completeness, Q&A activity | Per audit |
| Local Falcon | 5×5 grid scans for share-of-local-voice across a 5-mile radius | Per audit |
| County building permits | HVAC install & replacement volume by ZIP, normalized per 1,000 households | Quarterly |
| State contractor licensing boards | License status, complaints, disciplinary actions, named-operator structure | Quarterly |
| U.S. Census ACS | Household income, housing age, MSA population sizing | Annual (5-yr) |
| U.S. Bureau of Labor Statistics | Occupational counts, wage data, regional labor density for HVAC workforce | Annual |
| NOAA climate normals | Cooling-degree-days, heating-degree-days, climate-zone classification | Annual |
| Vendor & manufacturer pages | Dealer locator records, certified contractor lists, financing partner footprint | Per audit |
| Mystery-shop calls | One client phone shop per buyer; data captured by typing in real-time | Per audit |
| Buyer-supplied agency reports | Translation of agency monthly reports the buyer pastes in — vanity vs. revenue | Per audit |
Cited numbers come from a named primary source with a citation in the article or report. Modeled numbers are estimates we derive — they're useful, and they're labeled as such inline.
Every cited fact has a source row, a retrieval date, and a public URL. If you can't find the citation, it isn't cited — and the assertion needs to be labeled modeled or pulled.
Where we estimate from headcount, ticket-size benchmarks, or industry medians, the inline label reads "modeled estimate" and the assumption set is footnoted.
One call. One operator. The buyer's own line. The benchmark numbers come from the top-7-on-file already in our city archive — we don't shop competitors on the buyer's behalf.
A single inbound call to the buyer's main line, posing as a homeowner with a service or replacement inquiry. Not three. Not "your competitors." One.
Comparison numbers (answer rate, hold time, booking ask offered, fee quoted) come from prior research on the top 7 operators we already have on file for that metro. The benchmark is a median, not a real-time scan.
Data is captured by typing in real time as the call happens. We do not record audio. One-party-consent jurisdictions only — Florida and Nevada shops are not conducted under this protocol.
Script under two minutes. Single attempt — no retries. Caller does not pretend to book a real job that the operator dispatches; the call ends at the booking ask.
The operator (company) is named in the buyer's private report. The individual customer service rep who answered is anonymized as "CSR-01" — performance attaches to the operation, not the person.
Private-equity-backed and franchise operators are named in published research because identification of market structure is in the public interest — buyers and sellers in the HVAC space deserve to know who controls how much of which metro.
Personal names of CSRs, technicians, and unnamed dispatch staff are anonymized in every artifact. If a CSR's performance is part of a finding, it's described by role, time of call, and shop date — never by first name, last name, or voice characteristics.
Email hello@builtontenth.com with the article URL and your evidence — a citation, a screenshot, a public record. We respond within 5 business days. Confirmed errors are corrected with a dated note at the top of the article. Personal opinions, predictions, and interpretation disagreements aren't corrections; they're letters to the editor and get a separate response.
Marketing agencies cannot pay us to recommend them. Operators in published reports cannot pay to be excluded.
Vendor pages, software, and service providers cited in articles or reports paid us nothing. Citations are editorial.
We do not consult, retain, or accept ongoing fees from operators in markets we publish on. The diagnostic is a one-time deliverable.