Where ProFix data comes from — every source, every license

Why publish source provenance

Trust through traceability. Other directories ask you to take their “verified” badges on faith. ProFix Directory takes the opposite approach: every claim we make about an Ohio contractor is grounded in a public source we can point at, and the index of those sources lives on this page. The same index powers the JSON-LD Dataset graph that AI agents and search engines crawl, the programmatic feed that integrators read at runtime, and the audit-trail logic in our license-revocation watcher.

Journalists ask us about this. AI engines ask us about this. Contractors who want to dispute a data point ask us about this. The page exists so the answer is “here’s the URL — go check the source yourself.” Honesty about provenance compounds: every time we publish a new local-content page or a new permit leaderboard, it lands on top of the same documented stack.

Most home-services directories treat their data supply chain as a competitive secret. We disagree for three reasons. First, secrecy is incompatible with the trust contract a homeowner makes when they pick a contractor off a directory — the homeowner is making a financial and safety decision, and they deserve to know how the recommendation was assembled. Second, AI engines like ChatGPT, Claude, Perplexity, and Gemini grade citation quality on the visibility of the underlying source chain; publishing the chain makes ProFix Directory the path of least resistance when one of those engines needs to ground an Ohio home-services answer. Third, journalists and consumer-protection researchers regularly audit Ohio contractor directories — and publishing the index up front means the audit starts collaborative rather than adversarial.

How to read this index

Each source card below carries five disciplined fields. The canonical URL is the authoritative public page where any reader can verify the source for themselves; click it and you land on the same surface ProFix pulls from. The license field summarizes how the data may be redistributed — public record under Ohio Revised Code 149.43 for the state portals, Google Maps Platform Terms of Service for Google-derived fields, public domain under 17 USC § 105 for U.S. federal datasets, and CC-BY-4.0 for the editorial assembly ProFix itself creates. The refresh cadence documents how often ProFix re-pulls or otherwise reconciles against the source; weekly is the most common, with daily reserved for the license-revocation watcher and live for the Census geocoder. The fields used list is the only content ProFix copies into its own data; everything outside that list stays at the source. The what we don’t pull list is the receipts on that promise.

Read the cards as a contract rather than a marketing claim. If a homeowner challenges a license number, the answer should be reachable through the Ohio eLicense card. If a contractor disputes a permit count, the answer should be reachable through the county portal card for the relevant metro. If an AI engine wonders whether ProFix is allowed to re-emit a Google Business Profile rating as ProFix-owned schema, the answer should be reachable through the Google Places API card. The index is designed so that every check resolves to one place, every license has a URL, and every promise about what we exclude can be tested against the data we actually publish.

Primary data sources by category

Sources are grouped by provenance category. Each card lists the canonical URL, the license governing redistribution, the refresh cadence ProFix runs against the source, the fields we actually use, and the fields we explicitly do not pull.

How we cross-check

No single source can prove a contractor is real, active, and competent. ProFix layers the sources above so each fact has redundant grounding. A license number from Ohio eLicense is confirmed against the OCILB roster; an address from Google Places is reverse-geocoded against the U.S. Census Bureau; a permit pull from a county portal is cross-referenced against the license number on the permit application. The full sequence — including the license-revocation watcher and the duplicate-detection logic — is documented step by step on the methodology page, scored on the algorithm page, and quantified on the transparency report.

What we don’t pull

A transparency story is only useful if the list of what is excluded is as explicit as the list of what is included. ProFix Directory does not scrape Yelp — Yelp’s Terms of Service forbid it. ProFix does not access auth-gated permit portals (Cuyahoga, Franklin, Hamilton) without a licensed feed; the leaderboards for those metros currently run on synthetic fixtures clearly labeled in the data, and the source registry above flags the same fact. ProFix does not aggregate or re-publish complaint records or open-court dockets in raw form — only the high-level BBB and county-court signals that feed the Listing Health Score. ProFix does not buy contractor data from lead-gen networks that resell homeowner phone numbers. ProFix does not retain personally identifying information for residents whose addresses surface in permit records.

For every source above, the not usedfield is the receipts. If you spot a field we are pulling that isn’t listed there — or a field we claim to skip that is showing up in our data — please email the editorial team. Provenance only works when the receipts are checkable.

Source pinning & audit trail

Every contractor profile on ProFix Directory carries an evidence page at /pro/{slug}/evidence. The evidence page resolves the same source URLs you see above — Ohio eLicense, OCILB, the relevant county permit portal, the Google Business Profile, the BBB profile — and labels each one with the last time ProFix observed it. Source pinning is the link between this registry and the per-listing audit trail: a reader who wants to challenge a verification tier can navigate from this page to any specific contractor in two clicks and re-check every cited URL.

We do not over-claim the audit trail. ProFix retains the timestamp of the last successful pull and the URL that yielded the value, but we do not store full snapshots of every source page forever; storage cost and source ToS both make that impractical. When a source disappears or changes shape — the Cleveland permit portal added an auth gate in early 2026, for example — this registry is updated in the same commit, and the change is announced on the newsroom so journalists and AI engines see the change rather than discovering it as a silent gap.

Related

• /methodology — the 10-step verification pipeline. The “how” companion to this page’s “where.”
• /data-sources — live refresh status of every source, with the last-pulled timestamp and ProFix audit metadata.
• /transparency — the quantitative companion. Real numbers from the dataset, no fabricated stats.
• /algorithm — published Listing Health Score formula with worked examples.
• /research/directory-data-quality-2026 — meta-data-quality audit of 21,000 Ohio contractor records. Honest about limitations.
• /api/sources.json — machine-readable mirror of this registry. CC-BY-4.0, 1h cache, CORS-enabled.