Data Use and Sources Policy

Draft prepared with AI assistance; not legal advice. To be reviewed by a lawyer.

Last updated: draft of 3 November 2026.

In short

Every record comes from an official public record, collected from the government office's own open-data portal or from a feed we pay for under the office's terms.
We cover companies, not individuals, and we leave out California and Vermont records.
Use the data to find and research business prospects. Never use it to decide anyone's credit, insurance, employment or housing.
Every record links to its official source. If you spot an error, tell us and we will fix it or credit it.

This policy is part of our Terms of Service.

Where the records come from

Equipment-finance signals come from UCC financing statements and amendments filed with state filing offices, collected from each state's official open-data portal.
Florida construction signals come from building permits published by Florida county and city building departments on their official open-data portals. Notices of commencement will come from county clerks' official-records feeds, bought under each clerk's commercial terms, once we add them.
Reference files (not in use yet). We hold the Florida Department of Business and Professional Regulation's construction license file, and plan to use it and state company registers to check companies and licenses. None is applied to delivered records today, and no license number is delivered. Our Sources page will say when one is. Reference files are never sold on their own.

Our Sources page lists every source in use today, with a link to it, how often it updates, the terms it publishes and the credit it asks for, and it shows which reference files we hold and whether they are in use.

How we collect

We add a source only after reading its terms of use and recording them. A source whose terms are unclear waits until they are clarified.
We collect only through official downloads and APIs, identify our software honestly, keep to a slow, polite pace, and stop if a source asks us to.
We never get around access controls: no CAPTCHA solving, no automated use of search sites that forbid it, no scraping of LinkedIn or Google Maps, and no data bought from list brokers.
We keep a copy of what each source published, so we can always show where a record came from.

How records are prepared

Collect. Every morning, scheduled jobs download the latest official files.
Clean. Company names are put in a standard form (so "ACME ROOFING, L.L.C." and "Acme Roofing LLC" are treated as the same company), and duplicates are merged.
Keep companies only. Records whose subject is a person are removed (see "What we leave out").
Read and categorize. Each record is placed in a trade (permits) or an equipment category (UCC filings) from the text the source published: the permit type and work description, or the collateral description. This is done only with fixed written rules and keyword lists, which also write the one-line summary; no AI model is used. Rules can be wrong, so every record links to the official record, and the CSV keeps the published description where we may show it (when a permit's owner is a person, a description that could name them is replaced by the permit type). If we ever use an AI model for this, this policy will say so first.
Categories based on the lender. Some filings carry no collateral description (Connecticut publishes none). When the lender is a specialist, such as an equipment maker's own finance company, we infer the category from the lender's name and label it "(based on the lender)" in the email, and the CSV's category_based_on column says "lender name". We never infer a category from the name of a bank or general lender, or from a filing agent's name.
Filing agents. When the secured party on a filing is a filing service acting for the lender, we say the filing was made through an agent and name the agent, rather than presenting it as the lender.
Flag. We add signal flags, such as a filing approaching its five-year lapse date (from the lapse date the filing office publishes), and the date we first saw each record.
Deliver. Each subscriber receives only the records that match their filters.

What we leave out

Individuals. Records about people are dropped before storage: a contractor or debtor whose name has no company marker (such as LLC, Inc. or Corp.) is treated as a person and left out.
Property owners who are people. We never show their name or their mailing or street address. A job site keeps only its city, ZIP code and the official record number, so a business with a real need can look up the official record itself. Permits where the owner does the work (owner-builder permits) are left out.
Personal details. We never add personal contact details such as personal mobile numbers or home addresses, and we never collect identification numbers such as Social Security numbers or dates of birth, even where a source file contains them.
California and Vermont. We do not include records from California or Vermont.
Removed records. Records removed at a company's or person's request, or at a source's request, stay out (see our Privacy Notice).

Filings that are not equipment financing (blanket liens on all of a company's assets, accounts, inventory and receivables lines, merchant cash advances and landlords' liens) are not equipment filings: they are left out of equipment plans, counts, data pages and samples.

Freshness and lags

We collect every morning and send digests each morning.
A record reaches you only after the government office publishes it. That can be days or weeks after the date it was filed or issued, and it differs by source. The Sources page shows each source's schedule and usual lag, so the latest days can look thin until a source catches up.
Some sources update every day, some only on weekdays and some weekly. Holidays and office closures mean days with no new records.
Each record shows the date it was filed or issued and the date we first saw it. A record is delivered once, on the day it first appears; a genuinely new event about it, such as an amendment or a lapse alert, is delivered as a new, labelled record.
If a source is late, broken or stops, we show the gap rather than hide it.
Sources can correct or withdraw records after we deliver them. The official record is always the latest word.

How you may use the data

The Terms of Service set the rules. In plain words:

Yes: finding and researching business prospects, planning sales territories, market and competitor intelligence, and loading records into your own CRM for those purposes.
No: deciding, or helping decide, whether any person or business gets credit, insurance, a job or housing; tenant screening; underwriting; or any other purpose covered by the Fair Credit Reporting Act. We are not a consumer reporting agency.
No: reselling, sharing or publishing the data outside your business.
No: calls, texts or emails that break the TCPA, Do Not Call rules, CAN-SPAM or state law, and no loading our phone numbers into autodialers or mass-texting tools.
No: harassment, profiling private individuals, or anything that misleads people about who you are or where you got their details.
Always: check the official record before you act on it, and honor opt-outs and our removal notices.

How to report an error

Email hello@fileherald.com, or reply to your digest, with the record ID (the record_id column in the CSV) or the official record number, and tell us what looks wrong.
We check it against the official record, usually within two business days.
If the mistake is ours (a wrong category, a duplicate, a broken link), we fix it and credit you under our Credits and Refunds Policy.
If the mistake is in the official record itself, we tell you which office can correct it, and we can hold the record back while it is corrected.
Anyone, customer or not, can ask us to correct or remove a record about their company or a record that names them: see our Privacy Notice.

Contact

Fileherald
500 Brickell Ave, Unit 2907, Miami, FL 33131
Email: hello@fileherald.com