Mason

Field guide · 15 free records · UK, US and global

Which free public records date
a buying event to the day?

15 registers and filings from our own research, what each one dates, and what it actually gave us.

Read the findings Get the atlas
Research run Aug to Nov 2026By Shreyan Rajesh

Mason / Research · Published 9 November 2026

116,732payment reports read from one free register, to find 6 companies worth showing
5,118companies from 7,166 breach filings, read free in under 6 minutes
156 of 340funded companies appeared only in a securities filing, in no funding news

Free public records (registers, filings and notices) date a change at a company to the day. We used 15 of them in real research runs. They are the hardest evidence we have, and the yields are small: one register of 116,732 reports gave 6 companies worth a call. The record dates the change. Your reading decides what it means.

THE QUESTION

Which records say when,
and what do they really give?

Job ads and news are easy to search and easy to misdate. A register entry is dated by the government that keeps it. The question is how much of it survives a careful read.

HOW WE CHECKED

Only records we used in a real run.
Each with that run's own funnel.

For every record we give what one run read, what it kept, and what it showed. The funnels come from the runs' own logs; the bars were set per run, so compare them with care.

6 UK records6 US records3 global records$0 to read each one

01 / The records

The yields are small, and that is the point

A register lists everyone it covers. Almost all of them are not in the market for anything. The work is the narrowing, and each step has its own count.

116,732

payment reports in one register

6

companies worth showing, after four checks

15 free records from our runs. "What it gave" is that run's result, not a rate you should expect.
RecordDatesWhat one run read, and what it gave
UK payment practices register UKeach statutory payment report116,732 reports; 2,028 filed in the window; 195 worse by 10+ points; 87 also worse on a second measure. 6 shown.
The Gazette, insolvency notices UKwinding-up petitionsAbout 725 petitions not brought by the tax office, 517 by companies. 4 shown.
Companies House accounts and charges UKfilings, officers, charges871,121 active companies in the relevant industry codes, filtered from the free bulk file; 489 accounts hits judged, 31 customer failures in companies' own words.
Find a Tender and Contracts Finder UKa public buyer going to market29,264 + 3,935 + 4,643 notices and 8,194 awards. 5 buyers going to market for the system in their own words.
Traffic Commissioner Applications and Decisions UKmore vehicles, new operating centres3,638 licence variations in 90 days. 10 shown, with procurement notices.
HSE releases, MAIB and RAIB reports UKa serious incident, named59 releases read. 2 shown.
SEC Form D USa private raise8,122 records from four sources. 156 of 340 final companies were in Form D only.
SEC full-text search USwording in 8-K and 10-Q filings79 phrases, 46 filers, 1,054 documents read. 4 strong.
FDA warning letters and recalls USa quality failure67 letters read in full; 388 recalling firms. 7 shown.
CMS Star Ratings USa health plan's rating, year on year76 contracts fell; 32 below 4 stars at 28 parents. 10 strong.
State Attorney General breach notices USa data breach notice7,166 filings became 5,118 companies in under 6 minutes. 148 shipped.
OSHA injury register (Form 300A) USinjuries and hours worked383,283 rows; 4,283 establishments in 6 target sectors at 500 to 5,000 employees; 117 also hiring the right seat. 10 shipped.
Science Based Targets file Globala climate commitment and its deadline856 companies in one industry. 3 shown.
Certificate transparency logs Globala new test, staging or internal web addressOne buyer study of 20 deals: 3 of 20 in the buying year against 3 of 20 the year before.
Ransomware leak-site listings Globala public ransomware listing12,366 victims listed free. In one buyer study all 28 buyers were checked: 1 was listed, before its buying year.

02 / The second record

A record dates the change. A second one makes it a reason.

One record alone

A single entry proves something happened. It does not prove the company needs anything. In the payment register, 2,028 companies filed a report inside the window. That is a list, not a target.

The same holds for job ads: the safety-manager job ad from one of our 67 pitched signals put incident investigations in the duties on 12 of 14 posts. It fires on almost everyone.

A change, then a second dated record

Compare the company's report with its last one: 195 were worse by 10 points or more.

Check a second measure in the same filing: 87 were worse on both.

Read both report pages in full, and look for a second dated record in the company's own words: 38 read, 6 shown.

03 / Who, not when

Some records fire on everyone, or after the fact

A record can be dated to the day and still say nothing about timing. We checked these against each buyer's own year before, as in our study of 89 purchases.

28 of 28

Compliance deadlines.

Every buyer in one study had a binding deadline inside its buying year. A filter that keeps everyone sorts no one.

0 of 28

Cyberattacks.

5 of 28 buyers had been attacked at some point, 1 of them on a leak site. None in the year before they bought.

0 before, 3 after

Ownership changes.

In one study of 20 purchases, the register recorded the sale or new entity only after the purchase.

3 of 20 vs 3 of 20

Test and staging web addresses.

Exactly as common the year before. They tell you which companies build in public, more than when they buy.

04 / What is missing

Some records hide most of a market

A free register looks complete. Check what it leaves out before you trust its count.

156 of 340

funded companies were only in Form D, in no funding news

but
4,646

of 7,946 records were funds, vehicles or public companies

01

Tenders hide the choice.

Many public buyers choose software through a framework and publish only the award. Council forward plans were readable for 65 of 325 councils, with 1 hit.

02

Headcount is per site.

The injury register counts employees per establishment. Three household-name giants read as mid-market. Take size from another source.

03

Search returns boilerplate.

In full-text filing search, "wage and hour" returned 2,113 hits and "manual processes and spreadsheets" returned 9. Only the rare, specific phrase is an admission.

05 / Honest notes

Reading traps we hit

  • The page date was not the event date. One press release carried a meta date of 30 April 2026; its dateline and web address both said 11 November 2025. Trust the dateline and the address.
  • Filers type their own dates. In one breach register, 27 filings had a breach date after the reported date and 694 had none. Use the date the register received it.
  • A host stopped answering. After a burst of about 130 pages a minute, a government page host refused reads; it answered again within half an hour. After that we read one page every 2 seconds.
  • A filter caught everyone. A deadline list and a cyberattack list both looked like signals until we checked the year before.

Doors that are closed

  • The main certification registers, global and EU. One blocks its record pages to crawlers, one blocks search, one is free to read but carries no dates, and one needs a full browser. A certification signal there can find a yes but never earn a no.
  • Limits of this guide. Each funnel is one run with its own bar. These are research results, not sales. Registers change their formats; check the endpoint before you build on it.

06 / Takeaway

The free records atlas

All 15 records in one sheet: where to read each one, what it dates, which date to trust, what one run got from it, and the trap. It works the same for an agency or a market-entry consultant building a client's list.

Download the atlas (CSV)

Before you build on a register
  1. 01

    Name the change you need, not the record: a worse payment report, a new licence, a filed incident.

  2. 02

    Pull the whole register once and keep it, so a new window costs nothing.

  3. 03

    Compare each company with its own last entry. The change is the signal, not the entry.

  4. 04

    Use the date the register gives, not one the filer typed or the page stamped.

  5. 05

    Check it against the year before. If it fires as often, it says who, not when.

  6. 06

    Ask for a second dated record from the same company, in its own words.

  7. 07

    Read slowly. A burst of about 130 pages a minute got one government host to refuse reads.

Want the records read for your market?

We read the registers for your market and show you the companies whose records changed.

Book a call See the case studies