Where private company data actually comes from

There is no secret source of private company data. Every vendor, from the largest bureau to a freelancer with a spreadsheet, is working from the same four kinds of public material. What separates a useful file from a useless one is not access. It is how carefully the four are joined together, and how honestly the joins are labelled.

Knowing which source a field came from tells you almost everything about how far to trust it.

1. The statutory register

The spine of every company record. In each jurisdiction some authority maintains the list of incorporated entities: Companies House in the United Kingdom, the Companies Registration Office in Ireland, ACRA in Singapore, the Companies Registry in Hong Kong, ASIC in Australia. In the United States there is no federal equivalent, and the register is whichever state the entity was formed in.

What it gives you reliably: the legal name, the company number, incorporation date, legal form, current status, registered office, and the officers as last filed.

What it does not give you: whether the company trades, what it actually does today, where it operates from, or how to contact anyone there.

The trap: treating register presence as evidence of a business. Registers are cumulative records of incorporation, not curated lists of going concerns, and in every jurisdiction worth covering the non-trading entities are numerous enough to swamp a naive export.

2. Filed accounts

Where a jurisdiction requires companies to file financial statements publicly, those filings are the only genuinely reliable financial data on a private company.

The catch is the exemption regime. Small companies in most jurisdictions may file abridged accounts: a reduced balance sheet, no profit and loss, no turnover. Some jurisdictions exempt most private companies from public filing altogether. So the honest position is that for the large majority of private companies, revenue is simply not public — and any database showing it for most of its rows has modelled it.

What it gives you reliably: balance sheet items in jurisdictions that require them, filing dates, auditor where one is appointed, and dormancy, which is often stated explicitly.

The trap: reading a modelled turnover figure as a filed one. If a file does not distinguish them, treat the entire column as an estimate.

3. The company’s own website

The most under-used source in company data, and the only one that tells you what a business does now rather than what it registered as doing years ago.

What it gives you reliably: the current activity in the company’s own words, the trading address, the general enquiries email, the switchboard, the trading name where it differs from the registered one, and often the senior team.

What it does not give you: anything verifiable without a second source. A website establishes what a company says about itself, which is exactly what you want for activity and exactly what you should not rely on for existence or status.

The trap: stale sites. A live website is weak evidence of a live business on its own, which is why it belongs alongside a registry status rather than instead of one.

4. Public announcements and regulatory filings

Transaction announcements, tenders, licence registers, court and insolvency notices, and — for anyone with a US public parent or a public counterparty — the disclosure filings that name them. EDGAR’s full-text search is genuinely useful here for exactly this reason: private companies appear constantly inside other people’s filings.

What it gives you reliably: events. An acquisition, a licence, an insolvency, a name change, a contract award.

What it does not give you: a population. Announcements are heavily biased towards companies that do newsworthy things, so building a list from them systematically excludes the quiet, profitable businesses that are usually the point of the exercise.

The join is the product

Any one of these sources is available to anyone. The work — and the only thing worth paying for — is joining them correctly for one company at a time:

• Matching a registry entity to the real trading business, which for group structures means deciding which of several entities is the one you would actually contact.
• Reconciling the filed industry code with what the company says it does, and keeping both rather than picking one.
• Separating the registered office from the trading address instead of collapsing them into a single “address” column.
• Confirming that the entity is trading, using a status from the register plus an independent signal, rather than assuming it from active status alone.
• Leaving a field blank when no source supports it.

That last one is the discipline that decides whether a file can be put in front of a client. Every other step adds information. That one preserves the meaning of everything else in the row, because a column where blanks have been filled in is a column you can no longer cite.

How to read a vendor’s methodology page

Three questions get to the bottom of most of it.

Which of the four sources does each field come from? A vendor who has thought about this can answer field by field. One who cannot is selling a join they have not inspected.

What is the rejection rate? Real research rejects candidates — dormant entities, unmatchable records, businesses that turn out to have ceased trading. A process with no rejections is not a process.

What does a blank field mean? If the answer is “not published”, the file is citable. If the answer is anything else, find out what fills the gaps and treat those columns accordingly.

Sources

1. Companies House — Government of the United Kingdom; accessed September 6, 2026
2. Companies Registry — Government of the Hong Kong Special Administrative Region; accessed September 6, 2026
3. Companies Registration Office — Government of Ireland; accessed September 6, 2026
4. EDGAR full-text search — U.S. Securities and Exchange Commission; accessed September 6, 2026