What public data can you actually get about a private company?

Almost every disappointing piece of competitive research fails in the same place. Someone sets out to answer "how big are they?" and, finding no answer, quietly settles for a number that looks like one — a traffic estimate, a headcount guess, a revenue figure from a directory that got it from another directory. The number then hardens into a fact, gets put in a deck, and decisions are made against it.

The fix is not to try harder. It is to know in advance which questions have public answers, which have only public estimates, and which have none at all. Then you ask the answerable ones.

Tier one: things that are genuinely on the record

These come from sources with a legal or contractual reason to be accurate. They are dated, attributable, and checkable by someone who doubts you.

Corporate registry filings

Nearly every country maintains a public company register. What it contains varies enormously — this is the single biggest thing people get wrong when they generalise from one jurisdiction to another. In the UK, Companies House publishes incorporation dates, registered addresses, directors, shareholdings and annual accounts, and even the smallest companies file something. In the US, registration is state-level and typically confirms far less: existence, formation date, registered agent, and sometimes officers, with no financial statements. Other jurisdictions sit at every point between.

So the reliable move is to check what that register actually publishes rather than assuming. Where accounts are filed, note that small-company filings are often abbreviated and lag by many months — they tell you about a year that has already ended, not about now.

Intellectual property and regulatory records

Trademark registers show what names and marks a company thought worth protecting, and when — useful for dating a pivot or spotting a product not yet launched. Patent databases do the same for technical work. Sector regulators publish licences, registrations and enforcement actions. If a company operates in a licensed field, its licence status is a hard public fact, and a lapsed one is a significant finding.

Court and enforcement records

Litigation, judgments and insolvency proceedings are largely public in most common-law jurisdictions. Access varies and some of it is paywalled, but existence is usually discoverable. Treat with care: a filed claim is an allegation, not a finding, and repeating it as fact is both wrong and risky.

Their own public statements

Pricing pages, terms of service, job postings, changelogs, documentation, status pages and support policies are all things the company published deliberately. They are marketing, so they are not neutral — but they are also commitments, and companies are notably more careful in a terms-of-service document than in a headline. The terms often reveal the real shape of the business: minimum contract lengths, what is actually included, how they handle refunds, whether there is a hidden onboarding fee.

Dated third-party mentions

Press coverage, conference speaker lists, partner directories, integration marketplaces, procurement and grant records, and archived versions of their own site all carry dates from outside the company's control. Archived pricing pages in particular are one of the most underused sources in competitive research: they turn a static claim into a history.

Tier two: real signals, but estimates

These are useful and worth gathering. They are also modelled, sampled or self-reported, and they must carry that label all the way into your conclusions.

• Traffic and keyword tools. Modelled from panels and clickstream data. Directionally informative when comparing two sites measured the same way; unreliable as absolute numbers, and weakest exactly where small businesses live — low-traffic sites.
• Headcount from professional networks. A count of people who chose to list an employer, not a payroll. It undercounts non-desk workforces badly and overcounts stale profiles. The trend over time is more informative than the number.
• Review counts and ratings. Real evidence of customer sentiment, drawn from a self-selecting minority of customers. Never convert a review count into a customer count or a market share.
• Technology-detection tools. Genuinely useful — they read what a site actually loads. But they see the front end only, they miss server-side systems entirely, and a detected tag may be a leftover.
• Directory and data-provider profiles. Often the most confidently wrong source of all. Revenue and employee figures in company databases are frequently modelled or scraped from another database, then cited by a third as though independently established. If you cannot trace a number to a filing or a statement, it is not a fact.
• App-store rankings and download estimates. Ranks are real and observable; download figures are inferred.

Tier three: not obtainable

For a typical private company with no public debt and no filing obligation, no legitimate research process will give you these:

• Revenue, margin or profitability, unless they published it or a filing carries it
• Customer count, churn, or contract values
• Cash position and runway
• Cap table beyond what registries or announcements disclose
• Anything about their pipeline, roadmap or internal plans

If a research output hands you these numbers for a private company, the correct first question is: from where? Occasionally there is a good answer — a founder said it on a podcast, a filing carries it, a customer disclosed a contract value in a public tender. Usually there is not, and what you have is an estimate that lost its label somewhere upstream.

The line we will not cross

There is a second category of unavailable: things that could technically be obtained but should not be. Scraping behind a login, creating a fake account to see a competitor's gated content, misrepresenting yourself to their sales team to extract a quote, buying scraped personal data, or pulling anything that violates a site's terms. These carry legal and reputational risk, and they tend to produce information you then cannot use or cite anyway.

Everything in this guide is obtainable from public sources without deception. That constraint is not a limitation on the research — in practice it barely narrows what you can learn, and it means every finding survives being questioned.

How to use the tiers

The practical method is to write the question down first, then locate it in the tiers before doing any work.

If it lands in tier one, go and get it, and record the source and date alongside the finding. If it lands in tier two, gather it, label it as an estimate every single time you write it down, and never build a conclusion on one estimate alone — require two independent ones to agree. If it lands in tier three, stop and reformulate: replace the unanswerable question with the answerable one that would inform the same decision.

That reformulation is where most of the value is. "What is their revenue?" is unanswerable. But you rarely need revenue — you need to know whether they are growing, whether they can outspend you, or whether their model leaves a segment unserved. "Are they hiring, shipping and raising prices?" is answerable from tier one, and it usually informs the decision better than a revenue figure would have.

What an honest finding looks like

Two sentences about the same company. The first: "They do around $3M and have about 20 staff." The second: "Their filed accounts to March 2025 show a balance sheet consistent with a small operating business; a professional-network count suggests roughly 20 people, which is an estimate, not a payroll; they have had three delivery roles open since April 2026 and raised their entry price twice in eighteen months per archived pricing pages."

The second is longer and much less satisfying to put in a slide. It is also the only one of the two you can defend, act on, and revisit later when the facts change — because every claim in it carries where it came from and when.

The short version

• Sort every question into on-the-record, estimated, or unobtainable before researching it.
• On the record: registries, IP and regulatory records, courts, the company's own published commitments, dated third-party mentions.
• Estimated: traffic tools, headcount counts, review volumes, tech detection, directory profiles. Keep the label attached.
• Unobtainable for most private companies: revenue, customers, churn, cash, roadmap. Reformulate rather than guess.
• Registry contents vary hugely by jurisdiction — check what that register publishes instead of assuming.
• An untraceable number is not a fact. Cite the source and the date, or leave it out.