Using Companies House Data to Find Acquisition Targets: A Practical Guide

Companies House might be the most underused free dataset in British business. Every limited company in the UK is on it: 5.48 million records, of which 4.93 million are live once you strip out those in dissolution or liquidation (Companies House, March 2026). For each one, you can see who runs it, how old they are to the month, how long they've been in charge, what the company owes money against, and how promptly it files its paperwork.

For anyone trying to buy a business, it's a map of every possible target in the country, with the owner's circumstances noted in the margins.

I've spent a lot of time in this dataset building ProspectMap, and this guide is what I wish someone had handed me at the start: what's in a company record, which fields matter for acquisition work, and where the data will mislead you if you take it at face value.

What you can get, and what it costs

The full register costs nothing to search or download, and that still surprises people. Since 2015 it's been free to search and free to download in bulk. There are three ways in:

1. The search site (find-and-update.company-information.service.gov.uk): fine for looking up one company, useless for building lists.
2. The API: free, with generous limits, and it returns everything the search site shows. This is how you enrich records one by one.
3. The bulk data products: a downloadable snapshot of basic details for every live company, refreshed monthly. One file, every company in the country. This is where target lists start.

The bulk snapshot gives you the skeleton: name, number, registered address, incorporation date, status, SIC codes, and the category of accounts filed. Directors, charges, filing history and PSCs come from the API, record by record.

The fields that matter for acquisition work

SIC codes: useful, flawed, unavoidable. Every company declares up to four SIC codes: five-digit numbers from the ONS's 2007 classification describing what it does. There are around 730 of them. SIC codes are how you cut the register down to your sector, and you need to know three things about them.

They're self-reported. Nobody checks that a company registered under 43220 (plumbing and heating installation) installs anything at all. Codes go stale too: plenty of companies picked one at incorporation in 2004 and never revisited it as the business changed. And the same real-world trade spreads across neighbouring codes: a heating firm might legitimately sit under plumbing installation, electrical installation, or "other specialised construction." Search one code and you'll miss real targets sitting one code over.

The practical fix is to search generously: take the whole cluster of related codes, then filter the noise out afterwards, rather than search narrowly and never know what you missed.

Officers: the most valuable table in the register. Every current and former director is listed, with their appointment date and, this is the part that matters, their month and year of birth. That's public, for every company in the country.

Combine three fields and you can read an ownership situation from the outside: director count (one director usually means owner-operator), appointment date (twenty years in post means the business is them), and date of birth (an owner born in 1958 is thinking about things an owner born in 1988 isn't). None of this proves the owner wants to sell. It shows you where the odds are better, and when you're choosing 50 letters to write out of 500 candidates, odds are the whole game. I've written up how we combine these into a succession score in the succession risk guide.

Accounts: reading size without seeing turnover. Most small companies don't publish turnover: they're allowed to file abbreviated accounts. But the category of accounts a company files is itself a size signal, because the filing regimes have thresholds. As of accounting periods beginning on or after 6 April 2025, a micro-entity must stay under £1m turnover and roughly 10 staff; a small company under £15m and 50 staff; a medium company under £54m and 250. Thresholds rose about 50% in the April 2025 changes (before that it was £632k and £10.2m for micro and small), and most accounts currently on the register were still filed under the old thresholds, so read older filings against the old limits.

So "total exemption full" or "unaudited abridged" in the accounts field doesn't give you a revenue figure, but it brackets one. For sizing a first acquisition, brackets are usually enough. Where a company does file fuller accounts, you get balance sheet detail too: net assets, and sometimes employees and turnover.

Charges: the borrowing history. A "charge" is security registered against the company's assets: a bank loan secured on the premises, an invoice finance facility, sometimes a director's own loan formalised. The register shows every charge ever created, who the lender was, and whether it's outstanding or satisfied.

For a buyer, this section is one of the most useful in the whole record. No outstanding charges on a 25-year-old business often means it's debt-free. A stack of recent charges from alternative lenders can mean growth, or trouble. Either way, you walk into the first conversation knowing the shape of the balance sheet before the owner has told you anything.

Filing history: the company's pulse. Every company must file accounts and an annual confirmation statement, on deadlines. The pattern is telling. Twenty years of on-time filings is the signature of a business, and an accountant, in good order. Recent slippage into lateness is worth noting: sometimes it's nothing, sometimes it's the outward sign of an owner losing interest or a business losing its grip. We flag anything that drifts past its deadlines rather than excluding it. A flag is information; an exclusion is a guess.

PSCs: who owns it. The People with Significant Control register lists who owns 25% or more of a company. Usually it confirms what the officer list implied: the sole director owns the thing. When it doesn't, when the PSC is another company, or someone who isn't a director, you've learned something important before wasting a letter on the wrong person. The person running the business may not be the person who can sell it.

The cleaning problem nobody warns you about

Run a raw SIC-plus-location query and you'll feel briefly rich: hundreds of matches. Then you look closer. Dormant companies that trade nothing. Property shells holding a single building. Companies incorporated eight months ago. Holding companies with no operations. Registered offices that are an accountant's address in a town the business has never operated in.

On real runs, half or more of the raw matches turn out to be companies you'd never write to. The filters that do the heavy lifting:

• Status — active only.
• Accounts category — drop dormant; "initial" means brand new.
• Company age — a minimum of two or three years removes most shells.
• Filing history — a company that has never filed accounts is usually not a trading business.
• Address type — a check for registered offices that look like an accountant's address rather than trading premises.

This cleaning is exactly the sort of work software should do. Done badly, it quietly ruins a campaign: every letter to a dormant shell is postage spent learning nothing.

An honest note on data quality

The company register is self-reported: SIC codes can lie, addresses can mislead, and businesses can be dissolved in practice but look alive on paper for a year or more. Companies House has been tightening verification under its 2025–2030 reform programme, but treat any single field with some scepticism. A company that looks real across multiple independent signals usually is.

The register can't tell you whether the business is any good, or whether the owner will sell. What it does is narrow 4.9 million companies down to the few dozen worth asking those questions of, in person. That's real, useful work, and it's only the first step.

---