What's actually in the Companies House bulk data download

Companies House gives away the entire UK company register, every month, for nothing. If you can handle a 2.8GB CSV, that is the right answer and we'll say so before we say anything about ourselves — every commercial UK company-data product, ours included, is built on top of this file. What almost nobody tells you is what is not in it, and how much of what is in it means something different from what it looks like. So we counted the whole thing.

The file, measured

Every figure on this page is counted from the 1 July 2026 snapshot in one streaming pass over all 5,734,779 rows — not sampled, not estimated, not carried over from a previous month.

The status split is the first thing that catches people out. A filter of status starts with “Active” pulls in 404,834 companies that are being struck off the register. Exact-match Active is what you want, and it is a one-character difference in most people's code.

What isn't in it

Checked against the header of the file itself, these columns do not exist: website, email address, phone number, turnover, employee count, director or officer names. Not empty — absent. That is the entire commercial gap in UK company data, and it is why an industry exists on top of a free file. Two of those are worth being specific about:

• No website. There is no way to answer “which of these companies has a live site, and is it any good” from the bulk file. Resolving a company name and registered address to the right domain — and proving it is the right one rather than a near-match — is the single most valuable thing you can add to this data. It has its own comparison here.
• No size, in the way you mean it. There is no turnover and no employee count. The closest proxy the file carries is the accounts category filed, which is a legal filing band rather than a measure of trade: 1,669,698 active companies file as micro entity, 583,301 as dormant, and only 79,334 file full accounts with 6,148 filing as medium. If your brief says “companies turning over £1m+”, this file cannot answer it and neither can anyone selling you a filter derived only from it.

The SIC code is always there — and often says “nothing”

Good news first: every one of the 5,213,454 active companies carries at least one SIC code — zero blanks in the whole file, which is better coverage than most people assume before they open it. 4,147,844 declare exactly one and 212,149 declare the maximum four — which is its own trap, because a company filing four codes is telling you it does four things, and picking the first one silently is a decision, not a read.

The catch is that the field is never blank, so the codes have to carry the “no industry” cases themselves — and 216,366 active companies file the literal value None Supplied. If your pipeline checks for an empty cell, it will find none, and quietly treat all 216,366 as a valid industry. On top of those, 100,647 declare 99999 — Dormant Company and 39,148 declare 74990 — Non-trading company: live companies whose valid industry code tells you they aren't trading. Together that is 356,161 active companies — 6.8% of the register — carrying an industry code with no industry in it. Filtering by SIC properly is its own how-to.

A third of the register lives at 5,313 postcodes

The registered office is the one address field the register publishes for everybody, so it is the field people build location lists from. Over the whole active register there are 944,546 distinct postcodes, and 373,075 of them hold exactly one company. Then there is the head of the distribution:

These are company-formation, accountancy and mail-forwarding addresses, not enormous businesses — we count the postcodes, we don't name the occupants, because the register gives us the counts and not a verified occupier. The rows are still real companies and may be exactly who you want; they are simply not evidence of a local presence. Cutting the register by town, county or postcode goes into what to do about it.

Three things in the file that are better than people expect

It is a free file and most of this page is about its limits, so here is the other side, all of it free and all of it usable as signal. Outstanding charges: 621,217 active companies (11.9%) carry at least one unsatisfied mortgage or charge — secured borrowing is a real indicator of a company with assets and lenders. Overdue accounts: 97,773 were past their accounts due date on the snapshot day, which is a distress flag you would otherwise pay a credit agency for. Previous names: 481,200 active companies have been renamed at least once, and the file ships up to ten former names with the dates — which is how you match a company to a brand that no longer matches its registered name.

And the register is growing fast enough to be worth watching on its own: 728,290 of today's active companies were incorporated during 2025, and 398,186 more in the first half of 2026 alone. Pulling just those is the new-incorporations cut.

What we deliberately don't do

We don't pretend the free file is worse than it is — for a lot of briefs it is genuinely all you need, and we would rather tell you that than sell you a CSV you could have downloaded. We don't sell named individuals, email addresses or direct dials: company-level data only, a deliberate GDPR stance rather than a gap we're planning to fill. And every run we ship is a dated snapshot of a register that changes daily — we say so, rather than implying continuous monitoring.

Frequently asked questions

Related: the full UK company data providers guide compares every route end to end, including Endole and the technology a company's site runs.