Most UK company accounts do not contain a turnover figure

People arrive at UK company data expecting revenue. It is the first question anyone asks about a business, the register is free and public, so the number must be in there somewhere.

For most companies it is not, and no API can conjure it. We measured how often it is actually present across 26,112 real company lookups, and the answer surprised us even though we have been parsing these filings since February 2026.

What you get when you ask 26,112 times

What came back	Share
Accounts we can parse into figures	42.3%
Accounts filed, but no machine-readable version published	35.9%
Company dissolved	14.5%
Never filed accounts at all	7.3%

Before any question of turnover, more than a third of lookups end at a document that is not machine-readable. Companies House holds the filing and will show it to a human, but there is no structured version to extract figures from. Older registrations are worse: for companies incorporated before 1998, 55.4% have no machine-readable accounts.

Of the accounts that do parse, almost none carry turnover

Of the 11,049 filings we could parse, roughly 6 to 8 per cent carry a turnover figure. Across all lookups, that is around one in thirty.

The range is deliberate. Our measured figure was 6.4%, but we re-checked 40 filings by hand against the original documents and found we had missed a turnover tag in 4 of them, so our own parser understates it by something like a tenth. We would rather publish the honest range than a precise number we know is slightly low. The underlying defect is now on our issue tracker.

The reason so few carry it is law, not obscurity. 85% of parsed filings are micro-entity, dormant, or small and filleted accounts, and none of those formats require a profit and loss account to be filed. A company can trade profitably for twenty years and never publish a revenue figure, entirely legitimately.

Even the full accounts are not a safe bet: of filings Companies House labels full, only about 36% contain turnover.

Among the companies that do publish it, the median turnover is £12.1m - which tells you who they are. Publishing revenue is mostly a big-company behaviour.

Three real filings

Every one of these is checkable today:

Gloucester Rugby Limited (00034603) files full accounts, and its turnover is right there: £13,438,261 for the year to 30 June 2025, against £14,922,948 the year before. This is what people imagine the register is like.

Liverpool Law Society (00004302) files small accounts. The balance sheet parses cleanly - net assets £298,161 - but the document contains only a revenue-recognition policy note, describing how the society would recognise revenue, with no revenue figure anywhere. A tool looking for words about revenue finds some. A tool looking for a number finds nothing, because there is nothing.

Lloyds Bank PLC (00002065) filed its group accounts on 19 March 2026, and Companies House serves that filing as a PDF only. One of the largest banks in the country, and there is no machine-readable version of its accounts on the register.

Why we return "not available" instead of a number

This is the part that matters if you are building on any of this.

When a filing does not contain turnover, there are two things an API can do. It can say so. Or it can produce something turnover-shaped from whatever else is in the document - gross profit, a figure from a policy note, a number scraped out of a PDF - and hand you that.

The second is worse than useless, because you cannot tell the difference. A wrong revenue figure looks exactly like a right one, and it will be wrong in a way that survives all the way into someone's credit decision or pricing model.

So our financials response carries an explicit available: false with a reason when a filing cannot be parsed, and returns a null turnover, rather than a guess, when the document simply does not contain one. Each response also carries a data_quality block naming which fields were attempted and which were found. That is a narrower promise than "we have financials for every UK company", and it is the only one that survives contact with the register.

Anyone handing a raw filing to a language model and asking it for turnover is doing the second thing, whether they mean to or not. The model will nearly always return a number.

Method, and what would make these figures wrong

What we measured. All 88,546 rows of our production financials cache, read on 24 September 2026. The headline percentages use the 26,112 rows written since 6 September 2026, the point from which company status was recorded alongside each result, so that dissolved companies could be separated out instead of hiding inside the other buckets. The three named examples were re-checked against the live Companies House register the same day.

The main caveat. This is a sample of companies people looked up through Registrum, not a random sample of the register. It skews toward companies that someone had a reason to research, which means it under-represents the smallest micro-entities. Those are the least likely of all to publish turnover, so the true proportion of UK companies with no published revenue figure is probably higher than the figures above, not lower. Read every number here as "in a sample of 26,112 lookups", never as "of UK companies".

What we left out. We deliberately publish nothing here about profit before tax, tax or depreciation. Our extraction of those fields is incomplete, so any figure we quoted would measure our parser rather than the register. Two issues are open to fix that, and we will publish those numbers when the fields are trustworthy.

Checking this yourself

• Pick any company on find-and-update.company-information.service.gov.uk, open its filing history, and look at an accounts filing. The format is stated on the filing.
• Compare with ours at https://registrum.co.uk/company/00034603 for a company that publishes turnover, and https://registrum.co.uk/company/00002065 for one where no machine-readable accounts exist.
• If you find a filing that contains a turnover figure we report as absent, that is exactly the defect described above, and we would like to hear about it.