Data · Method

How we built the UK company cash dataset

Draft, not published

Source

Companies House publishes a monthly archive of every set of accounts filed digitally, free to download and with no key required. We used the archives covering June and July 2026. Nothing on these pages comes from any other source: no third-party statistics, no survey, no estimates. The build files it under one dataset line, which is what a monthly refresh replaces: Companies House monthly accounts bulk data, Jun+Jul 2026 filings.

Sample

10,000 filings randomly sampled (seed-reproducible), 6,863 with usable iXBRL facts; SIC joined from BasicCompanyDataAsOneFile-2026-08-01. The sample is drawn with a fixed seed, so the same filings can be pulled again and the figures reproduced. Industry codes come from the free BasicCompanyDataAsOneFile-2026-08-01 snapshot.

Balance sheet facts were extracted from each filing at its latest balance date: cash at bank, debtors, creditors due within and after one year, current assets, net current assets, equity.

What the sector pages leave out

Three groups of companies are in the UK-wide figures but not in any sector page. 81 dormant companies, 220 with no usable industry code, and 196 in sectors that did not reach the minimum number of sampled companies we require before publishing a page (agriculture 68, education 128). The UK-wide rows therefore describe the filing population as a whole, and the sector rows describe only the sectors big enough in this sample to publish.

What each figure means, exactly

These are the definitions the code applies, not a paraphrase of them. Note the denominators: they are not the same group of companies from row to row.

• With negative net cash. Cash at bank plus debtors, less creditors falling due within one year, is below zero, for companies that tagged all three. Reported as a percentage of that group (1,615 companies UK-wide).
• In net current liabilities. The company's net current assets, as tagged in its own filing, are below zero: what falls due within twelve months is larger than every current asset combined. A broader test than negative net cash, on a much larger group (6,424 companies UK-wide), and a separate measure rather than a restatement of it.
• Holding less cash than creditors due within one year. Cash at bank is lower than creditors falling due within one year, for companies that tagged both (2,311 companies UK-wide).
• Median cash as a share of creditors due within one year. The middle value of cash divided by creditors due within one year. This row has its own smaller denominator, 2,249 companies UK-wide, because it also requires creditors above zero: dividing by zero is not a ratio. 48% means the middle company's cash covers 48% of that bill.
• With £1,000 or less in the bank. Cash at bank at or below £1,000 at the balance sheet date, as a percentage of companies that tagged cash (3,347 companies UK-wide).
• Median cash at the balance sheet date. The middle value of cash at bank across the same group of companies that tagged cash.

The UK-wide figures, with their counts

The denominators below are the reason each page prints an n beside every value. They are not one number, and treating them as one would overstate what the sample can support.

Four caveats, stated up front

• The balance dates are mostly 2025. Accounts arrive at Companies House up to nine months after the year end, so filings made in June and July 2026 overwhelmingly describe balance sheets dated mid to late 2025. These pages describe the recent past, not this morning.
• There is no profit and loss in this data, so cover over time cannot be computed. Small companies file balance-sheet-only ("filleted") accounts. There is no monthly cost base in the filings, which means no honest way to express any of this as a period of cover. Every claim on these pages is therefore cash against what is owed, and nothing on them should be read as a length of time.
• Each figure has its own n. Companies do not all tag the same fields, so the denominators differ from row to row, and every table on this site prints the count next to the value. The median ratio row is smaller again, because it needs creditors above zero as well. Where too few companies in a sector tagged a field, we omit the figure rather than publish a thin one. The sample is companies that filed digitally in those two months; micro and dormant entities are included.
• Net current liabilities is not the same as insolvent. In the words of the source method: some healthy businesses run negative working capital by model, for example where customers pay up front. It is still a fair "no cushion" proxy for typical small companies, and that is a proxy, stated with its caveat, not a finding about any company in the figure.

What we will not claim from it

The dataset cannot say how long any company could keep trading, and we will not imply that it can. It cannot say whether a company is solvent. It cannot say what happened after the balance sheet date. It cannot separate small companies from large ones, because the filings do not reliably carry size. Where a figure is thin, it is absent from the page rather than dressed up.