What UK Company Data covers, what it does not, and how to cite a dataset

Coverage in one table

UK Company Data is built from two public Companies House sources. The figures below describe what a researcher can expect before designing a study.

Processing is described in full on the methodology page.

How often each field is filed

Availability depends on what companies are required to publish. For accounting periods ending in 2023 or later:

Balance sheet fields are widely available across all company sizes. Profit and loss lines are rare and come mostly from medium, large and group companies.

Why turnover and profit are rare

Most UK companies are small or micro-entities, and the Companies Act lets them file accounts at Companies House without a profit and loss account. Many file only a balance sheet and notes, often as micro-entity or filleted accounts (see abridged and filleted accounts). The turnover is not missing from our processing: it was never published.

This matters for research design:

• A study of profitability or sales growth across all UK companies is not possible from public accounts. Samples with turnover are dominated by medium and large companies, so results do not generalise to small firms.
• Balance sheet measures such as net assets, cash and employee numbers cover most companies and suit studies of small firms.
• Use the has filter (for example has=turnover) to build a sample where a field is present, and report the selection effect in your method section.

What is not included

• Companies outside the UK.
• Ownership trees and group structures.
• Credit scores or ratings.
• M&A transactions.
• Director networks.
• PSC data (people with significant control). See what a PSC is.
• Historical SIC codes. The SIC code is the company's current one from the register, so a company that changed activity appears under its present code for every year.

Paper and scanned accounts

Accounts filed on paper or as scanned PDFs are not in the Companies House bulk data, so they are not in the standard datasets. We extract them on request. In our tests on 140 documents, field accuracy was 97% to 98.5%. Up to 200 company-years a year are included in the campus licence, and these extractions are checked by hand.

Data rules researchers should know

• One row per company and accounting period.
• Amended accounts replace the original. When a company files amended accounts, the later filing replaces the earlier one.
• Blank is not zero. A blank cell means the figure was not filed or not extracted. Never treat it as zero.
• Implausible values are removed. Values over GBP 100bn are treated as parsing errors and removed.
• Every row is traceable. Each row links to the company's filing history at Companies House, so any figure can be checked against the original document.

Building a dataset

The dataset builder and the API (/v1/dataset for JSON, /v1/dataset.csv for CSV) accept these filters: sic (SIC prefix of 1 to 5 digits, comma list), town, postcode (area such as L or district such as L18), numbers (company numbers for a peer group), status, accounts (accounts type), inc_from and inc_to, year_from and year_to, min_employees and max_employees, min_assets and max_assets (total assets less current liabilities), has, latest=1, financials=0 (register only), fields and as_of. A campus licence allows 50,000 rows per query.

Reproducibility with as_of

The as_of parameter (format YYYY-MM-DD) rebuilds a dataset from filings published up to that date. Because companies keep filing new and amended accounts, the same query run a year later would otherwise return different rows. Recording the query and the as_of date lets an examiner or reviewer reproduce the dataset exactly.

How to cite a dataset

Include the attribution, the date and the exact query. A suggested format:

Keep the downloaded CSV with your thesis or deposit it with your paper.

FAQ

Why is turnover missing for most companies?

Small and micro companies are not required to publish a profit and loss account at Companies House. Turnover appears in 1.7% of filings for periods ending 2023 onwards, mostly from medium, large and group companies.

Does a blank value mean zero?

No. A blank means the figure was not filed or not extracted.

Can a dissertation dataset be reproduced later?

Yes. Add the as_of date (YYYY-MM-DD) to the query and the dataset is rebuilt from filings published up to that date.

Are scanned PDF accounts included?

Not in the bulk data. We extract them on request, and up to 200 company-years a year are included in the campus licence.

Sources: Companies House register and accounts data (Open Government Licence v3.0), legislation.gov.uk and GOV.UK guidance where linked. This page is general information, not legal, tax or credit advice. Spotted an error? Email info@globalmindsetmarketing.co.uk.