Teaching kits with real UK company accounts: three ready-made datasets

What the kits are

Each kit is one CSV of real UK company accounts, frozen at a snapshot date so every student gets the same numbers and the exercise can be repeated next year. Every row is one company and one accounting period, taken from the digital (iXBRL) accounts the company filed at Companies House, with a link to the original filing in the last column. Blank cells mean the item was not filed, never zero.

The kits are free for anyone to download. A campus licence adds the live dataset builder with up to 50,000 rows per query, so students can build their own versions for coursework and dissertations.

Source line for slides and reports: "Contains Companies House data, Open Government Licence v3.0; processed by UK Company Data."

Kit 1: liquidity and solvency of Liverpool restaurants

For accounting and finance modules, first or second year. 60 minutes.

Learning outcomes: read a small company balance sheet; compute the current ratio and net current assets; explain why negative net assets are common in hospitality; recognise what small company accounts do not disclose.

Session plan:

1. 10 minutes. Open one company's row and its original filing from the Source link. Match each column to a line on the balance sheet.
2. 20 minutes. Compute the current ratio. Only 42% of rows have "Creditors within one year", but 93% have "Net current assets": derive creditors as current assets minus net current assets. Discuss why the two columns differ in coverage.
3. 20 minutes. Count the share of rows with negative net assets (about 28%) and the median current ratio for rows where it can be computed (0.91: more than half of these restaurants owe more in the next 12 months than they hold in current assets). Ask: is this a sign of distress, or of how restaurants are financed (supplier credit, director loans, leases)?
4. 10 minutes. Turnover is filed for only 18 rows. Why? Link to the small companies regime and the glossary entry on micro-entities.

Teacher notes: the median net assets jumps between years (for example £1,454 in 2016 and £10,114 in 2021) because the set of companies filing digital accounts changes each year. This is a good moment to introduce survivorship and composition effects before students compare years.

Kit 2: Merseyside IT companies, a dissertation-style study

For business research methods or a final-year project. 90 minutes, or a two-week assignment.

Learning outcomes: turn a research question into dataset filters; describe a sample honestly (what is in, what is out); identify survivorship bias; write a reproducible data section for a methodology chapter.

Tasks:

1. Write one research question that this dataset can answer and one it cannot. (It cannot answer anything about profitability for most firms, because small companies rarely publish turnover or profit.)
2. Run the R script: companies per year, share reporting employees, median net assets by year.
3. Compare net assets in 2018 and 2024 for companies present in both years. Then explain why this "survivors" sample overstates growth: the dataset contains active companies only, so firms dissolved before October 2026 are missing.
4. Rebuild the dataset with status set to All in the dataset builder and discuss what changes. Write the methodology paragraph, citing the rebuild link and the snapshot date.

Assessment idea: a 1,000-word data note marked on question fit (30%), sample description and limitations (40%), reproducibility (30%).

Kit 3: a North West employer panel for data science

For data science, analytics and AI programmes. 90 minutes in a lab.

Learning outcomes: profile missing data in a real panel; compute ratios without dividing by zero or by missing values; build a prediction target without look-ahead; split train and test data by time; check an AI assistant's answer against verified figures.

The notebook builds a simple model of "net assets negative in the next period" from the current ratio, cash ratio and whether net assets are already negative. With a time split (train before 2022, test from 2022) it reaches an AUC of about 0.89 on 6,957 company-years, most of it from persistence: companies with negative net assets tend to stay negative. Students then re-run with a random split and explain why the score changes.

Teacher notes: a small number of rows have negative current assets or cash, which are filing or tagging errors. The notebook drops them after inspection; ask students to find them first.

FAQ

Can students use these datasets in assessed work?

Yes. The data comes from Companies House under the Open Government Licence v3.0, and the kits are free to use for teaching and research with the source line above.

Why are the snapshots dated 21 October 2026?

The snapshot fixes the set of filings, so a marked exercise gives the same answers every time. The rebuild links recreate exactly the same rows from the live data.

Can you make a kit for our module?

Yes, as part of a pilot or a campus licence we adapt one existing assignment to this data. Email info@globalmindsetmarketing.co.uk with the module outline.

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.