UK Company Due Diligence for Data Teams: Beyond Companies House

If you work with UK company data, you already know Companies House is free, comprehensive, and patchy in all the places that matter. It tells you a company exists, who runs it, and what they filed. It doesn't tell you whether you should do business with them.

For data teams supporting risk, compliance, procurement or sales, this is the gap that quietly causes the most damage. Below is what a defensible due-diligence dataset actually contains in 2026.

The seven signals that matter

1. Live registry status. Active, dissolved, in liquidation, proposal to strike off. Re-pull this monthly at minimum — a "live" supplier from six months ago can be in administration today.
2. Sanctions and PEP screening. Run every director (not just the company) against UK OFSI, US OFAC, EU and UN lists. Politically Exposed Persons need enhanced review, not exclusion.
3. Phoenix-company detection. The same directors dissolving and re-incorporating under a new name is one of the cleanest fraud signals available — and Companies House won't surface it for you.
4. Charges and PSCs. Outstanding charges, recent satisfaction filings, and changes in significant control reveal financing pressure and ownership games that financial statements lag by 12 months.
5. Address red flags. Hundreds of companies registered to one residential address. A mass-market formation agent. A serviced-office that vanished.
6. News and adverse media. Industry-specific outlets matter as much as nationals. AI summarisation is fine for triage; always store the original citation.
7. Filing punctuality. A company filing accounts on the last day of the extension every year is a behavioural signal worth scoring.

Building this as a data team

The naive build is "scrape Companies House, then bolt on sanctions lookups." It works for a month and then collapses under rate limits, schema drift, and the sheer volume of free-text matching required across sanctions lists. A more durable architecture:

• Use Companies House as the identity layer only — company number is your join key for everything else.
• Materialise a director graph separately. Most interesting questions ("show me other companies these directors run") are graph queries, not table queries.
• Treat sanctions screening as a streaming check, not a batch one. Lists update daily; nightly re-screens catch newly-added names.
• Score, don't gate. Output a 0–100 risk score with the contributing signals exposed, so analysts can override with context.

The shortcut

If you'd rather not build all of this, our Company Checkr does it as a service: live UK registry data, a 0–100 risk score, sanctions and PEP screening on every director, phoenix-company detection, charges, PSCs, address red flags, and AI-summarised news — all cited, with bulk CSV checks, watchlists with email alerts, and director profile pages. It's the same workflow the rest of this post describes, productised.

How it plugs into the rest of your stack

UK due diligence sits naturally next to AI enrichment (what they do) and fuzzy matching (which entity they actually are). Together those three steps turn a raw company list into something you can confidently make decisions on — which is the whole point of a data team.