Methodology
Signals, not verdicts. Nothing here is a credit rating, a prediction, or a finding of wrongdoing.
What is checked, and from where
- Companies House (UK registry): status, incorporation date, overdue accounts and confirmation statements, Gazette strike-off notices, registered-office and name changes.
- GLEIF (LEI records): whether a Legal Entity Identifier exists, its status, and whether its name and address agree with the registry.
- SEC EDGAR (US registrants only): late-filing notices, non-reliance on past financial statements, auditor changes.
- OFAC SDN list and UK Sanctions List: name screening against the lists. A match is only ever a possible match for you to review.
- DNS and RDAP: the company’s website domain: age, mail setup, whether it resolves.
- crt.sh (certificate transparency): when the domain’s first public certificate was issued.
- Lookalike domains: similar domains that are registered and set up for mail.
- News mentions (GDELT): only when this source is switched on. It is not part of a report otherwise.
All lookups are passive reads of public sources. We never contact the company or scan its systems, and we do not look up individuals. Every signal in a report shows its source and when it was retrieved.
How the overall level works
- Each signal is Info, Low, Medium or High. Info never counts: it is context, such as the incorporation date (also for a company under 12 months old), an overdue confirmation statement, a recent name change or a clean sanctions check.
- Any High signal makes the overall level High. Three or more Medium signals also make it High.
- One or two Medium signals, or three or more Low signals, make it Medium. Otherwise it is Low.
- Insufficient data appears when no source could be checked, or when the company’s main registry (Companies House for a UK company, otherwise GLEIF) or the sanctions screening could not be queried and nothing Medium or High was found anyway. Without the registry, “nothing found” does not mean Low.
- The level is a way to sort what to read first. The signals behind it are what matter.
The back-test, in plain English
We checked whether the Companies House signals separate UK companies that later went into liquidation or administration from comparable healthy ones, using only what the registry showed 12 months before the event. Two samples: 100 failed companies with 100 healthy matches, then an independent sample of 299 and 299. Healthy companies were matched on type, industry and age.
- Most companies that later failed showed no registry signal 12 months earlier. In the larger sample 254 of 299 (85%) were Low overall.
- Accounts overdue was about twice as common before failure (14% of failed companies against 5% of healthy ones in the larger sample), yet 1 in 20 healthy companies shows it too.
- Gazette strike-off notices were about 2.7 times as common (11% against 4%), mostly notices that were later withdrawn.
- Repeated registered-office changes were more common among failed companies (13 against 2 over both samples). This level is to be re-checked on a third sample.
- Confirmation statement overdue and a company less than 12 months old showed little or no difference in the larger sample, so they are shown as context only and never count towards the level. The same goes for a recent name change, which was too rare to judge.
Diligence does not predict insolvency. In real life failure is rare, so even a signal that is twice as common before failure mostly appears on healthy companies. A signal is a reason to look closer, not a verdict. The full numbers, intervals and method are on the back-test page.
Limitations
- Only the UK registry signals were back-tested. Their levels are set from that evidence.
- Not back-tested: the website footprint (DNS, mail, certificates), lookalike domains, sanctions name matches, SEC EDGAR signals and LEI consistency. Their levels are cautious starting points and capped accordingly, because these sources only show today’s state and cannot be rebuilt for the past.
- The samples are small and cover one lead time (12 months), UK limited companies and PLCs only. They describe those companies, not every company.
- Registers can be late, incomplete or wrong, and a source can be unavailable. The report says how many sources were actually checked.
- Sanctions results are possible name matches. A match needs a human to confirm it.