3 Reasons Why UBO Identification Systems Fail at Scale
In November 2022, the EU shut its beneficial ownership registers to the public. In March 2025, the United States paused its own reporting rule. The United Kingdom never had a UBO register at all; it holds Persons of Significant Control, which is not the same thing. In other words, the data these systems need to work with is getting harder to reach.
That is the first of three reasons UBO identification systems fail at scale. The second is that the shareholder data analysts fall back on is rarely structured, often crosses borders, and frequently runs through jurisdictions that publish little or nothing. The third is that trusts are built to hide ownership, and the registers that might pierce them are slow and closed. I'll take each in turn.
1. Declared UBO disclosures are hard to get
Since the EU ended public access in November 2022, every EU register now requires applicants to pass a legitimate interest test before it releases anything. A legitimate interest means a recognised reason to see the data, such as fighting money laundering, that you must prove before the register will open to you. Access runs to journalists, obliged entities (including third-country ones), and civil society workers fighting financial crime.
Proving it is slow. The 6th Anti-Money Laundering Directive tells member states to hand over UBO data as machine-readable files through an API, but only Denmark has done so. Everywhere else, the process grinds along case by case. Transparency International waited a year for German disclosures and several months for Hungary. Ireland has released nothing since 2022.
The United States went the same way. In March 2025, the Trump administration removed the reporting rule for domestic companies. Under the interim rule, every US-formed entity and US person is now exempt, with a final rule still to come.
Even the UK, which built the Persons of Significant Control register in 2016, offers little of use. A PSC is not a UBO.
Analysts can't simply look up the answer. They must work it out from primary source shareholder data.
The C/D distinction: calculated vs declared
One is filed. The other is figured.
A declared UBO is the owner that a company files to an authentic source, such as a national register. It carries legal weight because someone filed it under a duty to be accurate.
A calculated UBO is the owner that a system works out by tracing shareholdings. If Company A owns 60% of Company B, and B owns 80% of Company C, the person behind A holds an effective 48% of C. Nobody filed that figure. The system computed it.
On the page, they look identical: a name and a percentage. But a calculated UBO and a declared UBO carry different evidentiary weight, and a system that pours both into one field, without telling you which is which, hides the seam.
Any system worth buying must be open about the C/D distinction. That means recording how it resolved each entity in the tree, whether a machine matched it or an analyst confirmed it, and keeping the shareholder's name and address exactly as the register filed them.
2. Cross-border complexity kills
Only 13% of company registers return structured shareholder data through an API, according to Business Registers Insights' 2024 report. The other 87% hand back images, mostly PDFs, which means an automated system must read them with OCR before it can use them. At scale, much of the work stays manual.
I made this point to the Expert Taskforce on Interoperable Beneficial Ownership Data, convened by Open Ownership with the Global Coalition to Fight Financial Crime and LSEG Risk Intelligence. The biggest barriers lie in the fragmentation of registry systems across jurisdictions, from modern REST APIs to PDF-only registries, with no standardised identifiers and highly variable data quality in free-text shareholder declarations. The task force's own conclusion backed it up: interoperability can't make up for weak source data. If the register holds little, no amount of clever processing downstream will conjure the rest.
The situation is even more complicated. Some registers publish no shareholder data. Others publish only a name, which makes entity resolution hard because one name string can point to two or three different legal entities. This is where most systems fail. Faced with two plausible matches, they pick one and move on, and the mistake sits buried three levels down before anyone spots it.
The honest move is to stop. Kyckr's UBO Verify does that: when a name resolves to more than one entity, it halts, flags the doubt, and returns the candidate companies with their addresses so the analyst can choose.
Then there’s the ownership itself/ A UBO must be a natural person, but shareholders are often other companies, many in jurisdictions that publish nothing useful. The BVI publishes no shareholder data. Luxembourg allows nominee shareholders. The Cayman Islands show only the founding shareholders. A good system knows when to hand off to deep investigative tools like Sayari, which draw on third-party data, and when the trail has simply run cold.
3. Trusts are usually dead ends
If an owner wants to hide, a trust lets them – usually. A trustee owns the trust in law, not the person who benefits from it, and trusts need not file their ownership anywhere publicly. In secrecy jurisdictions, they exist precisely to keep the beneficial owner out of sight.
A few jurisdictions run trust registers, but they are rare and hard to use. The UK's Trust Registration Service, held by HMRC, is the obvious case: you must prove a legitimate interest to see it, it holds no structured data, and it can take eight weeks to return a result. Most people who try get nothing back.
A good system is transparent about what it knows
No system can conjure ownership data that the registers don't hold. What a good one does is tell you plainly how far it got: how much of the chain it traced, where the trail stopped, and why. A coverage figure of 60%, stated openly, beats a clean-looking answer built on a guess. Identify what you can, document the rest, hand off to the deep tools when you must, and walk away from a high-risk client when the picture won't come clear.
What to demand from a UBO system
The test is whether the system can show its work. Ask for five things.
The full chain. A name and percentage with no tree is a conclusion you can't audit. Demand every link from the entity to the person.
A source for every link. Each step should say where it came from: which register, which filing, which document. A link you can't source should be labelled as such.
A clear line between calculated and declared. This is the C/D distinction, and it matters: the two carry different evidentiary weight. A system that merges them into one unlabelled field hides the seam.
Honest behaviour when the trail runs cold. Ask what the system does when a name resolves to three entities or hits a trust or an offshore nominee. Does it stop and flag it, or guess and move on?
A coverage figure that is stated plainly. It should tell you how much it traced. 60%, declared openly, beats 100% that crumples in guesses.
A system that can answer these is one you can put in front of a regulator.
