Ref CP-016 · 25 September 2026 · Joshua M. · Reading time 9 minutes

Shell companies, shelf companies, and the difference that matters

Most alerts that mention "shell company" mean nothing of the kind, and most files that should have flagged a shelf company never used the word. This note sets out what a company registry extract actually tells you, which combinations of facts carry weight, and how to write the difference up so a reviewer can follow you.

Two words that get used as if they were one

I have lost count of the number of escalations I have read where the analyst wrote "customer appears to be a shell company" and then stopped, as if the phrase had done some work. It has not. A shell company is a company with no substantial operations of its own. That describes a holding vehicle at the top of a perfectly ordinary group, a special purpose vehicle sitting under a property deal, a dormant subsidiary kept alive to protect a trading name, and a joint venture entity that has not started trading yet. None of those are wrongdoing. Most of the shell companies you will meet in a career exist because a lawyer or an accountant recommended them for reasons that have nothing to do with hiding anything.

A shelf company is a narrower thing. It is a company incorporated by a formation agent with no intention of trading, left sitting on the shelf, and sold later to a buyer who wants a corporate entity with an older incorporation date. That, too, is lawful in most jurisdictions where the practice exists. It is also, in my experience, a far more useful signal than the generic word "shell", because the shelf pattern has a shape you can see on a registry extract and the generic shell does not.

The difference that matters is this. A shell company tells you something about what the entity does. A shelf company tells you something about the history of control. The second is nearly always the more interesting question, because laundering typologies depend on control changing hands quietly, not on the balance sheet being thin.

Reading a registry extract properly

Registry quality varies enormously. Some jurisdictions publish full filing history, officer appointment dates, charges and beneficial ownership declarations free of charge. Others give you a company number, a name and nothing else, and some give you a certified extract that is only as current as the last time somebody paid for it. Know which kind you are holding before you draw conclusions from silence. An absence of filings in a registry that does not require filings is not a finding.

When the data is there, I read it in a fixed order, because reading it in a fixed order stops you anchoring on whichever field happens to look odd first.

Incorporation date against first economic activity

The single most productive comparison on the file. You have an incorporation date from the registry and a first-transaction date from your own systems, plus whatever the customer told you at onboarding about when they started trading. Put the three side by side. A company incorporated in 2019 that opened an account in 2019 and started invoicing in 2020 has a life story that hangs together. A company incorporated in 2012 with no accounts, no charges, no officer changes and no address changes until 2023, which then does all four in the same quarter and starts moving money the following month, has a different story, and the gap is the story.

The gap is not evidence of anything by itself. Companies get dusted off. Families reuse dormant vehicles. But a nine-year dormancy followed by a compressed burst of activity is a question worth asking the customer, and their answer is worth more than your speculation.

The registered office

Formation agents and company service providers house hundreds of entities at a single address, entirely legitimately. So do accountancy practices and serviced office operators. The mere fact of a shared address tells you the customer used a professional, which the vast majority of small companies do.

What carries weight is the combination. A registered office serving several hundred companies, plus a director who appears at that same address across unrelated sectors, plus a trading address that turns out to be a mailbox, plus turnover that would ordinarily need premises — that is four facts pointing the same way, and four facts pointing the same way is a rationale. One fact is a data point. I have had analysts escalate on the address alone and it wastes everybody's afternoon.

Officers and the nominee question

Look at the appointment and resignation dates, not just the current list. A director appointed and resigned inside three weeks around the time control changed is more informative than the person sitting in the chair today. Then look at what else that person directs. Somebody holding twelve directorships across logistics, cosmetics wholesale, construction and a currency exchange, all incorporated within the same eighteen months, is doing something that deserves an explanation — and the explanation may well be that they run a company formation and management business, which is a lawful occupation in most places.

Beneficial ownership declarations are a separate layer and deserve their own reading. The UK's PSC regime and the equivalents built after the FATF standards on transparency of legal persons use ownership and control tests that differ jurisdiction to jurisdiction, and a declaration of "no person has significant control" can be accurate, evasive or simply wrong. The discipline of tracing an ownership structure to the ultimate beneficial owner is a separate skill from reading a single extract, and where the layers cross borders it usually needs both.

The filing history is the actual document

If I can only have one page of a registry record, I take the filing history over the company profile every time. The profile tells you the present state. The history tells you the sequence, and sequence is where intent shows.

Things I read closely in a filing history: dormant or micro-entity accounts filed year after year with no movement, then a gap or a late filing; a change of registered office and a change of director filed within days of each other; a change of company name to something that describes a trade the company has never carried on; the registration or satisfaction of charges, which tells you somebody lent against the entity; and confirmation statements that keep the same shareholder for years and then do not.

Late filings on their own mean the bookkeeper was busy. Clustered changes mean somebody was preparing the vehicle for something. You are not asked to know what. You are asked to notice the cluster, record it, and put it in front of someone with the authority to decide.

What actually raises concern, and in what combination

I keep a short mental list, and I weight the items differently rather than treating them as a scorecard.

• A long dormancy followed by a change of control and immediate high-value flow. This is the shelf pattern proper, and in my experience it is the highest-yield single observation in this whole area, because the entity's age is being used to purchase credibility it has not earned through trading.
• No employees, no premises, no obvious operating cost base, against turnover that would normally require all three. Ask what the cost base looks like; for some agency and brokerage models it really is two people and a laptop.
• Counterparties that do not match the stated trade, particularly where payments arrive from or go to parties outside the commercial chain the customer described — the kind of pattern covered in more depth in the note on third-party payments moving through an account.
• Invoicing that cannot be reconciled to goods, pricing or shipping documents. Where the company is thin and the trade documents are thick, the invoice red flags in trade-based laundering are where the substance usually is, or is not.
• A director who cannot describe the business. Not a registry fact, but the fastest test there is.

The Wolfsberg Group's work on correspondent banking and beneficial ownership makes the same underlying point in a different register: the question is not whether a structure is complex, it is whether the complexity has a purpose the customer can articulate. FATF's standards on the transparency of legal persons exist because opacity is exploitable, not because holding companies are suspicious.

Writing it up without overreaching

Here is where I see good analysis get damaged. Somebody assembles five solid registry observations and then writes a conclusion the observations do not support: "the customer is using a shelf company to launder criminal proceeds." You do not know that. What you know is that the entity was dormant for seven years, that control changed in November, that the first payment was three months later and that the volume exceeded the declared turnover by a factor of four within a quarter. Write that. The arithmetic is strong enough on its own, and it survives challenge in a way that adjectives never do.

Keep the registry extract itself on the file, dated and sourced, along with a note of what you searched and what you could not obtain. A reviewer coming back in eighteen months needs to see the record as it stood when you looked at it, because registries change. The broader question of what belongs in an enhanced due diligence file applies here in full: the corporate history section should read as a chronology, not a list of assertions.

And if it goes to a disclosure, the shelf pattern narrates well because it is inherently sequential. Incorporation, dormancy, change of control, change of name, account opening, first credit, escalation — that is a timeline, and a timeline is the backbone of a usable narrative. The worked example of a SAR narrative shows the shape. What the receiving financial intelligence unit does with it, and what feedback if any comes back, depends entirely on where you have filed; the Egmont Group's members operate under quite different regimes.

One last thing I say to every new analyst. The company registry is a record of what people have declared, not a record of what is true. It is a starting point of unusual quality in some jurisdictions and near-worthless in others. Treat a clean extract as the absence of a contradiction, never as confirmation. Most of the companies you look at will be exactly what they say they are, and the discipline is in being able to say why the few that are not look different.

Takeaway: The age of a company tells you nothing on its own; the gap between when it was incorporated and when it first did anything tells you a great deal.