The cohort · the edge of the record

The filing is the smaller half.

Everything on the two pages before this one was measured from public filings. Here is the honest boundary of that record, stated before anyone has to ask: it covers listed shares held long, and nothing else a family owns.

From the cohort, read when this page was built

One sleeve

of the balance sheet is reportable. The other sleeves — private funds, property, operating companies, cash, debt, tax — are why 40 reviewed offices can be read this closely and still not be known.

The finding

What does a public filing leave out?

Not in the record

Private funds and direct deals

Commitments, calls, distributions and the marks everything else is built on. None of it appears.

Not in the record

Property and operating businesses

Often the largest thing the family owns, and the reason the listed sleeve exists at all.

Not in the record

Cash and debt

The reserve, the credit lines, the margin, the covenants. A filing shows assets, never the claims on them.

Not in the record

Who owns it

Trusts, holding companies, partnerships, and which heir ends up with what. Filings report an entity, not a family.

Not in the record

Fees and tax

What was paid to hold the position, and what is owed on the gain already sitting inside it.

Not in the record

Everything between two dates

A position opened and closed inside a quarter was never held, as far as the record is concerned.

The second boundary

Even who counts as a family office is a judgement.

There is no public list of family offices. A single-family office is deliberately absent from the adviser rosters, so the first signal is an absence — which on its own is wrong often enough to be dangerous. Of 51 filers the screen proposed, a person accepted 40, rejected 7 as something else, and left 4 unresolved rather than guess.

We publish the count of the ones we could not settle. A cohort with no unresolved cases has either perfect data or a low standard of proof.

No performance figures here

Why there is not a single performance figure on these pages.

Turning holdings into returns needs licensed prices, and the result would still describe one sleeve of one entity, blind to everything traded between snapshots. So nothing in this cohort is a return, a ranking or a backtest. It is structure: how many names, in what weights, held for how long, and how much of that is shared with everyone else.

Structure is the part that transfers to your book anyway. A return someone else earned on a book you cannot see was never going to tell you anything you could act on.

For your own book

The only complete book is the one you already have.

Every gap listed above exists because an outsider is looking in. Inside a family office none of them is a mystery: the commitments, the credit lines, the trusts, the tax already sitting inside the gains. They are simply spread across custodians, entities, statements and inboxes, which is a different problem and a solvable one.

The questions worth asking are the ones a filing could never answer. That is the whole of what we build.

The seven questions no dashboard answers →

The limits

Stated plainly, once more.

  • Only listed shares held long are reported. Private funds, direct deals, property, cash, debt, fees and tax are all outside the filing.
  • A filing is a photograph of one day, published up to 45 days later. It cannot show what was bought and sold in between.
  • Only offices above the reporting threshold appear at all, so the cohort skews larger than the category.
  • No prices are licensed here, so nothing in this cohort is a return, a ranking, or a backtest. Holdings facts only.
Where the numbers came from

How the cohort was assembled.

A screen proposed 51 filers as candidate family offices. A person read every one: 40 were accepted, 7 were rejected as something else, and 4 were left open because the evidence did not settle it. Only the accepted ones are in any figure on this page.

That is 874 filings and 61,471 reported positions across 22 quarters, each one read as of the day it became public rather than with the benefit of hindsight. A filing that arrived late is missing from its quarter instead of being back-filled into it.

The cohort is other firms' public quarterly filings, read as they stood on the day each one became public. It is not our clients, not their data, and not a performance record. Counts as at 19 September 2026.

Public filings only, and only the part of a book they cover: listed shares held long. No private holdings, no cash, no debt, no fees, no tax. Offices that stopped filing stay in every figure. Nobody here is named. Figures above are computed from the filings themselves, as they stood on 2026-09-20. A record and an analysis, not investment, tax or legal advice; decisions rest with the family and its advisers.

The whole table, every measure →