How we think about a business, and about being wrong
Our research process is deliberately slow and deliberately adversarial. Nothing enters a portfolio because it is interesting. It enters because it survived four attempts to kill it.
- I
Origination
Weeks 1–4
Ideas arrive from screens, from operators, and most usefully from the failures of businesses we already own. We read filings before we read research. We speak to customers before we speak to management.
Primary sources logged
- II
Interrogation
Weeks 4–12
A second analyst is assigned to argue the short case in writing, with a budget to travel and a mandate to embarrass the sponsor. If the short memo is more persuasive than the long memo, the idea dies here. Most do.
Adversarial memo required
- III
Sizing
Committee
Conviction and position size are separate decisions. Size is a function of what the position does to the portfolio when the thesis fails, not what it does when the thesis works. We model the failure first.
Downside modelled before upside
- IV
Custody
Years
Every holding is re-underwritten annually as though we did not already own it. Ownership is not a reason to keep owning. The re-underwriting memo is written by whoever is most sceptical, never by the original sponsor.
Annual re-underwriting
Volatility is the price of admission. Impairment is the bill you cannot appeal.
We measure two things that most managers report as one
The industry reports risk as standard deviation because standard deviation is easy to compute and comfortable to explain. It describes how much a price moved. It says nothing about whether the business underneath it is still worth owning.
We separate the two explicitly in every quarterly letter. Drawdown is reported because you will feel it. Impairment — capital we do not expect to recover — is reported because it is the only number that has ever permanently changed a client's outcome.
Cash held against opportunity
The process is the reason people stay.
If the way we work resembles the way you already think about capital, the next step is a conversation rather than a document.
Request access