Process
What the belief looks like on an ordinary Tuesday.
Philosophy is cheap. Process is the part that survives contact with a bad quarter, a crowded trade, and a position that has gone against you for eleven months.
Everything below is repeatable by design. The point of a documented process is not bureaucracy — it's that it keeps working on the days when judgment is compromised by stress, by a drawdown, or by having been right too recently.
The stages are sequential, but the loop runs constantly. Most of my time is spent on stages three and six.
-
STAGE 01
Define the fishing ground
Before screening anything, I decide where I'm willing to own a business at all. That means explicit constraints on liquidity, disclosure quality, governance, and capital structure — applied at the universe level rather than argued case by case when I've already grown fond of a name.
In Canada this step does more work than people expect. It removes a large tail of companies where the analysis would be sound but the position could never be exited at a sensible price.
- Minimum tradeable liquidity, expressed as days-to-exit at a realistic share of volume
- Governance and control-structure screens — dual-class, related-party exposure, board independence
- Accounting red flags applied mechanically, before narrative gets a vote
-
STAGE 02
Generate ideas from more than one direction
A single idea source produces a portfolio of correlated mistakes. I run several deliberately uncorrelated funnels: systematic screens for statistical candidates, a watchlist of quality businesses waiting for a price, and situational triggers — spin-offs, forced selling, index events, management change, post-guidance-cut wreckage.
The screens are described in more detail under Models. Their job is not to pick stocks. It is to reduce a market to a reading list.
-
STAGE 03
Do the primary work
This is where the time goes, and it is the only stage that generates genuine informational advantage. Filings and the full history of management commentary rather than the last two quarters. Unit economics rebuilt from the disclosure rather than lifted from a model. Competitors, suppliers, and customers read alongside the company, because in a market this concentrated the useful signal is frequently in somebody else's disclosure.
The deliverable is a written thesis with three things stated before any capital is committed: what has to be true, what would prove me wrong, and what I think the market currently believes. If I can't articulate the third, I don't yet understand why the opportunity exists.
-
STAGE 04
Value it across scenarios, not to a point
A single target price implies a precision that doesn't exist. I work in scenarios with explicit probabilities — a downside anchored in what the business is worth if the thesis simply fails, a base case, and an upside that doesn't require heroic assumptions to reach.
The number that drives the decision is not the expected value. It's the relationship between the downside and the current price, because that is what determines how large the position can responsibly be.
-
STAGE 05
Size it, then check what it does to the portfolio
Sizing follows from the downside case and from how quickly the position could be unwound — not from enthusiasm. Then the more important question: what does adding this do to everything else I own?
In Canadian equities this check is non-negotiable. Six independently researched ideas can turn out to be one oil price and one credit cycle wearing six different tickers. I look at factor and macro exposure at the portfolio level explicitly, because the correlations are not obvious from the company descriptions.
-
STAGE 06
Monitor the thesis, not the price
Each position carries the specific operating milestones that the thesis depends on. Reviews are scheduled against those milestones, so that the decision to revisit isn't triggered by the stock having moved — which is exactly when judgment is worst.
There are three reasons to sell, and I try to be honest about which one applies: the thesis played out, the thesis broke, or I found something meaningfully better. "It's gone down and I feel bad" is not on the list, and neither is "it's gone up and I'd like to book the win."
The stage most people skip.
Every closed position gets written up against the thesis that was filed when it was opened. Not to assign blame — to separate the two questions that matter: was the reasoning sound, and was the outcome a consequence of that reasoning or of something I never modelled?
Done consistently, this produces the only genuinely proprietary dataset a manager has: a long record of their own errors, classified by type. Mine cluster in identifiable places — [the two or three recurring error types you'd be willing to name]. Knowing that has changed how I size certain kinds of idea.
This is also where the quantitative work started. Several of the models exist because a post-mortem kept surfacing the same error, and I wanted to know whether it was a pattern or a story I was telling myself.