Actionable insights — Why We Are Partners

1. Define the universe with two or three business archetypes, before any screening

1. Choose a small number of structural categories that you believe produce durable excess returns, and refuse to look outside them.
2. Support each with evidence, not preference — an academic result, a long-run study, or your own recorded base rate.
3. Label every candidate with its category at the top of the analysis, so a name that fits none is rejected before the numbers start.
4. Accept the cost: whole sectors become uninvestable, and that is the point.

• Cannibal status that is really dilution offset. A shrinking share count alongside a 22%-of-profit share-compensation bill is not the same thing.
• Founder-outperformance studies. The effect is real in the literature but selection-heavy, and the 3.7%/3.9% figures are quoted without a source.

2. Fix the candidate checklist and the mandate constraints in writing

1. Write down the characteristics every holding must have, and treat the list as a gate rather than a scorecard.
2. Write down the portfolio constraints separately: geography, number of positions, turnover, and whether you will time the market.
3. Also write the negative screen — who this strategy is not for — so that expectations are set before the first bad year.
4. Date the document and revise it deliberately, not silently.

• Constraints that drift without an announcement. The 1 September issue revisits exactly this — a stricter bar and a re-affirmed 15–20 stocks — which is how a mandate change should be done.
• "Fair valuation levels" as the eighth characteristic. It is the only subjective one, and it is where every disagreement in the archive actually lives.

3. Apply a two-question test to anyone recommending an investment product

1. Ask: are you invested in this yourself, with meaningful money?
2. Ask: does this product have a long track record of beating a relevant benchmark?
3. Proceed only on two clear yeses; treat a hedged answer to either as a no.
4. Apply the same test to the person publishing the research you follow, including this one.

• The second test being the weaker one for a young strategy. Three years is inside the luck window the same author defines.
• Alignment as a substitute for skill. Someone can be fully invested alongside you and still be wrong.

4. Pre-commit to a base rate of failure so a bad outcome does not become a strategy change

1. Write down, in advance, how often you expect things to go wrong: market drawdowns, individual disappointments, years of underperformance.
2. When one occurs, check it against the pre-committed rate before concluding anything about the process.
3. Only revisit the strategy when the realised rate exceeds the expected one over a meaningful sample.
4. Publish the rate, so that neither you nor your readers can retrospectively pretend to have expected better.

• A base rate used as pre-emptive cover. It explains one loss in three; it does not explain three losses in three.
• Whether the realised rate is ever measured against the stated one. It is not, in this archive.

5. Publish the mistakes by type, not just by name

1. Keep a running list of decisions that went wrong, with the date and the original reasoning.
2. Classify each: was it a buying error, a selling error, or a reasoning error — a good decision with a bad outcome?
3. Look for the repeated type. The category that recurs is the process defect; the individual names are noise.
4. Act on the ones where the conclusion has already been reached, and say so when you are not acting.

• Selling errors going unrecorded elsewhere. They never appear in a portfolio statement, so only a written list catches them.
• A name appearing as both a "steal" and a "mistake" within 48 hours, as Novo does here and on 24 May. Both can be true; the reader needs the reconciliation stated.

6. Never let cheapness be the reason for a purchase

1. Write the one-line reason for every purchase before you place the order.
2. If that sentence is about the price rather than the business, stop — cheapness is a permission, not a thesis.
3. Require an independent statement of why the business will be worth more in ten years.
4. Review purchase rationales periodically as a group, looking for the ones that were price-led.

• Value traps generally: a falling multiple on falling estimates is not getting cheaper.
• The mirror error. Declining wonderful businesses purely on price is also price-led thinking, and the archive does that too.

7. Judge a research product by its capacity, not only its ideas

1. Establish who does the work, how much of it, and whether the output volume is sustainable by that person.
2. Check whether the deep research is original or syndicated from guests, and whether that is disclosed.
3. Ask what happens to the product if the individual stops — key-person risk applies to newsletters as much as to companies.
4. Weigh disclosed effort as evidence of diligence, but not as evidence of accuracy.

• Output volume as a proxy for quality. Two issues a week is a publishing schedule, and schedules generate content whether or not there is something to say.
• The author's own caveat, which is the honest one: "The most important thing at the end of your life is not how much you worked."