Actionable insights — Time to Raise the Bar (Part I)

1. Rank every position on the same two axes, and read the four corners

1. Build one table with every holding, a forward multiple, and an expected multi-year growth rate — the same source and definition for all of them.
2. Add the portfolio average and the index on the same two axes, so the aggregate claim is testable.
3. Name the extremes out loud: most expensive, most attractive, fastest growth, slowest growth.
4. Read the corners. Expensive-and-slow is where the next mistake lives; cheap-and-fast is where the next addition should come from.
5. Do it on a fixed cadence, not when something feels wrong.

• The corner that goes unmentioned: Games Workshop is the most expensive and among the slower growers, and it is not on any action list.
• Mixing definitions — this table uses analyst estimates while the firm's own models use its own, and the letter flags the resulting 15%-versus-20% gap on Brookfield.

2. Plot current weights against conviction and treat every mismatch as a decision you have not made

1. Chart the book by weight, descending.
2. Beside it, list the positions in order of how much you would want to own them today, ignoring what you paid and when.
3. Any large divergence is a decision by default. Name it and resolve it in one direction or the other.
4. Diagnose why the divergence exists — drift from performance, a recent purchase not yet built up, or a conviction that has quietly changed.
5. Fix it by adding to the under-weighted rather than only by cutting, so the correction does not become market timing.

• A "skew" that is really a purchase schedule. Two of the three underweights were bought within the last six weeks — that is not a conviction error.
• Rebalancing only among the small positions: the six names involved are all between 2.6% and 5.0%, so the top of the book is untouched.

3. Split an expected return into the part the business earns and the part a re-rating supplies

1. Start from owner's earnings: expected EPS growth plus dividend yield. That is what the businesses deliver.
2. Add the multiple effect separately, stating the current level, the level you assume is fair, and the number of years over which they converge.
3. Report both components, not just the sum, so a reader can discount the part you assumed.
4. Cross-check with an independent construction that does not need a re-rating.
5. Judge the outcome against the first component alone, because the second is not in your control.

• "Cheapest valuation level ever" as the basis for assuming reversion — the same sentence has appeared in this archive since January while the multiple kept falling.
• A "fair" yield chosen by the person who benefits from the answer, with the words "I'm probably even being conservative here" doing the work of an argument.

4. Score the portfolio in look-through cash flow, per position and per share

1. For each holding, multiply your share count by the free cash flow per share. That is the cash the business generates on your behalf.
2. Total it, and restate it in per-month, per-week and per-minute terms so it is felt rather than read.
3. Track the series over years and separate the growth that came from the businesses from the growth that came from your own contributions.
4. Rank positions by cash contribution as well as by market value — the two orders will differ, and the difference is informative.
5. Set the goal in cash terms, and state the assumptions the projection needs.

• A rising cash figure driven mostly by contributions. At $50,000 a month, new money dwarfs the compounding in the early years, and the projection does not separate them.
• Cash per position as a hidden concentration measure: cash contribution and weight diverge sharply here, and the cash view is the one that matters for the stated goal.

5. Write down the pattern of your own mistakes, then change the rule rather than the holdings

1. Review several years of decisions and look for the recurring shape of the errors, not the individual names.
2. State it in one sentence, plainly enough that it constrains future behaviour.
3. Translate it into a rule change — a tighter screen, a smaller maximum count, a narrower geography.
4. Publish the new rules as a list so they can be checked against later decisions.
5. Then apply them position by position, and accept that some current holdings will fail.

• The tension with the firm's own flagship format: Best Buys works by shopping the worst performers of a pre-vetted universe, which is a cheapness-first habit. Subordinating cheapness to quality changes what that list is for.
• Rules announced and not yet applied. Until the count actually falls to 15-20, this is an intention.

6. Put your realised return next to your operating statistics, and sit with the gap

1. Tabulate the portfolio against its benchmark on balance sheet, capital intensity, returns on capital, margins, cash conversion, historical growth and forward growth.
2. In the same table, put the realised total return over three and five years.
3. If the operating measures win and the return loses, name the reason: multiple compression, timing of purchases, or a thesis that is simply wrong.
4. Decide what evidence would make you conclude the third, and by when.

• "CAGR since IPO 18.8% vs 9-11%" placed in the same block — a differently measured, differently dated number that softens the two that matter.
• A PEG ratio that is worse than the index (1.8x vs 1.5x) sitting inside a valuation block presented as favourable.

7. Apply a new rule to every position, including the ones you like

1. Write the new rule, then run the entire book through it mechanically before deciding anything.
2. List every holding that fails, without exception, and give a reason for each one you nonetheless keep.
3. Check the rule against your own tables — a rule contradicted by a printed number in the same document is not yet a rule.
4. Publish the exceptions. A framework's credibility lives in the names it keeps despite itself, not in the ones it cuts.

• Reductions justified on growth when a cleaner rule (geography, or the valuation bar) would have reached the same conclusion — using the weaker reason leaves the rule untested.
• Part II as the place where this gets resolved, or does not: the count has to fall from 21 to 15-20, so at least one name must actually go.