Actionable insights — How We're Going Forward

1. Run Buffett's ten-year-closure test as a written exercise, name by name

1. List every holding.
2. Ask one question of each: if the market closed for ten years and you could not sell, would you still want to own this?
3. Write the answer down. Allow "not sure" — it is the answer that carries information.
4. Do it in one sitting so the standard is consistent across names.
5. Compare the result to your existing conviction ranking. Agreement validates both; disagreement tells you which one you actually believe.

• No outright "No" appearing anywhere. Either the portfolio is genuinely clean or the scale is missing its bottom rung.
• The test being run and then not acted on. Only one of the three "Not sure" names is even proposed for sale.

2. Audit your losses for a repeated cause, not for individual errors

1. List the positions that went wrong, including ones you still hold.
2. For each, write down the reason you bought it rather than the reason it fell.
3. Look for a repeated purchase rationale across the list. That is the systematic error; the rest are noise.
4. Convert the finding into a rule that would have blocked all of them.
5. Re-read it before every subsequent purchase.

• The opposite error the new rule creates. "Quality at a fair price" is how portfolios end up paying any price for quality — the published target prices are the guard against that, and they only work if they bind.
• Survivorship in the audit: only failed purchases get examined. The cheap-and-good ones that worked are not listed.

3. Filter on the shape of growth, not just its rate

1. For each candidate, look at ten years of revenue and earnings growth as a series, not as a CAGR.
2. Score the variability: how many down years, how wide the swings, whether the trend is smooth.
3. Prefer the smoother of two businesses growing at the same average rate — it is worth more because the future is more forecastable.
4. Apply the same test to the business model, not only the accounts: recurring revenue, contracted volumes, non-discretionary demand.
5. Be explicit when you make an exception, and say what you are accepting in return.

• Smoothness that comes from unquoted marks rather than from the business — the private-asset argument made for KKR and III.L in this same issue cuts both ways.
• Your own exceptions. FFH.TO is admitted as a buy candidate with "earnings can be lumpy quarter to quarter" written in its own paragraph.

4. Name in advance the three fundamentals you will track instead of the price

1. Choose two or three portfolio-level fundamental series and commit to them: look-through free cash flow, an aggregated quality scorecard, owner's earnings.
2. Rebase each to 100 at a fixed start date and update annually, so the chart is comparable over time.
3. Report them whether or not the share prices have followed.
4. Use price only at the two moments it matters — buying and selling.
5. If the fundamentals stall, the thesis is broken; if only the price stalls, it is an opportunity or a style drawdown.

• The gap persisting. Owner's earnings at 19.7% a year against a 3-year share-price CAGR of 5.6% is either a coiled spring or evidence that the market disputes the fundamentals.
• Metrics chosen because they flatter. Owner's earnings is a defensible measure, but it is also the one that looks best here.

5. State the size of your investable universe, and read your own buy count against it

1. Count the names that have passed your quality screen and are eligible to be bought.
2. Express it as a share of the global listed universe, so the filter's strength is visible.
3. Then track how many of those eligible names are currently rated a buy.
4. A high proportion is a statement about valuations across the whole quality cohort, not about your stock picking.

• The universe quietly expanding to accommodate ideas. 153 is only a meaningful filter if the count is stable and the entry criteria are fixed.
• A record-high buy count being reported as opportunity when it may be a factor drawdown in progress.

6. Distinguish "not marked daily" as a benefit to the manager's attention from a benefit to the asset

1. When a business is praised for owning unquoted assets, separate two claims: the underlying assets are better, or the absence of a quote improves behaviour.
2. Only the second is usually true, and it is a claim about psychology, not value.
3. Ask who does the marking and how often, and what an independent transaction would show.
4. Treat a smooth NAV as unverified, not as low risk.

• Volatility that has been relocated rather than removed. It reappears at the exit, all at once.
• Using the same feature as a virtue in one paragraph and a misunderstood risk in another.

7. Commit to announcing a switch before you execute it, and then check whether you did

1. State the rule: any new position that requires funding will be flagged, with the sale, before it happens.
2. Keep the shortlist short enough that the reader can anticipate the pair.
3. When the transaction comes, publish the size, the limit and the reasoning.
4. Afterwards, compare the executed trade with the shortlist. If they do not match, say why.

• Adding to existing positions as a way of avoiding the switch decision. It is defensible, but it is not the decision that was announced.
• A shortlist that quietly expires. Of the five, only Fairfax was bought within four months.

8. Separate a ranking from a decision, and date both

1. When you name a "most likely sell candidate", record the date and the conditions that would trigger the sale.
2. Revisit on a schedule. A verdict without a trigger has no expiry and will drift.
3. If the position is still held three months later, either the conditions were never met or the verdict was soft — decide which.
4. Publish the rating in the interim so the drift is visible rather than silent.

• Re-underwriting that never concludes: "we are currently re-investigating" has now appeared twice, nine days apart, with identical text.
• The opportunity-cost framing being used to defer rather than decide — the sale only happens if something better is bought, and the buy list was not executed either.