Actionable insights — Portfolio Update 2026 (Part I)

1. Re-underwrite every holding with an identical four-line block

1. For each position record: forward PE against its own five-year average; the Earnings Growth Model return; the reverse-DCF growth the price requires; and the long-term expected EPS growth to compare it with.
2. Mark each line pass, fail or questionable, with the threshold printed beside it.
3. Add the ten-year cash-flow CAGR as the reality check on whether the required growth has ever been achieved.
4. Finish with a single-word verdict and one sentence of justification — then move on. Uniformity is what makes twenty positions comparable.

• A company's own five-year average as the benchmark. It bakes in whatever the last five years' sentiment was — useless for a name that was systematically overpriced.
• Positions reviewed in an order that lets the small ones be skimmed. Sorting by weight ascending, as here, puts the least-examined names first.

2. Let the verdict override the checklist — in both directions, and visibly

1. Run the models first and publish their output unedited.
2. Then state the decision, and where it disagrees with the models, say which is binding and why.
3. Require the override to name a business fact — a structural growth problem, a strategic asset — not a feeling.
4. Track the overrides separately over time; they are where a process either adds judgment or leaks discipline.

• Overrides that only ever go one way. Upgrading past a failed test and never downgrading past a passed one is not judgment.
• A metric bar carried over from a different business shape. Brookfield's <12x P/DE bar predates the asset-management mix it now has.

3. Distinguish a cheap stock from a structurally broken one — the test is growth, not price

1. When a holding is cheap on every measure and still uncomfortable, ask specifically whether the growth has stopped, not whether the price is low.
2. Decide whether the stall is cyclical (a funding cycle, a soft insurance market) or structural (commoditisation, permanent competitive loss).
3. If structural, no multiple is low enough — the correct question becomes opportunity cost, not valuation.
4. Say so in the review before you act, so the decision is on the record ahead of the trade.

• The reverse DCF passing on a business whose earnings are about to stop growing. It solves for growth; it cannot tell you the growth is over.
• Comparing a stalled name against your own cost rather than against the best alternative available today.

4. Concede a disruption premise, then locate the barrier somewhere else

1. Grant the strongest version of the threat outright — arguing against the technology loses credibility and usually the argument.
2. Decompose the value chain and ask which step the technology actually removes.
3. Name the remaining steps and test each for a barrier: integration depth, absence of alternatives, relationship, regulation.
4. Check the same reasoning against a counter-example you do not own, to be sure it is a test and not a defence.

• The argument being applied only to holdings. It is a general test or it is a rationalisation.
• Distribution moats in markets where the buyer is a procurement process rather than a relationship.

5. Build a full position into a decline, and record when you did it

1. Where a great business is too expensive, open a partial position and state that it is partial and why.
2. Define in advance the fall that would complete it — a percentage from the peak, or a multiple.
3. Execute when it triggers, and publish the date.
4. Expect the position to show a loss afterwards; that is the arithmetic of averaging down, not evidence against the decision.

• Averaging down without a pre-set trigger. "It fell further so I bought more" is a different activity.
• A completed position with nothing left to add if it halves again. Decide the maximum weight before the first tranche.

6. Underwrite a founder succession on pattern match, not on reassurance

1. When a founder-run compounder changes chief executive, treat it as a new investment decision rather than an event.
2. Check the successor's background against the specific model the business runs, not against general seniority.
3. Look for an overlap period long enough to transfer judgment, and for the founder staying in a governance role.
4. Decide in advance what evidence in the first two years would confirm or break it — deal cadence and price discipline, not reported earnings.

• A succession happening while the end market is also weak, as here — two variables changing at once makes the first two years unreadable.
• Credentials from a similar-sounding model. Halma and Renishaw are the right comparison; a large-cap operator would not be.

7. Disclose the portfolio's construction buckets and the drift between them

1. Define a small number of business-type buckets that reflect why you own things, not sectors.
2. Publish the weight in each, and the change over the year.
3. Say whether the drift was deliberate or a by-product of performance — they demand different responses.
4. Use the buckets as a construction constraint, so a single style cannot quietly become the whole book.

• Nearly three-quarters in one bucket described as diversification. It is a concentrated style bet, made explicitly.
• Buckets that overlap. Several holdings here are plausibly two of the three at once, so the percentages depend on the classifier.