Actionable insights — Portfolio Update: April 2026 (Part II)

1. Give every holding a conviction tier, and make the bottom tier a question

1. Assign each position to a small number of named tiers — four is enough — and write a one-line definition of each that you would be willing to publish.
2. Phrase the bottom tier as a question about selling, not as a weaker endorsement. "Medium conviction" invites drift; "should we consider selling these?" forces a decision.
3. Re-rank on a fixed cadence, and publish the ranking so the previous one can be compared against it.
4. Review the tiers from the bottom up. The top of a portfolio needs no defending; the bottom is where the returns leak away.

• Tiers that never change. A conviction scale is only doing work if names move down it as well as up.
• The slide and the text disagreeing: here 17 names are shown against a stated 18 holdings, with Visa missing. Check the published artefact against the position list.

2. Test a retailer's like-for-like growth against its own input inflation, not against zero

1. Take the like-for-like (same-store) sales growth rate the company reports.
2. Find the relevant price inflation for what it actually sells — food inflation for a grocer, not headline CPI.
3. Subtract. What remains is real volume growth in the existing estate.
4. If the answer is negative, the reported growth is coming from prices alone, and every future year depends on new stores rather than on the base business.
5. Separate that from the store-opening pipeline, which is a different question with different evidence (capex commitments, land, distribution capacity).

• Margin compression arriving at the same time — cost inflation passing through faster than selling-price inflation is the second half of the same problem.
• Companies that stop disclosing LFL separately once it turns negative in real terms.

3. Read discretionary capex as management's own forecast, and check how it is funded

1. Look for capital commitments that only make sense under a specific growth assumption — distribution capacity, land banks, plant.
2. Size them and date them: how much, completing when, supporting how many units.
3. Check the funding source. Cash from operations is a forecast management is willing to fund itself; debt is a forecast someone else is underwriting.
4. Treat the commitment as harder evidence than the guidance, because it is expensive to reverse.

• Capacity built for a growth rate the recent like-for-like numbers no longer support — the two signals in this write-up point opposite ways, and both are true.
• Completion dates: capacity that lands in 2027 says nothing about 2026 volumes.

4. Distinguish a cheap stock from a governance problem — the cure for each is different

1. When a holding is unusually cheap, ask whether the discount is about the business or about who controls it.
2. Score the disclosure separately from the numbers: does management explain capital-allocation changes, or announce them and go quiet?
3. Enumerate the possible explanations for an unexplained action, and note how many of them are good for a minority holder.
4. If a controlling shareholder is accumulating, model the exit you are actually offered — a takeover premium off a depressed price, not a re-rating to fair value.
5. Decide in advance what would resolve it: a buyback programme, a stated policy, an independent board response.

• Buybacks as the tell. "In an ideal world, the company would start to heavily buy back shares" — if the cash is genuinely for shareholders, that is where it shows up.
• Third-party theses (here a widely-quoted tweet) doing the analytical work. Useful, but source it and check the incentives.

5. Put a struggling holding on an explicit opportunity-cost test rather than defending it

1. Separate the diagnosis ("what went wrong") from the decision ("is this still the best use of the capital").
2. State the unresolved question in one line — usually "temporary or structural?" — and admit when you cannot answer it.
3. Rather than forcing an answer, declare the position available: it will be sold if something better is found, not because the thesis is disproved.
4. Then go and look. The test is only honest if it is paired with an actual search for the alternative.

• "Under review" as a permanent state. The sequence here is dated and finite, which is what makes it a process rather than a hedge.
• Selling the wrong one: opportunity cost tests the whole book, so the position sold should be the worst one, not the most painful one.

6. Decide with an ownership question, not a valuation, when the uncertainty is about the business

1. Do the valuation work anyway — two independent methods if you can.
2. Then ask the ownership question separately: would I want to own this for the next ten years, knowing what I know?
3. If the answer is no, treat the valuation as irrelevant. A cheap price is not compensation for a business you do not want.
4. Write down what would change the answer, so the position is reviewable rather than abandoned.

• Sum-of-parts arguments that value the troubled segment at zero. They are usually right that the market is doing so, and usually silent on why.
• Quoting a fellow holder's mea culpa (Smith's "triumph to a tragedy") — powerful, but it is an argument about the past, not a forecast.

7. Publish the portfolio-versus-index scorecard including the columns you lose

1. Build one table comparing your holdings, aggregated, to the index on balance sheet, capital intensity, returns on capital, profitability, growth and valuation.
2. Include realised performance in the same table, over several horizons.
3. Do not drop the horizons you lose on. Those are the ones a reader would otherwise have to reconstruct, and they are the reason the review is happening.
4. Read the combination: better businesses at a lower multiple that have underperformed is a factor drawdown, not a stock-picking failure — but only if the fundamentals in the table are real.

• Aggregated ratios hiding a single outlier — a 265% FCF/net income for a portfolio is an average worth decomposing.
• The "cheapest valuation ever" refrain appearing on most holdings at once. That is a factor statement about quality as a style, not six independent observations.