Actionable insights — Portfolio Update April 2026 (Part I)

1. Rank your own holdings into conviction tiers, and write the reason for the tier

1. Periodically stop screening for new ideas and instead rank what you already own, from most to least sure. Four bands is enough: favourites · no doubts · proud owner but real risks · should this be sold.
2. For every name, answer two questions in writing — how does it make money, and why is the conviction at this tier rather than one higher.
3. Force the second answer to name a single specific reason. A tier demotion with no nameable cause means the ranking is a feeling, not an analysis.
4. Treat the bottom tier as an explicit sell queue, reviewed on a schedule rather than when the price moves.

• Tier inflation — if nothing ever sits in the bottom band, the ranking has stopped doing work.
• A demotion reason that quietly becomes a thesis break; the tier should move to a sale, not stay parked, once the cause is confirmed.

2. For a serial acquirer, value the cash available to owners — not reported earnings

1. Identify the accounting distortion first: an acquisitive company must amortise the purchase price of what it buys, which suppresses reported profit without any cash leaving the business.
2. Rebuild the owner's number instead. For a software roll-up that is Free Cash Flow Available To Shareholders (FCFA2S) — free cash flow after the claims of minorities and debt-holders.
3. Apply an explicit, stated growth assumption for the coming year rather than a multi-year model, and say what it is so a reader can disagree with one input.
4. Express the result as a forward yield on that cash and compare it with the company's own history, not with a sector average.

• The growth assumption doing all the work — at +0% instead of +15%, the yield and the "cheapest ever" claim both change materially.
• Whether the acquisition pipeline still absorbs the cash at good returns; the whole model depends on redeployment, not on the yield itself.

3. Use NPATA (or any owner-earnings adjustment) where amortisation is not a real cost — and build the multi-year ladder

1. Where a business acquires intangible assets — client lists, customer relationships, books of business — add back the amortisation of those intangibles to get the cash profit an owner actually receives. Kelly Partners' name for it is NPATA; Buffett's is Owner Earnings.
2. Apply the current price to that figure for a forward multiple.
3. Then lay management's own multi-year targets against today's price to produce a ladder of multiples, so the question becomes "do I believe the plan?" instead of "is the multiple high?"
4. Keep the targets as management's, explicitly attributed — the ladder is a test of credibility, not a forecast you have made.

• Add-backs that are real costs — if the acquired client relationships genuinely run off and must be replaced with new acquisitions, the amortisation is a maintenance expense in disguise.
• The first missed rung on the ladder; a serial acquirer's forward targets assume deals that have not yet been done.

4. Treat a founder's pledged shares as a governance red flag — and re-underwrite the person, not the price

1. Check whether insiders have pledged shares as collateral for personal borrowing. Skin in the game financed with debt is not skin in the game.
2. Model the reflexive loop: a falling price triggers a margin call, forced selling pushes the price lower, which triggers more calls. The founder becomes a forced seller precisely when the company most needs a stable register.
3. Separate the two questions the event raises — has the business deteriorated (usually no), and has your read on the operator's judgement changed (possibly yes).
4. Respond by contacting management and setting a date to conclude, rather than by trading the headline in either direction.
5. Cap the position's conviction tier until the question is closed, and say out loud what would resolve it.

• Whether remaining pledges are disclosed and unwound — one margin call rarely exhausts the borrowing.
• The distinction between an owner-operator and a promoter, which usually shows up in guidance quality rather than in one event.

5. Judge a serial acquirer on deal cadence, not just deal price

1. Establish the acquirer's normal deal size relative to its own market value; that cadence is the model you underwrote.
2. When a transaction is an order of magnitude larger, treat it as a change in the risk profile even if the price looks defensible — many small bets average out, one large bet does not.
3. Do the dilution arithmetic explicitly: capital raised, resulting share-count increase, and the incremental profit at the acquirer's own margins, so you can see whether the deal pays for its own dilution.
4. Then decide whether it changes the conviction tier rather than the ownership.

• Integration evidence over the following two to four quarters — whether the acquirer's margins actually get applied to the acquired revenue.
• Organic growth being masked by the acquisition once the two are combined in the reported line.

6. Buy a listed vehicle at a discount to a published asset value — and check the assets first

1. For a trust, holding company or asset manager that publishes a net asset value, analyse the underlying holdings before the wrapper: growth, margins, and whether the assets are the kind that hold their marks.
2. Compare the share price with the published NAV per share and express the gap as a discount.
3. Compare that discount with the vehicle's own history — a wide discount is only an opportunity relative to where it normally trades.
4. Recognise the two ways you get paid: NAV compounding, and the discount narrowing.

• How the NAV is struck — private holdings are marked by their owner, so the discount is only as real as the marks.
• A discount that never closes; without buybacks, realisations or a catalyst, you are relying entirely on the assets compounding.

7. Value the same company two independent ways and require both to agree

1. Pick two measures that fail differently — an asset-based one (NAV, sum of the parts) and an earnings-based one (a multiple of the cash profit).
2. Run both at today's price and state each result as a single number.
3. Act only when both say the same thing; a disagreement is information about which measure is wrong, and the work is not finished.
4. For the earnings leg, show today's multiple and the multiple on management's out-year guidance, so the implied bet is visible.

• Both measures sharing an input — if the NAV and the earnings forecast both come from the same management model, they are not independent.
• The 2030 figure being a plan, not a fact; a 25% compound growth rate is a strong assumption to embed in an entry price.

8. Build an expected return from disclosed inputs instead of a target multiple

1. Start with the earnings yield (earnings ÷ price) rather than the P/E, so the number is already in units of return.
2. Multiply it by the share of profits actually returned to shareholders in dividends and buybacks to get the cash-return leg.
3. Add the organic revenue growth rate and the earnings growth rate the company has been delivering.
4. Sum to a range, and present the inputs so any single one can be challenged without discarding the framework.
5. Note what is deliberately excluded: no multiple re-rating. A return that needs the multiple to expand is a different, weaker claim.

• Buybacks executed above intrinsic value — the cash-return leg only counts fully if the shares retired were cheap.
• A fee-based earnings yield that is a cyclical peak; the whole sum rests on the E in the yield being durable.

9. Test pricing power with an actual annual number, not an adjective

1. Ask what the company raised prices by last year, and the year before, and whether volume grew anyway.
2. A moat claim that cannot produce that pair of numbers is a story; one that can is measurable and repeatable.
3. Then look for unpriced optionality attached to the same asset — licensing, adaptation rights, a partner distributing the IP — and treat it as upside rather than as part of the case.

• Price rises that start buying volume declines — the first year the pair breaks is the signal, not the price rise itself.

10. Anchor a demand thesis to a social trend rather than a product cycle

1. Ask what slow-moving change in how people live drives the spending, and whether it is reversible within your holding period.
2. Prefer spending driven by emotion or obligation over discretionary spending — it survives a recession better.
3. Then check the price separately; a good trend at a high multiple is still a bad entry.

• The trend being real but already fully priced, which is the usual failure mode of demographic theses.

11. Notice when the whole book goes cheap at once — and ask why

1. Track each holding's valuation against its own ten-year history, not against the market.
2. When most of a portfolio simultaneously reaches decade-low multiples on unchanged fundamentals, the cause is almost certainly the discount rate or sentiment rather than the businesses.
3. Separate the names where the fundamentals also weakened — those are not in the same category and should not be bought on the same reasoning.
4. Use the observation to decide sizing (adding to the strongest convictions) rather than to add new names.

• The possibility that the market is right about a common factor you own repeatedly — here, quality compounders re-rating together on AI and rate fears.
• Concentration creeping in through additions rather than through a decision.