Actionable insights — Best Buys March 2026

1. Hunt the toll bridge — get paid a small fee on someone else's transaction

1. Look for a business that collects a small, mandatory fee whenever an activity happens, without carrying the risk or the capital of that activity.
2. Confirm the "mandatory" part comes from something durable — regulation, a standard, a network everyone already uses — rather than from a superior product.
3. Check that the fee is trivial to the payer relative to the value received: that is what allows price rises without churn.
4. Then confirm the volume driver is structural, so the toll grows without new decisions being made.

• A regulator or standards body blessing an alternative — the only thing that reliably breaks a toll bridge.

2. Against an AI-disruption narrative, ask who owns the data

1. Separate the model layer from the data layer: models are commoditising, proprietary data is not.
2. Identify data that cannot be recreated — decades of history, private commercial records, exclusive collection rights.
3. Check the company is deploying AI on its own data to raise the value of that data, rather than defending against AI.
4. Invert the conclusion where it holds: the data owner is a beneficiary of the AI build-out, not a victim of it.

• Contractual leakage — data licensed to model developers on terms that let it be learned once and used forever.

3. Write the full bull case, then be willing to pass on price

1. Complete the moat and quality work even for names you expect to reject — the analysis is reusable when the price moves.
2. Anchor the valuation on where the multiple came from, not just how far the price has fallen: a −30% move from >100x earnings is still expensive.
3. Push the earnings base out to a year where the growth is already in the numbers and check the forward multiple there.
4. State the pass explicitly, with the reason, and rank it against the alternatives: "there are more attractively priced Quality businesses elsewhere."

• Your own drift from "cheaper" to "cheap" — the discipline only exists if a named price level triggers the change of mind.

4. Diagnose a growth slowdown: deliberate mix shift or lost competitiveness?

1. When growth decelerates sharply, ask first whether management chose it — walking away from low-margin business is a different fact from losing customers.
2. Test the claim in the margin and mix data: profitability holding or rising while volume slows supports the deliberate reading.
3. Look for the reinvestment the slowdown is funding (a new product line growing off a small base).
4. Weigh founder ownership: an owner-operator can absorb a bad quarter to protect long-run economics in a way a hired manager often cannot.

• Take rate and margin trends over the following two reports — the deliberate story requires profitability to improve as volume growth slows.

5. Buy the collateral damage — a name sold for its exposure to someone else's problem

1. When a sector derates, find the businesses sold off for partial exposure to it and quantify that exposure as a share of assets or earnings.
2. Compare the price move with the exposure. A ~50% decline for a 7% exposure plus a two-cent miss is a mismatch worth investigating.
3. Check the core drivers are intact and, ideally, at record levels.
4. Ask whether the dislocation makes the business better off — a buyer with cash benefits when asset prices fall.

• The other reasons for the decline (private credit, recession fears) actually materialising — the mismatch argument only holds if the extra worries stay hypothetical.

6. Treat open-market insider buying as the confirming signal on a fallen quality name

1. After the quality and valuation work points to a buy, check whether insiders are buying with their own money in the open market.
2. Give weight to clusters — several insiders buying, or buying that continues over months — rather than one token purchase.
3. Read it as confirmation, never as the thesis: it corroborates that the people with the best information disagree with the market's fear.

• Purchases small relative to the insider's net worth, and buying that stops — a change of behaviour is as informative as the behaviour.

7. In a cyclical trough, value the installed base rather than the current earnings

1. Split the business into new-unit sales (cyclical) and aftermarket/consumables (recurring).
2. Note that a demand boom permanently enlarges the aftermarket, even as new-unit sales collapse afterwards.
3. Buy when the market extrapolates trough new-unit earnings across the whole business and the moat is untouched.

• The recurring share of revenue actually rising through the downturn — if it doesn't, the installed-base argument is not working.