Actionable insights — Best Buys: June 2026

1. Screen for a toll bridge: a fee that is mandatory, recurring and paid by someone else's activity

1. Look for a business paid per transaction on activity it does not fund and does not control.
2. Establish what makes payment unavoidable — regulation, market convention, or a system built around it.
3. Count the tolls. A business with two independent tolls on the same system is far more robust than one.
4. Check the fee is small relative to the transaction, so it is not worth fighting.
5. Confirm the revenue is recurring rather than cyclical with issuance volume.

• The toll being a regulatory permission that a regulator can withdraw. That is exactly what happened to FICO when the FHFA approved a rival.
• Cyclicality hiding in the ratings half. Bond issuance falls when rates rise, even though the oligopoly holds.

2. In asset management, buy the fee stream and check where the capital is permanent

1. Separate management fees (charged on assets, contractual) from performance fees (charged on results, volatile). Value the first far more highly.
2. Establish total assets under management and the annual gross inflow, since the fee base only compounds if new money arrives.
3. Identify how much of the capital is permanent — insurance balance sheets and evergreen vehicles cannot be redeemed at the wrong moment.
4. Look for a new distribution channel that expands the addressable pool, and check whether it is actually growing.

• Fee rates falling as the asset base grows. Assets can rise while fee income does not.
• Retail distribution into illiquid assets. K-Series-style vehicles have never been tested through a redemption cycle.

3. When a regulator ends a monopoly, test whether the product becomes additive or replaced

1. Identify the exact regulatory change and what it permits — approval to compete is not the same as a mandate to switch.
2. Ask whether a customer would run both products or replace one with the other. Additive outcomes barely dent revenue.
3. Estimate the switching cost inside the customer's own systems, in years rather than dollars.
4. Check what share of revenue is exposed at all, and how much sits in adjacent recurring software.
5. Then decide whether the price fall over-discounts the change.

• The same evidence supporting opposite conclusions. The 21 May deep dive used this identical risk to refuse the stock for the portfolio.
• Objections that quietly disappear. FICO's 22%-of-profit stock compensation, decisive three weeks earlier, is not mentioned in this write-up at all.

4. Before accepting an AI-disruption selloff, ask whether the company's data is an input to AI

1. Establish whether the asset is proprietary and licensed, or public and scrapeable. Only the second is genuinely at risk.
2. Check the depth of history — decades of consistently collected data cannot be recreated by a model.
3. Ask whether AI systems need this data to be accurate, which would make them customers rather than substitutes.
4. Look for licensing revenue from AI firms as the confirming evidence.

• Data that is proprietary but not exclusive. If a competitor holds equivalent data, licensing power is shared.
• Applying the "AI needs our data" argument selectively. The same author calls Adobe too hard on the identical question.

5. Test a razor-and-blade lock-in on capital cost, training cost and consumable capture

1. Establish the up-front cost of the installed base item to the customer — the higher it is, the longer the lock lasts.
2. Add the human cost: training, certification, workflow change. This is usually the stickier half.
3. Confirm the consumables and service are proprietary, so the annuity actually accrues to the manufacturer.
4. Estimate the installed-base growth rate, since that is what compounds the recurring revenue.

• The installed base saturating. Once the hospitals that want a robot have one, growth reverts to consumables alone.
• Competing robots being placed for free or on lease, which removes the capital barrier that creates the lock-in.

6. Look for the manufacturer whose financials read like a software company

1. Screen manufacturers on gross margin, return on invested capital, cash conversion and capital spending as a share of sales.
2. Flag any that combine a high margin with CAPEX below a few per cent of revenue and cash flow above reported profit.
3. Establish why: vertical integration, a proprietary software layer, or a recurring service contract attached to the hardware.
4. Then check what the cash is used for — this profile funds acquisitions without debt.

• Municipal demand, which is a budget decision. The current quarter shows revenue and EPS falling because "cities held off on spending for a while".
• Channel destocking, which flatters the prior year and depresses the current one. Both are noted here.

7. Read a worst-performers table as a candidate list, then apply the quality filter separately

1. Maintain a fixed investable universe so a monthly performance table is a screen, not a market summary.
2. Publish both tails. The best performers tell you what you would now be paying up for.
3. Treat the worst list as candidates for research only, and require the quality work before acting on any of them.
4. Separate the reason for each fall: a cyclical trough, a structural change, or an unrelated market move.

• A monthly table encouraging monthly action. The stated strategy is low turnover; a fresh list of fallers every month pulls the other way.
• No commentary attached to the tables. A -42.7% month is either an opportunity or a warning, and the table does not say which.

8. Know the construction rule of any ranked list before you use it

1. Find out what the list excludes. A ranking of ideas outside an existing portfolio is not a ranking of best ideas.
2. Check whether a name on the list is one the publisher would actually buy, or merely one they rank for readers.
3. Cross-reference the ranking against the same publisher's own portfolio decisions on the same name.
4. Treat any divergence as information about the list's purpose, not as a contradiction to be resolved.

• Ranked ideas the publisher does not own. A reader following the list mechanically ends up in a different portfolio from the author's.
• Internal inconsistencies that reveal the seam — the conclusion here calls S&P Global a "legal duopoly" after naming three raters in the body.