Actionable insights — Best Buys: July 2026

1. Keep a standing shortlist of what you would buy next, separate from what you own

1. Maintain two distinct lists: the portfolio, and a ranked shortlist of names that would be bought if cash or conviction allowed. Never mix them — a name cannot appear on both.
2. State the rule out loud so the shortlist cannot be misread as a conviction ranking: the portfolio names are excluded because they are all still wanted.
3. Refresh the shortlist on a fixed monthly cadence rather than when something feels interesting; the calendar removes the impulse.
4. Rank it. A shortlist without an order does not force the comparison that makes it useful.
5. When cash arrives, buy from the top of the list rather than starting a fresh search — the decision was already made under calmer conditions.

• A shortlist that never converts — if nothing is ever bought from it, it is a reading list, not a pipeline.
• The reverse failure: buying the #1 name automatically without re-checking the price that made it #1.

2. Screen your own watchlist by monthly drawdown, not by news

1. Define an "investable universe" in advance — the businesses you have already judged good enough to own at some price. This is the hard work, done once.
2. Every month, rank that universe by price change and print the five worst and five best.
3. Treat the worst-performer list as the sourcing queue: quality already established, price now lower. State the reason explicitly — "the cheaper we can buy great companies, the better."
4. For each faller, ask only whether the decline reflects a change in the business or a change in sentiment; you are not re-underwriting from scratch.
5. Use the best-performer list for the opposite discipline — a check on names that have run and may no longer clear the valuation bar.

• A universe that quietly expands to include whatever has just fallen — the pre-vetting is the whole point.
• One-month windows catching noise; the screen surfaces candidates, it does not decide.

3. Ask whether the scarce asset can be permitted, not just bought

1. Identify the physical or legal input a competitor would need to replicate the business.
2. Ask whether a well-funded rival could simply buy it. If yes, the moat is capital, and capital is available.
3. Ask instead whether they could get permission for it — zoning, environmental approval, a licence, a regulatory designation. A moat made of permissions cannot be out-spent.
4. Check whether that scarcity is tightening over time (cities expanding, rules hardening) rather than static.
5. Confirm ownership, not tenancy: an incumbent that rents the scarce asset is exposed to whoever owns it.

• Rules that can change — a permitting moat is a political asset and politics moves.
• Technology routing around the physical asset entirely (fewer total-loss cars, remote assessment).

4. Test an AI-disruption fear by asking what the customer is actually buying

1. When the market marks a business down for AI risk, separate the software from the job it performs.
2. Ask what happens to the customer if the job is done wrong. If the answer is regulatory penalties, lawsuits or unpaid staff, price is not the buying criterion — reliability is.
3. Check retention and tenure directly; a stated switching cost that does not show up in churn is a story.
4. Count the number of adjacent services the same customer buys. Each one multiplies the cost of leaving.
5. Verify the cash: free cash flow at or above net income says the reported profits are real, which matters most when you are betting against a narrative.

• Switching costs that protect the relationship but not the price — a captive customer can still refuse an increase.
• A new entrant that sells to the customers who have never chosen a provider yet, rather than trying to convert yours.

5. Notice when your candidate list is one idea repeated, and treat that as a concentration decision

1. After building the shortlist, describe each name in a single clause and look for the shared structure.
2. If they all share one economic mechanism, you have made a factor bet, not five independent selections — decide whether you intended to.
3. Ask what single condition would impair all of them simultaneously (in a transaction-toll book: a collapse in transaction volumes, or regulation of the fee itself).
4. Either diversify the mechanism deliberately or size the whole group as one position.

• Regulatory attention to the fee itself — interchange caps, ratings-fee reform, index-licensing scrutiny — which is the common failure mode.
• Volume sensitivity: a toll on transactions is not defensive if transactions stop.

6. Treat a returning operator with a measurable record as a dated catalyst

1. When a former CEO is brought back, quantify the record from their previous tenure with exact start and end dates rather than a reputation.
2. Ask what specifically has changed since they left — is the returning operator facing the same business or a different one?
3. Pair the appointment with the balance sheet: a returning operator with a large cash pile has more levers than one without.
4. Define what evidence of the turn would look like within a year, so the story cannot run indefinitely on the appointment alone.

• The prior record being a market cycle rather than an operator's skill.
• Buybacks used as the whole plan when the reason for the slowdown is demand.

7. Flag it when a company that always built in-house starts acquiring

1. Establish the company's historical capital-allocation habit — build or buy — as a baseline.
2. When the habit changes, treat it as information about management's view of its own runway, not as routine M&A.
3. Ask what capability is being bought and whether it is adjacent to the core or a diversification.
4. Watch integration in the next two or three reports; a first acquisition is where the discipline is tested.

• The second and third deals arriving quickly — a broken habit rarely breaks once.
• Culture dilution at a company whose entire advantage was one internally built platform.