Actionable insights — Buy-Hold-Sell List: June 2026

1. Price a hot listing by how long it needs, not by whether it is expensive

1. Take the market value and pick a multiple you would actually be willing to pay — separately on sales and on profits.
2. Solve for the revenue or the profit the company must reach to make that multiple true.
3. Divide by a generous growth rate to get the number of years the price already assumes. That number, not the multiple, is the thing to react to.
4. Assume nothing goes wrong along the way, then note explicitly that you have assumed it.
5. Compare against an independent fair value (an analyst house, an appraisal) purely as a sanity range.

• Conglomerate listings where one part is genuinely excellent — the article concedes SpaceX "is not exactly like Pets.com" and compares the pricing behaviour, not the business.
• Underwriting incentives: "they are making over $500 million just in fees." The seller chose the moment and the price.

2. Run three valuation models and let them disagree

1. Value every candidate three independent ways: an earnings-growth model (growth + dividend + an assumed exit multiple → expected return and fair value), a multiple-versus-own-history test, and a reverse DCF (what growth does today's price require?).
2. Record all three, including the ones that disagree with your conclusion.
3. Treat unanimity as a stronger signal than any single large discount — count the names that clear all three.
4. When they conflict, name which model you are trusting and why the others fail for this business type.

• The assumed "fair exit PE" doing the work: many fair values here rest on a 25x exit. Change that input and the discount changes with it.
• Models that are not independent — a forward PE and an earnings-growth model share the same estimate of next year's earnings.

3. Put the quality gate in front of the cheapness screen

1. Maintain a pre-vetted universe that has already passed a quality standard; run valuation screens only inside it.
2. Publish the screen outputs anyway — they are candidates for research, not for purchase.
3. Require a separate, explicit quality rating before anything on a cheapness screen can be bought.
4. Track which screen names never get a rating; that residue is the discipline made visible.

• A screen that is mostly the same de-rated sector — half these names are software and payroll, i.e. one bet with twenty tickers.
• Deep-value traps in the reverse-DCF list: a price implying negative growth is either an opportunity or the market's correct forecast.

4. Convert "do nothing" into a checklist you have to answer

1. Write four questions you must answer for every holding when its price falls: is the valuation reasonable, has anything structurally changed, are the fundamentals still strong, is the moat still durable?
2. Answer them in writing, per name, on a fixed schedule — not when you feel anxious.
3. If all four are yes, the required action is explicitly none.
4. If one is no, that specific answer, not the price, is the trigger to act.

• "Nothing" as a default rather than a conclusion — the checklist only works if a "no" is genuinely available.
• Opportunity cost as a legitimate sell reason (which is what was used on Judges Scientific), rather than as a rationalisation after a fall.

5. Mark the cases where your rating overrides your own models

1. After the models are run, list every name where the rating disagrees with all of them.
2. For each, write the reason the models cannot see — asset backing, holding-company structure, an accounting artefact, a management judgement.
3. If no such reason exists, change the rating or the model.
4. Review those overrides specifically at the next revision; they are where discretion, and therefore error, concentrates.

• Overrides that always run in the direction of the existing position.
• A model applied to a business it does not fit (a forward PE on an insurer's holding company) producing a number that then has to be argued away.

6. Publish the comparison that includes your own scoreline

1. Build a like-for-like table of your portfolio against the index on quality, capital intensity, growth and valuation.
2. Include realised return over several horizons in the same table, even when it is against you.
3. Use the gap between the fundamental columns and the return columns as the explicit statement of your bet.
4. Re-publish it on a schedule so the bet can be scored rather than re-argued.

• Portfolio aggregates flattered by one or two extreme constituents — an FCF/net income figure of 195.8% is not a normal steady-state number.
• The index's own capital intensity being the AI build-out; that column may be the most informative one in the table.