Actionable insights — Update Buy-Hold-Sell List: March 2026

1. Run three independent valuation screens and treat disagreement as information

1. Screen 1 — multiple versus its own history: current forward P/E against the five-year average. Cheap here means cheap relative to how this business has been priced, not to the market.
2. Screen 2 — earnings growth model: expected EPS growth + dividend yield ± multiple change to an assumed exit P/E, producing an expected annual return and a fair value.
3. Screen 3 — reverse DCF: solve for the growth the current price already requires, then subtract it from the growth you expect. A positive difference is the margin of safety.
4. Require agreement for the strongest ratings; where the screens disagree, name which one is doing the work and ask whether that is the right one for this business.

• Screen 1 rewarding a permanent derating. A business whose economics have changed should trade below its own history.

2. Use the count of Buys as a market-wide indicator, not just a shopping list

1. Keep a fixed universe and a fixed rating rule, so the number of Buys moves only when prices or fundamentals move.
2. Track that count over time — it becomes a breadth measure of your own opportunity set, independent of index level.
3. Add a stricter version: how many names are cheap on all screens simultaneously, which strips out single-screen artefacts.
4. Read an extreme reading as a signal about deployment pace rather than as a prediction.

• Universe drift. A count is only comparable over time if names are not being added and removed for reasons other than the rating rule.

3. Decompose the index before concluding the market is expensive

1. Measure the weight of the largest handful of constituents; above roughly a quarter, index statistics describe those names rather than the market.
2. Recompute the index return and the index multiple with them excluded.
3. Compare the two multiples. A large gap means "the market is expensive" is a statement about the top ten, not about what you can buy.
4. Direct the search at the excluded remainder, which is where the derating actually happened.

• The equal-weight index as the honest check. If it is also expensive, the decomposition argument does not hold.

4. Hunt the gap between business compounding and shareholder return

1. Pick a start date several years back and measure two things separately: the growth in revenue and operating profit since then, and the total return to shareholders.
2. A large business gain against a zero shareholder return means the entire change is multiple compression — which is recoverable, unlike a deterioration in economics.
3. Verify the compounding is still running now, not just historically, before treating the gap as an opportunity.
4. Then check the multiple against its own history to size how much of the gap could close.

• A starting multiple that was itself a bubble. Some of the gap is the correction of an unsustainable entry price, not a discount.

5. Distinguish "cheaper" from "no longer investable" — and remove rather than downgrade

1. When a business deteriorates, ask whether the change is to the price or to the economics.
2. Two categories force removal rather than a lower rating: the people who created the edge have left, and a regulator has capped what the product may earn.
3. Delete such names from the universe entirely, so a falling price cannot later pull them back onto the Buy screen mechanically.
4. Record the reason, so the decision can be reversed if the cause reverses.

• Screens that keep re-adding a deleted name. If the universe is generated automatically, the exclusion has to be enforced explicitly.

6. Publish the bear case inside the buy rating

1. State the competitive fact that caused the decline, in the competitor's favour, before making the case.
2. Separate the two questions: is the business worse (often yes), and is the price more than compensating (the actual decision).
3. Show the screen that disagrees rather than only the ones that support the rating.
4. Set the falsifier — the level of growth or share loss at which the valuation stops compensating.

• The dissenting screen being quietly dropped in later issues once the position is larger.

7. Rate your own holdings on the same sheet as everything else

1. Run every position through the identical screens you apply to candidates — no exemption for names you already own.
2. Expect the winners to migrate to Hold: a position that has performed well should get more expensive on its own screens.
3. Use the resulting spread to direct new money to the widest discounts inside the book before looking outside it.
4. Publish the whole table, including the holdings that fail, so the rating is a rating rather than a defence.

• Ratings and actions diverging. A Hold that fails all three screens is, in practice, an unexecuted sell — the sheet says so even if the text does not.

8. Hold downgrades to the same standard of explanation as upgrades

1. Write the reason for every rating change, in both directions, at the moment it is made.
2. Notice when only the positive changes are argued — that asymmetry is a tell about where the writer's attention is.
3. For an unexplained downgrade, reconstruct the likely cause yourself before accepting or rejecting it.
4. Keep the reasons, so the next rating change on that name can be checked against the last one.

• Silent downgrades preceding a quiet disappearance from the universe — the change that never gets explained is the one worth reconstructing.