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

1. Decompose every big price move into EPS and multiple before reacting

1. Write the price as EPS × P/E. Over the move in question, pull both series (trailing or forward EPS, and the multiple).
2. Attribute the change: how much came from earnings, how much from the multiple.
3. If EPS is intact and the multiple did the work, the move is about sentiment — "the fundamentals didn't change. The narrative did."
4. Put the current multiple against its own ten-year range, not just its average: a range of 31×–89× tells you how far sentiment alone can carry the price in either direction.
5. Only then decide whether the drop is an opportunity (multiple compressed, EPS intact) or a warning (EPS falling).

• A compressed multiple accompanied by falling forward estimates — that is the market pricing an EPS decline, and the decomposition should use forward EPS to catch it.
• Analyst price targets moving with the price (the Salesforce chart): targets are a lagging read of sentiment, not an independent input.

2. Classify the cause of a drawdown as temporary or permanent

1. List the stated causes of the fall.
2. For each, ask whether it attacks the long-run demand for the product, or only its timing.
3. Look for a structural floor under demand — regulation, necessity, contracts.
4. Check the next reported quarters: if EPS keeps growing through the scare, the thesis was right and the multiple should follow.
5. Resist both reflexes the post names — buying every dip, and assuming the market knows something you don't.

• A "temporary" cause that lasts more than a few quarters — check the customers' funding data, not just the company's commentary.
• The discount disappearing after the recovery: this month's sheet puts Medpace only 5% below fair value.

3. Run several valuation screens and write up one name per screen, bear case first

1. Rank the rated universe three ways: forward P/E versus own five-year average, an earnings-growth expected-return model, and a reverse DCF.
2. From each top ten, pick one name to examine.
3. State the market's worry in one or two sentences.
4. Answer it with a structural point (market structure, product role) and one datapoint (guidance, organic growth, buybacks).
5. Name the external catalyst that could prove the worry right, so it can be monitored.

• The FHFA decision on single or bi-merge credit reporting — a direct hit to bureau volumes if adopted.
• Names that appear on two or three of the top tens (Adobe, Gartner, lululemon, Paycom here) — the overlap is stronger evidence than any single rank.

4. Audit a published model month on month before trusting its rankings

1. Keep the previous issue's sheet and diff the same columns for the names you care about.
2. Flag any input that moves far more than the price did — a reverse-DCF requirement jumping from single digits to 70–90% on a small price change points to a changed input or formula, not to the business.
3. Check that the headline list is computed from the same figure as the prose.
4. Reconcile counts: the number of Buys claimed, the rows shown, and the holdings list.

• Next month's sheet: whether the reverse-DCF readings revert (a one-off data error) or persist (a methodology change nobody announced).

5. Treat the count of Buys as a market-level valuation gauge

1. Keep the quality universe fixed, so the number rated Buy moves only with price.
2. Log the count every month.
3. A rising count with unchanged business quality means multiples across the universe are compressing — a better time to add capital.
4. A falling count means the opposite: raise the bar for new buys and let cash build.

• A count that rises because the universe grows rather than because prices fall — the two must be separated.