← Analysis page  ·  Pieter Slegers hub  ·  Research hub

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

Three methods, one rating: how to build a ratings sheet that disagrees with itself on purpose, and how to read the disagreements.
2026-FEB-05 · Compounding Quality (Substack) · Pieter Slegers · read ↗ · full analysis · transcript
How to read this page: each insight is a method used in this issue, written so it can be rerun on other names. Written post, so no timestamps.

1. Rate every name on three independent valuation methods, then read where they disagree

The repeatable method
  1. Run the same three checks on every watchlist name, every month: (a) current forward PE against its own five-year average; (b) an Earnings Growth Model — EPS growth + dividend yield, plus or minus the move from today's multiple to a "fair exit PE" — giving an expected annual return; (c) a reverse DCF solving for the growth rate the current price already assumes.
  2. Record all three outputs side by side rather than collapsing them into a single score. The value of the sheet is in the spread between them.
  3. Rate Buy / Hold / Sell on the composite, and accept that a Buy will often fail one method.
  4. Treat a disagreement as the research prompt: a name cheap on the multiple but failing the DCF has a growth problem; a name expensive on the multiple but passing the DCF usually has an accounting problem worth understanding.
Here: BN is 39.1% overvalued on the multiple and carries a 10.6pp margin on the reverse DCF — resolved by using distributable earnings and Bruce Flatt's own 15% target. DFH is the mirror image: 9.9x forward, but the price implies 19.6% growth against 10% expected. WSO, ESQ, KPG.AX, KKR and MKL are all Buys trading above their own five-year multiples.
Watch for

2. Track intrinsic value and price as two separate series, and measure the gap

The repeatable method
  1. Once a year, estimate the change in what each holding is worth, independent of what the shares did.
  2. Put the two percentages side by side. If value rose and price fell, compute how much cheaper the stock became — that number, not the drawdown, is the improvement in expected return.
  3. Aggregate it across the portfolio to get a single sentence about the opportunity set.
  4. Use it to reframe a losing year: a portfolio whose value compounded while prices fell has a better forward return than it did twelve months ago.
Here: "The average company within Our Portfolio increased its intrinsic value by +13.8% last year. Stock prices didn't follow." BRO is the worked case — +13.6% intrinsic value against −21.9% price, so "the stock became 35% (!) cheaper."
Watch for

3. Use the count of names passing all three methods as a market-breadth gauge

The repeatable method
  1. Each month, count how many watchlist names clear every valuation method, not just one.
  2. Track that count as a time series and compare it to its own history rather than to an index level.
  3. Read a record high as evidence about your own opportunity set, not as a market forecast — the claim is "there is more to buy," not "the bottom is in."
  4. Let the count, rather than a macro view, set how much capital you deploy this month.
Here: "Today, 51 companies are undervalued on each valuation method in our watchlist. This number (51) has never been higher." Purchases follow within days — $50,000 on 1 February, another $50,000 on 22 February.
Watch for

4. Read the top of a valuation screen as a map of the market's current fear

The repeatable method
  1. Sort the watchlist by discount to its own historical multiple and look at the top fifteen as a group, not as fifteen separate ideas.
  2. Name the common factor. If one industry dominates the list, you are looking at a single narrative repriced across many tickers.
  3. Decide whether you are underwriting the narrative once or fifteen times — buying the whole top of the screen is one concentrated bet wearing a diversified costume.
  4. Then look for the names on the list that do not fit the theme; those are the idiosyncratic opportunities.
Here: the forward-PE screen is dominated by payroll, HR and data software — Paycom (61.6% under), Paylocity (54.1%), Gartner (47.6%), MarketAxess (46.6%), EPAM (46.4%), ADBE (44.0%), Enghouse (44.0%). The outliers are the interesting ones: Goosehead, LULU, Mips, LEM, DiaSorin, NVO.
Watch for

5. Decompose an "expected return" before trusting the ranking

The repeatable method
  1. For each high-ranked name, split the expected return into its parts: growth, dividend yield, and multiple change.
  2. Flag rows where the dividend does most of the work — the growth assumption may be low precisely because the business is not growing.
  3. Flag rows where multiple expansion does most of the work — those depend on the market changing its mind, which is not something you control.
  4. Prefer the rows where growth carries the return, and treat the others as income or as re-rating trades with a different risk profile.
Here: EVO.ST scores 15.3% on the Earnings Growth Model with only 5.7% EPS growth — a 5.0% dividend yield and a multiple moving from 10.3x to a 15.0x exit do the rest. Insperity (5.2% yield) and Progressive (6.9% yield) rank on the same mechanism. DNP.WA, by contrast, scores 15.9% on 15.0% growth and no dividend at all.
Watch for

6. Keep the mechanical rating separate from the thesis, and expect it to whipsaw

The repeatable method
  1. Let the monthly rating move purely on price versus modelled fair value, without rewriting the business case each time.
  2. When a rating flips, ask which input changed. If only the price moved, the flip is information about valuation, not about the company.
  3. Publish the flip anyway. A rating that never downgrades a favourite is not a rating.
  4. Keep the thesis in a separate document, so a Hold does not silently become a sell and a Buy does not silently become a conviction.
Here: POOL is Best Buy #1 in January, cut Buy → Hold here, and Best Buy #5 again in March on an unchanged installed-base argument. CHE is cut with no numbers given at all.
Watch for

7. Audit the Holds — that is where the discipline is actually tested

The repeatable method
  1. Run the same three methods on what you already own and publish the output, including the rows that embarrass the portfolio.
  2. Identify holdings the model marks as overvalued, and state explicitly why they are being kept rather than trimmed.
  3. Reconcile the sheet against the actual book each month — a stale row is a control failure, not a rounding error.
Here: GAW.L is marked 61.4% overvalued on a 6.0% expected return and held without comment; MEDP trades 13.3% above its own history; OTCM still appears as a Hold although the position was sold a week earlier on 29 January.
Watch for

Methods distilled from the archived Compounding Quality post for personal study. Not investment advice.