Pieter Slegers — research hub

Stock & name index

▲ Positive

► Neutral / referenced

▼ Negative

Overall thesis

• The edge is behavioural. Informational and analytical advantages are gone (everything's online; quants/PhDs/AI dominate large caps). Retail's one durable edge is behaviour — no clients to answer to, so you can hold cash and wait for the obvious pitch. He deliberately avoids the crowded big-tech corner where he'd have no edge.
• Six quality criteria, all required: a moat; founder skin-in-the-game (an HBR study: founder firms outperform ~3.9%/yr); low capital intensity; capital allocation (ROIC >15%); high profitability (>10% margin with >90% earnings-to-FCF conversion — "earnings are an opinion, cash flow is a fact," per Siegel); and attractive growth (>7% revenue, >9% FCF/share).
• The written cases run a 15-step worksheet with hard thresholds — gross margin >40%, ROIC >15%, ROE >20%, interest coverage >15x, goodwill/assets <20%, CAPEX/sales <5%, net margin >10%, FCF/net income >80%, SBC <10% of net income — rolled into a Total Quality Score /10 (CMG.TO 8.3; Brookfield 8–9 per section). Failures are recorded, not argued away.
• The golden goose: the rarest, best business is a high-ROIC compounder that can reinvest almost everything at that rate (Dino Polska) — vs a great business too cash-rich to reinvest, which just buys back stock (Apple). The listed variants he keeps returning to are the serial acquirer (Constellation, Topicus, CMG, Watsco, Asseco) and the toll bridge (FICO, S&P Global, MSCI, Cintas).
• Three valuation methods: forward PE vs its own 10-yr average · an earnings-growth model (EPS growth + yield ± multiple change) · a reverse DCF (solve for the growth the price implies, and compare it with delivered growth). "Roughly right beats exactly wrong — value it on a napkin."
• Entry discipline is a price, not a feeling. Quality alone never triggers a buy: FICO was passed over in March 2026 at 21.6x 2028 EPS ("still not very cheap") and bought in April after a further fall; Cintas is admired and left on the watch list with a stated limit — a 25x forward PE, i.e. $132 against $174.
• AI: concede the disruption where it's real, then test the moat. Software whose only value is completing a basic task is in trouble; quality names are defended by proprietary data (ADP, S&P Global), high switching costs ("most CFOs won't risk their entire operation to save a few dollars") and operating leverage from embedding AI in their own products (Adobe's Firefly, S&P's Kensho, ADP Assist). Confirming signal: open-market insider buying (CSU.TO, KKR, MSCI).
• The book is ranked, and every demotion has one named cause. The disclosed portfolio (18 names, April 2026) sits in four tiers — Very Strong · Strong+ · Strong · Medium ("Should we consider selling these stocks?") — and a name drops a tier for a single specific reason, never for price: BRO for one oversized acquisition ($9.83bn Accession — "we prefer Serial Acquirers to execute a lot of small acquisitions"), KPG.AX for founder governance after Brett Kelly was margin-called on $64m of pledged shares. MEDP is the control: a 30.3x forward PE keeps top conviction, because the tiers rank the business, not the entry.
• Match the profit metric to the business model. Reported earnings are the wrong number for an acquirer, so he substitutes the cash an owner actually receives: FCFA2S (Free Cash Flow Available To Shareholders) for CSU.TO and TOI.V — 4.8% and 5.1% forward yields, both all-time-cheap — NPATA for KPG.AX (Buffett's Owner Earnings, amortisation added back), and Distributable Earnings for BN. Asset-based vehicles get the NAV-discount treatment instead (III.L, HGT.L at ~30% off a £5.62 NAV, BN at $46.5 against a $68 intrinsic value).
• Buffett's good / great / gruesome taxonomy is the live filter. The question is what the owner must put in to earn another dollar out: good businesses have a moat but must buy growth with capital (FlightSafety); great ones grow without it (See's Candies — $8m of capital earning $5m in 1972, $40m earning $82m decades later); gruesome ones grow fast, eat capital and earn nothing (airlines). In a market that is expensive, modelling 24% S&P earnings growth for five years and "full of good and even gruesome businesses… priced for perfection," the answer is to narrow rather than hedge: "only the truly great businesses are good enough."
• Conviction is converted into size, with a price attached. The June 2026 transaction issue adds to names already ranked Very Strong rather than opening new ones, and specifies each add three ways — V $20,000 / 60 shares / limit $337 · KNSL $20,000 / 65 shares / limit $330 · AMP $10,000 / 22 shares / limit $460 — announced before execution. No new idea is introduced; the weights do the work.
• Two valuations that fail differently, and both must agree. A forward PE against the company's own 10-year average, plus a reverse DCF solved at a 10% required return: KNSL at under half its historical multiple and needing 3.3% FCF growth; IPAR at 23x against ~34x ("an undervaluation of almost 40%") on the same 3.3%; AMP at its average multiple but needing only 2% — the case there rests on the DCF alone. The same pair produced the HEICO pass: a 7.8/10 Quality Score with valuation scored 2/10, a ~53-57x forward PE and a reverse DCF demanding 21.4% growth against 8.7% delivered over a decade.
• Cyclicality is a two-sided test, not a blanket exclusion. Memory (MU, with WDC and SNDK alongside) is refused because "these companies don't have any pricing power" and today's margins exceed the 2018 peak that preceded a collapse — Grantham's "if you make obscene profits, you'll get ferocious competition." The same price-cycle logic makes Kinsale a buy, because a structural cost advantage (own technology, small-account E&S focus, in-house underwriting) survives the soft leg that removes competitors: "Mr. Market is pricing Kinsale like the soft market will continue forever."
• The tollkeeper filter, adopted openly from Chris Hohn. Businesses that "control essential infrastructure or platforms and charge users a small toll each time they pass through," with demand compelled by regulation, physics or an installed base — and preferably five or six overlapping barriers rather than one. The portable arithmetic: "if you can price 1% above inflation and you have a 20% profit margin, your profits will grow 5% faster than revenue." Slegers' comment: "At Compounding Quality, we try to do the same thing."
• In a drawdown, report the cash rather than the price. The July letter substitutes the portfolio's look-through free cash flow for its market value — $65,520/yr, $5,460/mo, $1,260/wk, $0.12/minute — on Graham's voting-machine/weighing-machine logic, and pre-commits to base rates: "one year out of three the market falls at least 10%; one stock out of three will be a disappointment; one year out of three we will underperform." The historical anchor is 1998-2000, when the Nasdaq rose 75% while Berkshire fell 18.9% and Fairfax fell 64% — both then beating the index by a wide margin. The one instruction is not to switch style at the extreme.
• Survival is a valuation input, and the lists that use it carry no price. The August 2026 Lindy series screens on corporate age alone — against a base rate of "50% of all businesses fail within 5 years; 80% are gone within 10" — and states the mechanism precisely: "the value of a business depends on all the cash it will generate in the future (discounted to today)… a longer life means more future cash flow." Its own counter-examples are left in (Disney flat for a decade; Pfizer at 9.8% a year against Eli Lilly's +17,000% on a near-identical description), and the second half withdraws the implied promise — "none of these companies are guaranteed to outperform the market." Read these as universe-building, not as buy lists: no fair value, multiple or expected return appears anywhere in the durability issues, which is the opposite of the portfolio updates.
• Growth investing appears once, as a guest, and is left unrebutted. The Kris Heyndrikx interview is the archive's only sustained argument for a different discipline: a three-filter funnel (do I find this interesting → 20%+ consistent revenue growth → management quality) that eliminates 90% before any number is examined, a founder's pull treated as an off-balance-sheet asset, and a fortnightly DCA schedule with an index-keyed escalator (-10% → invest 20% more, -20% → 50%, -30% → double). Disclosed entries: SHOP $5.58, NVDA ~$2.50, NET $39, CRWD $24.53. No valuation step appears anywhere in it — and the interview's own best lesson is a failure: Nvidia was skipped as an official pick for being "too big at around $60 billion" and returned 8,720%.
• The ratings are published for the whole universe, and they now change for business reasons. The monthly Buy-Hold-Sell List carries a standing rating on every name in the investable universe with all three valuations attached — 55 on Buy, 62 undervalued on all three methods, seven Strong Buys inside the portfolio — which makes the disagreements between the models visible (BN 49.9% under on fair value but 2.2% expensive on its own multiple; FFH second in the universe on the reverse DCF and 13.8% expensive on the multiple). August 2026 introduced two firsts: downgrades on competition rather than price (DNP.WA and NVO, the latter cut while showing a 54.5% discount and the book's highest expected return), and an outright SELL on valuation (WSM) — applied asymmetrically, since GAW.L is held in the same issue at 51.1% over fair value. The September list pushes the Buy count to a record 57 (one addition, ISRG; upgrades THEP and EVD; no cuts), finally marks DNP.WA HOLD on the portfolio sheet, and shows the book at 20 holdings with NVO gone — while several reverse-DCF readings jump implausibly month on month (BN 90.6%, EVO 74.6% required), so the sheet itself now needs auditing. Three days later the first post-reset transaction doubles Fairfax (30 more shares at CAD 2,300, ~2.7% → ~5.3%) and prints the whole book's Earnings Growth Model: 13.9%/yr on average, down from the 17.1% modelled on 1 September, with no reason given. Fairfax tops it at 18.6%, but its row is computed on revenue rather than EPS. The largest-but-lowest-returning positions (EVO 8.6%, GAW 6.7%) keep their size.
• Go-forward policy (1 September 2026): raise the bar and let the winners be bigger. "Almost every time I made a buy decision because I thought the company was somewhat quality but definitely cheap, it was a mistake in hindsight." The rules: 15-20 stocks (from 21), developed countries only, minimal trading, no market timing, and eight required characteristics ending in "trading at fair valuation levels" — with the target unchanged at S&P 500 + 3%/yr. The weights are re-aimed on quality, not price: BN, SPGI and FFH.TO up; LVMUY, NVO and DNP.WA down — note that ZTS, EVO.ST and AMP are named as the cheapest holdings and get no increase. Whole-book statistics are published for the first time (18x NTM P/E vs 20x, 14% vs 12% EPS CAGR, 6.0% FCF yield, 13.2% owner's-earnings growth → a modelled 17.1% return), and so is the record: a three-year CAGR of 1.3% against the index's 20.1%. Three names go unmentioned in the audit — GAW.L (33x, a 6.68% modelled return), III.L (−56% in twelve months) and the Polish holding that the new developed-markets rule excludes.
• The September policy has an April rehearsal, and the April version is the more testable one. The five-part April 2026 review ends on 28 April with the same self-criticism the 1 September letter repeats — "every single time I bought a company not because I thought it was the highest quality, but because it was cheap, it ended up being a mistake (so far)", naming Text SA, OTC Markets and Novo Nordisk — and converts it into two mechanisms rather than a sentiment. First, Buffett's ten-year-closure test applied to all 18 holdings and published name by name: 15 Yes and three "Not sure" (EVO.ST, JDG.L, NVO), which are exactly the three Medium convictions from Part II — two independent frameworks agreeing. Second, the linearity of growth as the new filter: two companies compounding at 10% are not equally valuable if one does it smoothly and the other swings +30%/−15%/+20% — "it's always Company A." The funnel is sized for the first time (153 stocks, 0.4% of the 40,000+ listed worldwide), and the formulation is the inversion the September letter later reuses: "We don't want good companies at cheap prices. We want wonderful companies at fair prices." The verdict produced one named sell candidate (JDG.L) and five buy candidates (KKR, III.L, FICO, FFH.TO, MSCI) — of which only Fairfax was bought, three and a half months later, and Judges Scientific was still held in August.
• Holding-company arithmetic: the accounts are the opportunity. The Fairfax India study sets out the general method — split the book into mark-to-market and management-estimate, then compare realised returns (19%/yr) against unrealised private marks (8%) to size the conservatism, multiply by the 67% of assets not marked to market, and demand a named event that converts the estimate into a price (the Anchorage IPO, which would take mark-to-market coverage from 31% to 85%). Proof is required in cash: Poseidon, carried at $15.50 and part-sold at $28.30. And the last step is a fee test — the exposure is taken through the fee collector (FFH.TO), not the payer.
• The funnel's first stage is a published screen — check that every filter bites. The September 2026 screener encodes the quality checklist as eight thresholds (5-yr revenue CAGR >5%, 5-yr diluted EPS CAGR >7%, 5-yr gross margin >40%, 5-yr net margin >10%, 3-yr FCF margin >10%, CapEx/revenue <5%, 3-yr ROIC >15%, 3-yr ROCE >20%), excludes banks (the metrics invert) and China (circle of competence), and cuts 50,000+ listings to fewer than 100. The shortlist step is a business-model paragraph and one operating chart per name — ASML, Fastenal, MSCI, Hermès, Mastercard — with no valuation, so the screen feeds the Buy-Hold-Sell work rather than replacing it. The post's own output shows the checks worth adding: the CapEx column reads 0 for every survivor (the filter was typed as 0.05), 100% gross margins on Mastercard and Rightmove are data artefacts, and domicile-based country exclusion still passes PDD.
• Sell rule: never on valuation (his biggest, expensive winner stays untrimmed); sell only when the investment case is no longer intact. Test: "would you own it if the market closed for 10 years?" His only two sells (Text S.A., Ulta) were both admitted thesis errors.
• Argued longs across the archive: Kinsale, Medpace and Dino Polska from the 2025 interview; CMG.TO as "the next Constellation Software"; and the 2026 monthly picks — ZTS, ACP.WA, FFH.TO, FTNT, POOL, III.L, CSU.TO, TOI.V, ADP, ADBE, ADYEN.AS, SPGI, KKR, MSCI, MELI, FICO. A coffee-can list of ten "own forever" names (BRK.B, BRO, WSO, MKL, GAW.L, CSU.TO, LVMH, SYK, ODFL, ZTS) sits behind all of it. The April 2026 Portfolio Update finally names what he actually holds: MEDP, GAW.L, KNSL, AMP, V, CSU.TO, TOI.V and BN at Very Strong, then KPG.AX, BRO, HGT.L and ZTS at Strong+ (Part II covers the remaining six). Summer 2026 adds IPAR — a thirteenth disclosed holding, presumably from the unarchived Part II — and a set of names carried in as illustrations rather than picks: TJ Terwilliger's twenty-year list (BRK.B, ADP, WM, BN, MA, SPGI, ROL, CTAS, SHW, GWW), Chris Hohn's tollkeepers (GE, V, MSFT, MCO, SPGI), and two explicit passes — HEI on price and MU on the cycle. August 2026 adds one real purchase — FFH.TO at a CAD 2,300 limit — and a large tranche of names carried in on durability alone, with no price on any of them: the Lindy twenty (MCD, MAR, DIS, MCO, BAC, MMM, PEP, JNJ, PPG, CVX, KO, LLY, BF.B, SHW, UNP, AXP, PFE, PG, DE, CL), Arka's twenty-year list (RACE, ODFL, TMO, AMAT, SU.PA, XYL, ASML, MSFT, GOOGL, V), the Diploma pitch (DPLM.L), and a guest's growth book (SHOP, NVDA, NET, CRWD, NU, GOOG). September 2026 adds the first full investment case on a bank — Nu Holdings at 7.6/10, constructive on its own scoring but an emerging-markets name the 1 September developed-countries-only rule keeps out of the book, with ITUB and BBD carried in as multiple comparators.

The product — Compounding Quality

Grounded in Slegers' own statements in the archived interview (2025-APR-20, Thoughtful Money) and in the archived newsletter issues themselves (2024-FEB-22 → 2026-APR-05).

• Coat-tail a disclosed, aligned portfolio. Every position is published and it's all his own money — the transparency and skin-in-the-game are the pitch, letting a retail investor mirror a quality portfolio without doing the tens of thousands of hours of work themselves.
• Act before he does. Sunday buy/sell announcements are structured so subscribers can transact ahead of his own execution — an unusually reader-first alert cadence.
• Learn the process, not just the tickers. The cases teach the repeatable screen and valuation math (the actionable-insights distillation), so a subscriber can eventually run it independently. The written issues go further than the interviews: the 15-step worksheet with its thresholds is published in full.
• A monthly discipline, not a feed of alerts. The Best Buys format repeats the same sequence every month — market reading, worst performers, spotlight, ranked five — so a reader inherits a habit (shop the drawdowns of a pre-vetted universe) rather than a stream of tips.
• Passes are published too. Names are written up in full and then declined on price, with the level that would change his mind stated — which is what makes the "buy" issues legible rather than promotional.

Transcripts

One dated page per appearance — each has its talking points and the saved transcript. Newest first.

To process — backlog

Pieter Slegers appearances discovered via search (Pieter Slegers Compounding Quality), not yet processed — verify publish dates & channels, newest first. Limited to the last ~2 years. None queued from YouTube yet. Known Substack gaps, all to be resolved on the next archives.json sweep of compoundingquality.net: (1) 10 Stocks for the next 20 Years — Pieter's own instalment: the 13 August issue is the third edition and refers back to "Pieter's selection" and "TJ's selection"; only TJ's and Arka's are archived here. (2) Time to Raise the Bar, Part II — the position-by-position review promised in Part I for "Thursday" (~2026-SEP-03), where the book is to go from 21 names to 15-20; URL not yet published. (3) The Lawrence Cunningham interview linked from the 4 August #QualityTuesday issue; URL not resolved. (4) The Owner's Manual (2023) — the standing statement of philosophy referred to twice in the 1 September letter ("still highly relevant today"); a subscriber document rather than a dated post, not yet captured. (5) Brookfield Deep Dive, Part 3 — the third instalment (valuation and conclusion) following Part 1 and Part 2, expected ~2025-DEC-18; URL not resolved. Deliberately skipped: "Taking a break" (2026-AUG-25) — philosophy and housekeeping only, no securities.

For personal study — not investment advice. Source material © the respective shows / Compounding Quality.