Li Lu checklist — v0.3.0

This profile carries the deep-fundamental owner territory: know the business better than almost anyone who owns it, and think in decades as a part-owner rather than in quarters as a renter of shares. Management is judged above all by what it does with retained capital, and value is only interesting if it is compounding rather than melting. It is the one lens with explicit Asia context — the framework is adapted to Japanese market structures: cross-shareholdings, parent-child listings, cash-hoarding culture, and the TSE's capital-efficiency reform pressure. Li Lu's public corpus is the thinnest of the four masters — two Columbia lectures, the Peking University talks, and the translated book and essays — so this checklist is deliberately smaller than its siblings: items are capped at what he actually taught.

Items tagged [PRACTITIONER] are answered by the practitioner, never by the profile agent — the agent marks them practitioner-pending and moves on; where an Agent proxy is given, the agent evaluates the proxy instead.

Evidence grades: quote-backed = verbatim-verified primary source; teaching-derived = faithful to published teaching, paraphrased; grades signal how much weight the citation itself can bear.

Knowledge bar

L1. Ten-year knowledge bar

• Asks: To predict this company's earnings power ten years from now, what are the two or three variables that actually decide the outcome — and do the filings (yuho, segment notes, mid-term plan) let you answer each one? Name the variables and map each to a specific disclosure or figure-table line.
• Good: A short, honest list — core-segment unit economics, pricing power against input costs, demographic exposure — with each variable traced to a public disclosure, plus a plain statement of what remains unknowable. If the decisive variables cannot be answered from the public record, the company is too-hard regardless of how cheap it looks.
• Source: Columbia lectures 2006/2010 (Greenwald course) — the Timberland account and the boundary-of-knowledge theme; Peking University 2015 (tr. Rhodes) — the ten-year minimum-knowledge standard. (teaching-derived)

L2. Ten-year worst-case picture

• Asks: Can you state, from filings alone, what this company looks like in ten years in the worst case — trough revenue, trough margin, earnings floor — and defend each number?
• Good: A written worst-case sketch where every input traces to a figure-table row (worst historical margin, demand floor evidenced by past downturns, recurring revenue base); not a base case with a haircut.
• Source: Peking University talk, Oct 2015 (tr. Rhodes) — "Before investing, we need to know at a minimum what a company will look like in ten years and how it will behave in a downturn." (quote-backed)

L3. Filed downturn record

• Asks: How did revenue, margins, and operating cash flow actually behave in the last two downturns this company lived through?
• Good: Figure-table rows for the actual stress years (2008–09, 2020, local crisis years) showing drawdown and recovery; a company with no lived downturn on record is flagged as an unknown, not assumed resilient.
• Source: Peking University talk, Oct 2015 (tr. Rhodes) — "After a lot of hard work, you will see which companies and industries have a clear picture of what they could become in the worst case." (quote-backed)

L4. Know / assume / pretend ledger

• Asks: Classify every load-bearing input of the thesis as verified fact, explicit assumption, or unknown — do you really know what you think you know, and is anything you are pretending to know carrying the verdict?
• Good: An explicit ledger next to the thesis with facts traced to primary sources; the verdict survives deletion of every "pretend" and "unknown" entry, and assumptions are counted and few.
• Source: Columbia lecture 2010 — "do I really know what I think I know... chances are you are gambling"; Greenwald conversation 2021 — "Be honest what you know, what you assume, what you pretend and what you don't know." (quote-backed)

L5. Unknowns absorbed by the discount

• Asks: Are the acknowledged unknowns items the purchase price already absorbs — would the investment still work at the current price if every listed unknown resolved badly?
• Good: Each unknown priced at its worst case with the margin of safety shown to cover the sum; the thesis requires no unknown to resolve favorably.
• Source: Columbia lecture 2010 — "There are plenty of things I don't know but they don't factor into the purchase because I am using a huge margin of safety." (quote-backed)

L6. Boundary conditions of the thesis

• Asks: Can you state the specific, observable conditions under which your view of this company would be wrong?
• Good: Two or three falsifiable conditions tied to reportable figures (e.g., incremental ROIC below 10% for two years, segment share loss to a named competitor beyond X points); a conclusion without stated failure conditions fails this item.
• Source: Peking University talk, Oct 2015 (tr. Rhodes) — "The most important part of the circle of competence is the boundary. An ability without a boundary is not an ability." (quote-backed)

L7. The halving test and the pre-committed add

• Asks: If the stock halved tomorrow on no new fundamental information, does the work already tell you whether it is headed back to $100 or on to $200 — and at what pre-committed price would you buy more?
• Good: A named, evidence-backed insight plus a written add price grounded in the worst-case valuation floor, set before any fall; if the honest answer to "would you buy more?" is no, the claimed understanding isn't real.
• Source: Columbia lecture 2010 — "when it goes from $100 to $50 you aren't going to know if it will back to $100 or $200"; Greenwald conversation 2021 — "when our favorite companies go on sale... That is the definition of whether we understand them." (quote-backed)

L8. Specialist edge, named concretely

• Asks: On what specific, checkable aspect of this company does your reconciled work give a more complete picture than the consensus narrative — and where in the filings does it live?
• Good: A named piece of filed evidence (segment economics, unit economics rebuilt from footnotes, a disclosure the market narrative ignores) — not a generalized claim of "deep understanding."
• Source: Greenwald conversation 2021 — "you better become a true specialist to the point where you know hopefully better than anybody in the world including the management team." (quote-backed)

L9. Opposite-view test — "What did I miss?"

• Asks: Have you written the strongest bear case — the one the most knowledgeable skeptic of this company would make, with the figures the bears actually cite — and answered it from filings, along with the psychological biases most likely operating on you?
• Good: A steel-manned bear case with a filed-evidence response to each point, plus named bias risks (liking the idea too much, commitment to sunk research) each answered or conceded into the unknowns column.
• Source: Columbia 2006 notes (paraphrase: the fourth checklist question, "What did I miss?", greatly affected by psychology); Peking talks 2015/2019 (paraphrase: test the circle against the most knowledgeable holder of the opposite view). (teaching-derived)

L10. Why the bargain exists

• Asks: Can you name, with evidence, why this stock is cheap and who is selling or refusing to own it — a structural mechanism (crisis headline, zero coverage, retail-dominated register, venue discount, control-structure turn-off) shown to be non-fundamental?
• Good: A documented cause checked against the fundamentals it supposedly reflects (Timberland: Asia panic, lawsuits, no coverage — none impairing the business), plus shareholder-composition and coverage facts; "the market is wrong" backed by a mechanism, not confidence.
• Source: Columbia 2006 notes (paraphrase: check why it became cheap and what others are looking at); Greenwald conversation 2021 — China retains "an enormous amount of inefficiency in the security market." (quote-backed)

L11. Feared exposure quantified

• Asks: Is the headline fear quantified from segment disclosures — what fraction of earnings and assets is actually exposed, and what would a total write-off cost?
• Good: Worst case stated in numbers reconciled to the segment note, Timberland-style: the feared Asian business was tiny, so a complete loss would cut earnings under 5% — against a stock priced for disaster.
• Source: Columbia 2006 notes (paraphrase: "Their Asian business is tiny — it would reduce earnings by less than 5%"). (teaching-derived)

L12. Accurate and complete information

• Asks: Is every load-bearing figure traced to a primary source and reconciled, with no material segment, asset, liability, or claim left unexamined?
• Good: A filing citation for each number and a completeness pass confirming nothing material was skipped (subsidiaries, litigation, off-balance items); the study visibly went to extra lengths beyond summary financials.
• Source: Columbia 2006 notes (paraphrase: most investment mistakes come from inaccurate or incomplete information). (teaching-derived)

L13. Projection honesty

• Asks: If the valuation relies on projections beyond a few years, why is this one of the rare businesses where a 10–20 year projection is defensible?
• Good: Either the value case needs no long projection, or durability evidence (decades of stable economics, structural lock-in) puts it among the handful one can confidently project that far — no banker-style projecting into infinity.
• Source: Columbia 2006 notes (paraphrase: across a 50-year career there are maybe 5–10 opportunities where you can confidently project the next 10–20 years). (teaching-derived)

L14. Key value bottleneck named

• Asks: What single component, technology, or economic factor determines where value accrues in this industry — and where does the company stand on it?
• Good: An explicit "the heart of this industry is X" claim (as the battery is for electric vehicles), with evidence of the company's position on X versus competitors and X shown to be where the remaining improvement curve lies.
• Source: Columbia lecture 2010 — the BYD discussion ("the heart and soul of the electric vehicle age... is the battery"). (teaching-derived)

Owner mindset

L15. Closed-exchange owner test

• Asks: If the Tokyo Stock Exchange closed for ten years, would you be content owning this business at today's price — does the expected return clear the hurdle with the exit multiple no higher than today's and no exit assumed?
• Good: An affirmative case from operating figures alone: IRR from dividends plus per-share earnings growth ≥ the hurdle, with multiple expansion treated strictly as a free option. If the case depends on someone paying a higher multiple later, the item fails.
• Source: Columbia lecture 2010 — ownership held as an owner would hold it; Peking Guanghua talk, Nov 2019 (tr. LONGRIVER) — "If you don't participate in speculation and stick strictly to investing in what you understand, then you won't lose money." (quote-backed)

L16. The inheritance test

• Asks: If you suddenly inherited 100% of this company, what would you do with it — can you write the one-page owner's brief from filings: where cash comes in, where it must go out, who is attacking you, and what you would do with the surplus?
• Good: The brief exists with every line sourced — cash sources by segment, non-discretionary uses (capex, debt service), named competitors, and a defensible use for the free cash — before any discussion of the stock.
• Source: Columbia lecture 2010 — "imagine a distant relative passes away and you find out that you have inherited 100% of a business they owned"; Peking Guanghua 2019 (tr. LONGRIVER) — "a long-lost uncle who died suddenly, leaving the entire company to them." (quote-backed)

L17. Whole-company framing, not per-share paper

• Asks: Treating the market cap as the price of the entire business, would a rational private buyer pay it for the whole company's earnings power and assets?
• Good: The figure table states whole-company values and conclusions in whole-business terms ("paying X for a business that earns Y"); per-share numbers appear only as derived.
• Source: Columbia lecture 2010 — "A stock is not a piece of paper, it is a piece of ownership in a company"; Peking 2015 (tr. Rhodes) — "a certificate bestowing fractional ownership of a company." (quote-backed)

L18. Dollar at fifty cents

• Asks: How far below a conservative, worst-case-anchored intrinsic value is today's price — is intrinsic value at least roughly twice the price paid, grounded in tangible assets and current earnings power rather than projections?
• Good: A stated intrinsic value with visible arithmetic at ≥ ~2x market cap, plus a sensitivity showing you lose little even if the key prediction is 10–20% wrong.
• Source: Columbia lecture 2010 — "Buying a dollar at 50 cents. So if things turn against you, you will be okay"; Peking 2015 (tr. Rhodes) — the price "should always be far, far below the company's intrinsic value." (quote-backed)

L19. Permanent-loss mechanisms, downside first

• Asks: Before any upside case, what are the specific mechanisms by which capital here could be permanently lost (leverage, product obsolescence, fraud, regulatory kill, customer concentration) — and what filed evidence bounds each one?
• Good: An enumerated failure-mode list preceding the bull case, each mechanism shown absent or bounded, plus pre-registered impairment criteria distinguishing a real mistake (heading toward zero) from a quotational 50% drop; price volatility appears nowhere on the list.
• Source: Columbia lecture 2010 — "you look at the downside before looking at the upside"; Peking 2015 (tr. Rhodes) — "it's the risk of a permanent loss of capital." (quote-backed)

L20. Survives the downturn without new money

• Asks: Can the company get through a severe multi-year downturn without issuing equity or refinancing at the market's mercy?
• Good: A trough scenario from reconciled figures — debt maturity schedule, fixed charges, covenant headroom — covered by cash on hand plus internally generated funds; no dependence on continuous capital-market access.
• Source: Peking 2015 (downturn behavior) plus his repeated rule against leverage, applied at the company level (paraphrase: leverage converts temporary declines into permanent loss). (teaching-derived)

Capital allocation

L21. Retained-earnings audit — trace every yen

• Asks: Over the last 5–10 fiscal years, where did retained earnings actually go — reinvestment, acquisitions, buybacks, dividends, or the cash and securities pile — and did each retained unit create at least a unit of per-share intrinsic value?
• Good: A reconciled table of cumulative net income against each destination, with Δ(book value per share + cumulative dividends per share) / cumulative retained EPS ≥ 1; buybacks executed when cheap, no serial empire-building, and any hoarding explained by the business rather than habit.
• Source: Peking talks 2015/2019 — rationality with retained capital as the defining management virtue; the retained-to-value test derived from the return-convergence frame (2019/2021, paraphrase). (teaching-derived)

L22. Incremental ROIC gate

• Asks: On an unleveraged, pre-tax basis, what return has the capital actually deployed in this business earned — and what has the capital added in the last five years earned?
• Good: Deployed capital and unleveraged returns shown across years (Timberland: ~$200M deployed earning ~$100M), plus ΔNOPAT over ~5 years against cumulative retained-plus-new capital clearly above cost of capital — not a high legacy ROIC on old capital (his "mathematics get interesting" bar: 40–100%).
• Source: Columbia 2006 notes (paraphrase: 50–100% ROIC and the mathematics get interesting very quickly); Greenwald conversation 2021, derived from the return-convergence teaching. (teaching-derived)

L23. Management inside its own circle

• Asks: Has management stayed within the business it demonstrably understands, or has retained cash leaked into unrelated ventures, trophy assets, or serial acquisitions?
• Good: Segment history shows focus over ten years; goodwill modest relative to equity; no pattern of buying outside the core followed by write-downs; new ventures adjacent to proven competence.
• Source: Peking talks 2015/2019 (paraphrase: an ability without a boundary is not an ability — for a company as for an investor). (teaching-derived)

L24. Cash returned when the runway ends

• Asks: In years when high-return reinvestment opportunities fell short, did cash actually come back to shareholders — or did it pile up at deposit rates or get spent anyway?
• Good: Payout (dividends + net buybacks) rises as growth capex falls in the filing record; no decade-long cash hoard earning near zero — the classic reason cheap Asian stocks stay cheap.
• Source: 2020s interviews and book themes (paraphrase: owning a fraction of the business only pays if the value can eventually reach the owner — as high-return reinvestment or as cash out). (teaching-derived)

L25. Management as fiduciaries of shareholder capital

• Asks: Is there concrete filed evidence that management treats shareholder capital as a trust — moderate self-pay, candid discussion of mistakes, per-share rather than empire metrics, dilution avoided?
• Good: Compensation small relative to profits and stake; share count flat or falling over ten years; commentary that owns errors and reasons in per-share terms; no serial capital raises funding expansion that returns less than it costs.
• Source: Peking 2015 (tr. Rhodes) — "treat every dollar of client money as though it were the fruit of your own parents' labour," transposed to management's treatment of shareholder capital. (teaching-derived)

Structure & Asia

L26. Structure check — cross-holdings, parents, and minorities

• Asks: What does the ownership and affiliate structure hide? Quantify policy cross-shareholdings from the yuho, identify any listed parent or controlling shareholder, and check whether the company is responding to the TSE's capital-efficiency reform pressure or ignoring it.
• Good: Policy holdings quantified as a percentage of net assets and demonstrably shrinking; no listed parent with conflicting interests — or a documented history of fair dealing; a visible, credible response to TSE reforms (stated ROE/PBR targets with actions), not boilerplate. Structural subordination of minorities is a failing answer even at a low price.
• Source: Core Asia-adaptation theme — value investing adapted to local institutions and ownership structures (Peking 2015/2019); the Japan-specific application is this profile's extension, not a direct quote. (teaching-derived)

L27. Hidden-asset sum versus price

• Asks: Does the figure table re-mark each balance-sheet item to realizable value — net cash against market cap, listed cross-holdings at current market price, real estate against recent comparable transactions — and sum them against the price paid?
• Good: A line-by-line table like his Korean example: 60M market cap versus 70M net cash, a stake booked at 30M worth ~80M at market, a hotel at 2–3x decade-old book — totaling ~320M of assets plus 30M of annual earnings for a 60M price.
• Source: Columbia 2006 notes (paraphrase: the Korean hidden-assets walkthrough — "you get 320M in assets that you are paying 60M for, earning 30M annually"). (teaching-derived)

L28. Sister-company look-through

• Asks: Did the study open the filings of the major cross-held affiliates one level down and check whether they show the same profile — their own net cash, hidden assets, and businesses invisible from the parent's book value?
• Good: Each significant holding examined at its own filings level (his department-store stake itself traded around cash and investments, owned real estate, and was the second-largest cable operator), with findings fed back into the parent's asset sum.
• Source: Columbia 2006 notes (paraphrase: "The department store has exactly the same profile"). (teaching-derived)

L29. Local-investor realization path

• Asks: For a stock priced by a local investor base, how do local investors see it and what would make them buy — what is the concrete path from hidden value to realized price?
• Good: Documented local ownership and sentiment, precedent re-ratings of comparable local names (the sister department store went from 22 to 100), and named realization routes (dividends, buybacks, governance or market reform) — not "value will out" by assertion.
• Source: Columbia 2006 notes (paraphrase: local investors need to be buying it for the price to go up). (teaching-derived)

L30. Controller alignment and the ten-year minority record

• Asks: Document insider/family ownership and any split between economic interest and votes — and over ten years of filings, has the controller ever moved value away from minorities (discounted insider placements, unfair asset shuffles, low-ball buyout attempts)?
• Good: An ownership table (Timberland: family owned 40% controlling 98% of the vote) plus a clean ten-year review resolving the milking-versus-building question from evidence — related-party flows, dividend history, litigation record, how the family made its money; one bad episode is disqualifying evidence about character.
• Source: Columbia 2006 notes (paraphrase: find out whether management is actually milking the business); 2020s interviews and book themes on controller treatment of outside shareholders (paraphrase). (teaching-derived)

L31. Related-party extraction check

• Asks: What is the total magnitude of related-party transactions (sales, purchases, loans, guarantees, balances) relative to revenue and profit, and are terms demonstrably arm's length?
• Good: RPT disclosures summed from the notes; small relative to the business (low single-digit % of revenue/profit), stable, with stated arm's-length terms; large or growing flows to controller-owned entities fail.
• Source: 2020s interviews and book themes (paraphrase: where governance is immature, establish that profits are not being channeled to the controller through side dealings). (teaching-derived)

L32. Insider-ownership economics

• Asks: Do the people running this company get rich the same way you will — through the value of shares they own — rather than through salary, fees, option grants, or private side vehicles?
• Good: Insider stake from the shareholder register is large, bought or long-held (not just granted), worth many multiples of annual compensation; no disclosed private vehicles competing with or supplying the listed company.
• Source: 2020s interviews including the Greenwald conversation 2021 (paraphrase: owner-operators whose wealth compounds through the same shares minorities hold). (teaching-derived)

L33. The listed entity's claim on the cash

• Asks: Does the security you would actually own carry an enforceable claim on the operating cash flows — do subsidiary dividends reach the listed parent, and for VIE/holdco structures, does economic interest match legal ownership?
• Good: A structure trace from filings (operating subsidiaries → intermediate holdcos → listed entity) with parent-only accounts showing dividends actually received; VIE arrangements documented with evidence of real cash movement, or the gap logged as a thesis risk.
• Source: Book and 2020s interviews on Chinese listing structures (paraphrase: the certificate you can buy must actually own the business that earns the cash). (teaching-derived)

L34. Legal/license barrier to entry

• Asks: Does the company hold licenses, concessions, or territorial rights that legally block entry — and what do filings say about their terms, duration, and renewal risk?
• Good: The actual license/concession terms cited from filings (scope, expiry, renewal history, pricing constraints); a "license moat" with a near-term politically exposed renewal is a risk, not a moat.
• Source: Peking Guanghua talk, Nov 2019 (tr. LONGRIVER) — the cable-TV example: "a local monopoly; if a company has a licence for a territory, other companies cannot enter." (quote-backed)

Value trajectory

L35. Growing value or melting ice cube

• Asks: Is intrinsic value compounding or eroding — does the 10-year record show the business getting stronger every year, and can you explain why it makes more and more money compared to others?
• Good: Per-share owner value (book value plus cumulative dividends, or earnings power) compounding over the decade with a causal explanation of the advantage, not just a favorable trend line; a statically cheap business whose value is melting is a pass, not a buy.
• Source: Peking 2019 talk and Civilization, Modernization, Value Investing and China (2020) — intrinsic value must keep growing; Columbia 2006 notes (paraphrase: identify businesses getting stronger and stronger every year). (teaching-derived)

L36. Long-run ROIC convergence

• Asks: What is the ten-year return on invested capital (and ROE) series for this company — the number a long holding period will make your return approximate?
• Good: A 10-year ROIC/ROE series computed from reconciled figures, comfortably above the hurdle and stable or rising; if that number doesn't clear the hurdle by itself, a long hold cannot.
• Source: Greenwald conversation 2021 — "If you own them through the ups and downs, your return roughly approximates the actual business return to the capital we invested." (quote-backed)

L37. Reinvestment runway

• Asks: What filed evidence shows this company can keep reinvesting a meaningful share of earnings at high returns for another decade — market headroom, capacity pipeline, penetration data?
• Good: Reinvestment-rate history from the cash-flow statement plus concrete runway evidence from filings (stated expansion plans, addressable-market disclosures, recent-cohort unit economics still holding).
• Source: Peking Guanghua 2019 (tr. LONGRIVER), the sustained-compounding frame (paraphrase: even 6–7% a year sustained over very long periods produces enormous cumulative results). (teaching-derived)

L38. Profitability under documented attack

• Asks: Competitors have attacked this company's profits — filings will show when. Did margins and market position hold through the attack?
• Good: A named episode of competitive entry or price war visible in the record, with the figure table showing margin/share resilience through it; untested profitability is marked as such, not presumed durable.
• Source: Peking Guanghua talk, Nov 2019 (tr. LONGRIVER) — "Profitable companies will attract competitors who will try to snatch market share and profits." (quote-backed)

L39. Growth quality decomposition

• Asks: Decomposing ten years of revenue and profit growth, how much is repeatable organic volume/price versus acquisitions, one-off gains, subsidies, or a cyclical spike?
• Good: A reconciled decomposition where the dominant driver is organic and tied to a durable demand trend; growth bought with serial M&A or riding a commodity peak fails.
• Source: Peking 2015, derived from the continuous, cumulative, compounding-growth frame (paraphrase). (teaching-derived)

L40. Best-versus-worst industry test

• Asks: Did the study compare the industry's best and worst players over time — and if the worst players still earn reasonable returns relative to the great ones, credit the industry rather than the company?
• Good: A multi-year return table for the leader and the laggard, with an explicit conclusion about whether the subject's results come from industry economics or company-specific advantage.
• Source: Columbia lecture 2010 — "Find the best of the players in the industry and the worst players. And see how they perform over time." (quote-backed)

L41. Metric trajectory at five-year intervals

• Asks: Does the figure table show return on capital and margins at roughly five-year intervals over the longest available history, revealing whether the economics are improving or decaying?
• Good: A GM-style series (return on capital 46% in the 1960s, 28% in the 1970s, 9% in the 1980s, 6% in the 1990s) built for the subject company, with the direction of travel addressed rather than a single-year snapshot.
• Source: Columbia lecture 2010 — "look at General Motors from the early days, look every 5 years and see how the performance metrics change." (quote-backed)

L42. Return decomposition — earnings first, re-rating as bonus

• Asks: Decompose the expected return into entry multiple, earnings growth, and exit multiple — does the base case work on earnings alone, without multiple expansion?
• Good: An explicit "it adds up" table in the Timberland pattern — bought at ~5x with earnings compounding ~30% a year; the 5x-to-15x re-rating was the bonus, not the requirement.
• Source: Columbia 2006 notes (paraphrase: stock up 700%, propelled by earnings — "It adds up"). (teaching-derived)

L43. Switching costs and the winner-take-all line

• Asks: What would leaving actually cost customers — and where does the company sit relative to the point where its market tips to winner-take-all?
• Good: Evidence of Bloomberg-style lock-in (daily-workflow dependence, behavioral hooks, pricing power once users have no real choice), plus an explicit judgment on whether the line to monopoly economics has been crossed, is approaching, or does not exist in this industry.
• Source: Columbia 2006 notes (paraphrase: at a certain point the switching costs become so high the winner takes all). (teaching-derived)

L44. Demonstrated adaptation over projected plans

• Asks: Where the company is changing fast or information is imperfect, does the thesis rest on what management has already accomplished — repeated, documented adaptation — rather than projected plans?
• Good: A record of surprises survived and pivots executed reconstructed from filings (his BYD case: $300K of pre-IPO funding to $6–7B revenue and $500M net profit), with the thesis explicitly resting on that demonstrated ability.
• Source: Columbia lecture 2010 — "how do you look at it as an investor with imperfect information? Well I suggest you look at what he has accomplished." (quote-backed)

L45. Twenty-to-forty-year mortality check

• Asks: What could make this company disappear over 20–40 years — technology shift, capital cycle, GM-style decay — rather than assuming a perpetuity by default?
• Good: Named extinction scenarios with the company's specific defenses assessed, acknowledging the base rate that two-thirds of top-50 companies are gone on that horizon.
• Source: Columbia lecture 2010 — "Look at the top 50 companies in America every 10 years. By the time 20-40 years go by, 2/3rds of them will be gone." (quote-backed)

Accounting credibility

L46. Trust in the accounting and the people

• Asks: Can the reported numbers and the people running the business be trusted? Check auditor continuity, restatement history, segment-disclosure quality, and the controllers' past conduct.
• Good: A stated, evidence-based conclusion: no auditor changes or restatements without a clean explanation, segments granular enough to expose the real economics rather than a consolidated blob, with extra depth where the jurisdiction or situation raises fraud base rates. Opacity here caps the verdict at watch, whatever the price.
• Source: Columbia lecture 2010, answering the China-fraud question — "whether you can trust the accounting and people running the business... spend a lot of time looking at these factors, especially if you are investing for the long haul." (quote-backed)

L47. What is in the book

• Asks: For any book-value-based cheapness claim — what is in the book, and how much is the book? Decompose book value into cash, working capital, and fixed assets, excluding goodwill.
• Good: Each remaining component tested for realizable worth (receivables collectible, inventory saleable, property at current rather than historical value), Timberland-style: a clean book of tangible liquid assets plus $100M in real estate.
• Source: Columbia 2006 notes (paraphrase: if you see a low P/B ratio, ask what is in the book; don't count on goodwill). (teaching-derived)

L48. Profits convert to owner cash

• Asks: Over ten years, does cumulative operating cash flow minus capex roughly match cumulative net income — do reported profits become cash an owner could actually take?
• Good: Cumulative (CFO − capex) / cumulative net income around 0.8+ with the gap explained by disclosed growth capex; receivables and inventory not persistently outgrowing revenue — the uncle who inherited the company would find real cash, not accruals.
• Source: Peking Guanghua 2019, derived from the inheritance exercise (paraphrase: an owner lives on the business's cash, not its reported accruals). (teaching-derived)

L49. Figures reconcile or become unknowns

• Asks: Do the load-bearing figures reconcile across statements — profit versus operating cash versus tax actually paid versus sum-of-segments — with every unexplained gap logged as an unknown rather than assumed away?
• Good: The figure table reconciles; each discrepancy is resolved from footnotes or explicitly carried as a thesis risk; no load-bearing number rests on "probably fine."
• Source: Peking University talk, Oct 2015 (tr. Rhodes) — "The most fundamental requirement of this profession is complete and utter intellectual honesty." (quote-backed)

Process

L50. Four-part long-hold gate

• Asks: Does the company pass all four gates — superior economics within its industry, an industry with superior economics, good management, and the right price?
• Good: Each gate answered separately with evidence from the figure table and record; failure on any one gate downgrades the idea from long-hold compounder to trade or pass.
• Source: Columbia lecture 2010 — "a business with superior economics in an industry with superior economics with good management and you get them at the right price." (quote-backed)

L51. Every court document read

• Asks: If the company has lawsuits, regulatory actions, or shareholder claims, did the study obtain and read the primary documents and classify what they actually allege — fraud versus technical or benign?
• Good: Every case listed with docket-level sourcing and a verdict in the Timberland pattern: the suits stemmed from missed guidance and the family going silent — annoying, but "they were not crooks."
• Source: Columbia 2006 notes (paraphrase: download every court document and lawsuit; read everything). (teaching-derived)

L52. Management judged by the trail of actions

• Asks: Is the management assessment built from the documented trail of what they have done across situations — capital decisions, crisis behavior, promises versus delivery — rather than from meetings or investor-relations narrative?
• Good: A reconstructed multi-year record of decisions from filings and contemporaneous sources ("most business owners leave a trail for you to follow"); meeting-based impressions clearly subordinated to the record of actions.
• Source: Columbia lecture 2010 — "You can know about management teams without meeting them... Obviously, actions speak louder. You want to see what they have done." (quote-backed)

L53. Third-party character evidence on the controllers [PRACTITIONER]

• Asks: Did the study gather evidence on the controlling owners' character from public records outside the company's own disclosures — press archives, public litigation records, and proxy/registry board-network mapping (which other boards the controllers sit on, and what happened at those companies)?
• Good: At least one independent, public-record line of character evidence documented and reconciled with the filings picture — a board network mapped from registries and proxies, a press-archive sweep, a litigation-record search. Any character judgment beyond what public records support belongs to the practitioner, not the agent.
• Source: Columbia 2006 notes (paraphrase: go to their community, meet their friends and neighbors, learn what these business people have done — the Timberland community visit is historical color; the operative channels here are public records). (teaching-derived)

L54. Ground-level verification of the earnings story

• Asks: Is the demand or margin story corroborated by public evidence outside the company's own reporting — listed customers' filings, industry shipment statistics, archived product or pricing evidence — rather than resting on the reported numbers alone?
• Good: At least one independent public corroboration whose findings reconcile with the reported trend — a listed customer's filed order or capex trend, an industry shipment series, archived product/pricing evidence. A demand story with no public corroboration routes to the unknowns ledger, not to a pass.
• Source: Columbia 2006 notes (paraphrase: Lu visited all the stores to see how margins improved — the store visits are Source color from the Timberland case; the pass bar here is public corroboration). (teaching-derived)

L55. Work finished before betting

• Asks: Was every research task identified during the study completed before the buy conclusion — or are there questions you liked-but-didn't-finish, betting on probabilities instead of analysis?
• Good: A closed loop: each open question resolved with cited evidence or explicitly moved to the unknowns column with its price consequence stated; no "probably fine" left standing on a load-bearing point.
• Source: Columbia 2006 notes (paraphrase: you make a mistake when you have not finished your work but like it enough). (teaching-derived)

L56. Historical analog studied

• Asks: Is the company framed against a historical analog — a mature company in the same industry whose long-run record, sampled at five-year intervals, shows how these economics evolve?
• Good: A named analog with its metric history summarized and applied (his BYD prescription: analyze GM's performance every 5 years for 100 years), concluding which phase of the analog's curve the subject sits in and what that implies.
• Source: Columbia lecture 2010 — "the study of history, of all the great corporations will give you a good insight in seeing what will happen with BYD." (quote-backed)

L57. Competitors studied directly

• Asks: Did the study examine the main competitors' filings and economics directly before concluding this company is the best vehicle for the thesis?
• Good: A competitor-by-competitor comparison from primary sources — not the subject company's own narrative about its rivals.
• Source: Columbia lecture 2010 — "Well because we also studied all those other guys. We will see when the winner emerges whether we are right or wrong." (quote-backed)

L58. Continuous-monitoring plan

• Asks: What will you keep studying after purchase — the trends and metrics to observe continuously, and what observed changes would force a thesis revisit?
• Good: A short list of trackable indicators (market share, margins, industry line-crossing, controller behavior) with stated thresholds — nothing is constant, so the analyst's job is to study the business all the time.
• Source: Columbia 2006 notes (paraphrase: study that business all the time and observe those trends); reinforced by Columbia 2010 ("The game of investment is really continuous learning"). (teaching-derived)

L59. Concentration warranted by the work

• Asks: Does the depth of verification documented justify a concentrated position — and does the study end with an explicit bet-big / small / pass conclusion rather than a token weighting?
• Good: The study links the work done to a sizing conclusion in Li Lu's spirit ($200 of a $900 portfolio, not the crowd's 50 basis points), or states plainly why conviction is insufficient to own the stock at all.
• Source: Columbia 2006 notes (paraphrase: think of how much work you did — you need concentration, a $200 position). (teaching-derived)

Verdict guidance

The knowledge bar gates everything. If, after the work, you cannot honestly claim to understand this business's economics and its next ten years better than most of the people who own it (L1–L14), the verdict is too-hard — intellectual honesty about the boundary of your competence is the discipline, and no price rescues a business you cannot predict (L1, L6). Structural failures behave the same way: minority subordination or a broken claim on the cash (L26, L30–L33) and accounting you cannot trust (L46–L49) end the analysis before valuation begins.

For businesses that clear the bar, the question is decade-scale: will a part-owner of this company be meaningfully wealthier in 10–20 years (L15, L35–L45), and is management a rational allocator of every retained yen (L21–L25)? Buy-below-¥X is set at a margin of safety to a conservative owner-value estimate built from the figure table — a dollar at fifty cents (L18), with no re-rating assumed (L15, L42). Because this lens invests with concentrated conviction (L59), the base rate is that most companies land at pass or watch; a buy requires the four-part gate (L50) answered well and a price that leaves room to be wrong.

Archive-depth rule. When the archived sources are shallower than an item requires (fewer years, missing competitor/customer filings, absent disclosures), the item is answered data-insufficient with the missing source named — never estimated from general knowledge without an explicit [general-knowledge: not-from-archive] flag. Load-bearing items stuck at data-insufficient push the verdict toward too-hard or watch, not toward a guess.

A too-hard verdict names the specific item ID(s) that triggered it. Item-level answers use pass / fail / data-insufficient / practitioner-pending; items after a too-hard early exit are marked not-evaluated, never silently skipped.