Kobayashi Pharmaceutical (4967): a fortress that failed, at a full price

Let me tell you what this company is before I tell you what I think of it, because that is the only honest order.

It makes small things people reach for without thinking. Throat lozenges, deodorizers, disposable heat pads, over-the-counter remedies — niche products where Kobayashi is the name on the shelf E1E2. Healthcare OTC did ¥58.3bn last year, household goods ¥50.8bn, the heat pads ¥6.2bn F70F71E2. That is a business a shopkeeper understands. Somebody's throat hurts, they buy the lozenge, and next winter they buy it again. You do not need a view on semiconductors or interest rates to own it. It sits inside my circle, and I can write down a rough guess at what it earns ten years out. That clears the first gate, which is the gate most things never clear.

Now the trouble. In 2024 one supplement line — red yeast rice, 紅麹 — was tied to health harm, and the company had to recall it and pay E3E14. What I care about, as an owner, is whether the customer kept buying everything else. Revenue tells me she mostly did: ¥165.6bn in the crisis year, ¥165.7bn the year after, against ¥173.5bn before F4F5F3. The register kept singing. That is the American Express salad-oil pattern — the scandal was ugly, but the franchise held. The mail-order channel shrank hard F72 and is being shut down entirely E13, but that was a small sliver.

So why did the bottom line fall through the floor? Net income went from ~¥20bn a year F17 to ¥10.1bn F18 to ¥3.7bn F19. Two separate things happened, and you must not confuse them. First, real operating profit fell about 40% — they suspended advertising and wound down mail-order, and operating income dropped from ¥24.9bn to ¥14.9bn F8F9. That is a genuine wound, but a healing one: they turned the television ads back on in July 2025 E13. Second, and larger, was a ¥14,645M writedown of two factories F60E10. That is an accounting entry, not a check that cleared. It is the tooth fairy in reverse — a one-time mark against plant, sitting below the operating line, and I do not let a single non-cash impairment convince me the earning power is gone. Strip it out and normalized profit looks a great deal more like the pre-crisis ~¥20bn D4 than the reported ¥3.7bn.

The balance sheet is the part I admire most. This is a fortress. Essentially no debt — ¥678M of lease obligations and nothing else F52. ¥62.3bn of cash F42, another ¥8.3bn in securities F43, and ¥31.8bn in investment securities beyond that F43F47. Equity is 76% of assets F29. Net cash of interest-bearing debt is ¥61.6bn D3. A company like this can survive an indefinite wait; the tide can go out for years and it will not be caught swimming naked. And through the whole ordeal they raised the dividend — ¥104 last year, guided to ¥106 E12. A board that lifts the payout while the earnings statement is on fire is a board telling owners it believes the fire is temporary.

Management earns a passing mark, which I do not hand out freely. They wrote a plain three-pillar recovery plan admitting the governance failure E4, moved from the founding family to a professional president E5, turned over four directors at the AGM E9, stood up a quality-assurance function and a compensation-response office E7, and are moving to an audit-committee structure with real outside directors — including a KPMG partner and a former Supreme Court official E6. Pay is modest and tied to results: ¥160M for the executive directors, roughly a third of it performance-linked F90E8. That is the confession-time behavior I look for, not boilerplate.

Now the only question left: is ¥5,505 a bargain? Here I have to be a spoilsport. Market cap is ¥409bn D1. Against the honest, normalized ~¥20bn of pre-crisis earning power D4, that is roughly twenty times — a fair price for a fine business, not a bargain. Against this year's guided ¥10bn F82 it is forty times, which is silly, but that guidance still carries the crisis hangover, so I set it aside. Owner earnings — reported profit plus the depreciation and impairment addbacks, less the real maintenance capex — land in the same neighborhood as that ~¥20bn once you normalize; last year's ¥25.6bn operating cash flow F41 is flattered by the non-cash writeback and cannot be taken at face value. The margin of safety I want is not there. The figures do not hit me over the head with a baseball bat. A quarter of the market cap is cash and securities F42F43F47, which softens the downside, but you are still paying a full twenty times normalized earnings for the operating business, and the ¥104 dividend on a ¥5,505 stamp is under 2% — no bargain yield.

So this is a wonderful little business, priced fairly, sitting on a fortress balance sheet, run by people cleaning up an honest mess. It is not cheap enough. That is a watch, not a pass — I would very much like to own it, at a price. Around ¥3,400 the normalized earnings yield reaches the double digits I require with a real margin of safety, and the cash pile does the rest of the work. Below that, I would swing.

Invert. The interesting question here is not how Kobayashi wins — a shelf of little category-leading OTC brands and household oddments E1E2 more or less wins by sitting there — but how it already lost. It poisoned its own customers. The 紅麹 red-yeast-rice matter E3 killed people. Start there, because a company selling things you swallow lives and dies on one asset — the trust that what is in the bottle is what the label says. Munger's rule: reputation and integrity can be lost in a heartbeat. This one was.

So the whole case reduces to a character-and-incentives test, and I will read it as one.

First, was this bad luck or bad wiring? A company whose single indispensable input is safety, that shipped a supplement that harmed customers, that then took until September 2024 to publish a three-pillar plan E4 — that is not weather. That is a control environment that let the most gameable thing in the business (product safety) go ungoverned. The prescribed antidote — Munger's cancer-surgery formula — is to cut the folly out and return to the wonderful core. They wound down the whole owned mail-order channel E13, deconsolidated three subsidiaries E15, suspended advertising and restarted only in July 2025 E13. That is closer to cutting than to denial. Credit where due.

Second, the incentives — the part that actually tells you whether it recurs. Read what changed, not what was said. The founding-family president stepped aside for a professional manager; the Kobayashi family member was moved to run the compensation-response HQ E5 — the man associated with the harm now owns the cleanup, not the strategy. Six outside directors, a KPMG partner and an ex-Supreme-Court lawyer among the auditors, a move to an audit-committee structure E6. Four directors turned over at the 2025 AGM E9. New Quality-and-Safety-Assurance and Compensation-Response HQ functions E7. This is the rare thing: a board that visibly overruled the captain and threw people off the plane. On M83 and M60 — who can tell the CEO no, and does bad news get faced — this scores better than most Japanese boards I would ever look at.

Now the incentive I do not like. Pay is re-tied to consolidated EBITDA margin and EPS E8. EBITDA is bullshit earnings, and I will not pretend otherwise. In a company that just took a ¥14,645M impairment F60, keying the bonus to a metric that adds depreciation and impairment back is precisely the metric under which this year's disaster never happened. That is the wrong lesson bolted onto the right reforms. It is not disqualifying — the fixed/performance split is modest, ¥160M total across five directors F90 — but it is a demerit, and it is the seam where the next mistake hides.

Third, and this is the part the governance applause tour skips: while the house was on fire, they misallocated capital. A ¥14,645M impairment F60 on a brand-new Sendai plant and the Thailand plant — assets so new they were still partly in 建設仮勘定, ¥17,656M of construction-in-progress still sitting on the balance sheet F50. You do not impair a new plant because customers got sick. You impair it because you built capacity for volumes that are not coming. So the same management era that failed at quality also failed at allocation — it built plants for a business it was simultaneously poisoning. M42, retained-earnings productivity: serial impairment is the tuition bill of bad allocation, and here the bill and the scandal arrived together. That is a lollapalooza on the wrong side — quality failure plus allocation failure, one culture producing both.

Set against that: the balance sheet is a fortress, and fortresses buy time to be wrong. Essentially no interest-bearing debt — ¥678M of lease obligations F52 against ¥62,314M of cash F42 and ¥8,300M of securities F43. Net cash of debt ¥61,636M D3. Equity ratio 76.3% F29. On M17, the fifty-percent-quotation test, and M2, permanent loss — there is no covenant, no maturity wall, no forced-seller trigger anywhere. The recall provision is down to ¥2,176M F58. The company cannot be killed by its balance sheet. It can only be killed by doing it again. Note one caution: after ALL liabilities, liquid net cash is only ¥6,294M D2 — the fortress is in the equity ratio, not in a giant spare cash pile, and the compensation liability E14 is not fully fixed in the filings.

The moat is real but I will not overpay for it. These are habit-and-association mini-monopolies (M27, M28) — you buy the throat drop you have always bought. But this scandal is exactly the event that poisons a habit-and-association moat: the association was "the reliable little company," and that is the association that broke. Recovery is "回復基調" E1, on a recovery trend — a trend, not a fact.

Price. The stamp is ¥5,505, market cap ¥409,235M D1. The whole question is what "normal" earns. Trailing EPS ¥49 F32 is crisis-plus-impairment and meaningless. Pre-crisis net income averaged ¥20,025M D4; FY2026 guidance is ¥10,000M F82, with operating income guided DOWN 16% F81 — so management itself says the recovery is not one year away. At ¥409bn against ¥20bn of maybe-someday normalized earnings, you pay ~20x D1D4 for a business that just demonstrated it can inflict a permanent-loss event on itself, and against FY2026 guidance you pay ~41x D1F82. On M35 and M44, the opportunity-cost and pari-mutuel tests: the quality is fully paid for, and the horse just fell down once. A payout ratio of 211% F35 — paying out more than twice earnings to hold the dividend E12 through a loss year — is admirable to a shareholder and slightly worrying to an analyst; it is not sustainable arithmetic, it is a signal, and I take it as one.

Verdict: watch, not buy, not too-hard. The reforms are real enough that this is no longer purely the too-hard pile — the board did the hard, credible things. But I do not pay 20x normalized for a company two years off a fatal quality failure, on trust that is a trend and not yet a track record. Trust lost in a heartbeat is earned back over years of clean filings, not one plan document. Let it prove the reforms with two or three uneventful years, and let the price come to me.

(First-person interpretation in Pabrai's Dhandho voice — reasoning reconstructed from the published framework; no invented quotations. His real checklist is undisclosed.)

Here's a setup that should be right in my wheelhouse. A boring, category-leading consumer-health franchise — a company that sells throat drops and heat pads and deodorizers E1E2 — gets clobbered by a scandal on one supplement line, the 紅麹 red-yeast-rice product E3. Net income falls off a cliff, from a steady ~¥20bn run-rate D4 to ¥3.7bn F19, and the stock is on the floor. Diversified brand portfolio, essentially debt-free F52, a pile of cash F42. This is exactly where I go hunting: a durable business, a nameable fear, and a market that might be confusing a wide range of outcomes with permanent loss. So let me count the downside first, the way I always do. Heads I win; tails I don't lose much — that's the only question that matters, and I answer it before I let myself dream about the upside.

Start with P1, and remember my own scar tissue: an equity cushion is not an asset floor. The balance sheet shows ¥211bn of net assets F22 and a 76% equity ratio F29, and every promoter of this stock will wave that at me. I don't care about the book equity line. I care about marked-down liquid and hard assets minus all liabilities. Cash F42 ¥62.3bn plus securities F43 ¥8.3bn is ¥70.6bn; total liabilities F25 are ¥64.3bn. So the truly liquid net-cash floor after paying off everyone D2 is about ¥6.3bn — roughly ¥85 per share on 74.3m shares [F95−F96]. Against a ¥5,505 price and a ¥409bn market cap D1, that "floor" is a rounding error. Add the long-term investment securities F47 at ¥31.8bn and haircut them, and you buy maybe another few hundred yen per share. This is the thing I have to be honest about: there is no hard asset floor here. Kobayashi is not Frontline with ships you can sell, and it is not a net-net. The downside protection, if it exists, has to come from earning power, not from assets. That already tells me the margin of safety must be earned on the P&L, and it moves several of my asset-floor items (P4, P5) to fail or not-applicable.

So the whole case rests on normalized earning power surviving the scandal — P15, P18, P56. And on the operating numbers, the news is genuinely good, which is what makes this interesting rather than a quick pass. Revenue barely moved: ¥165.6bn F4 to ¥165.7bn F5, within a whisker of the pre-crisis ¥173bn F3. Healthcare and household goods are on a recovery trend E1, TV advertising restarted in full in July 2025 E13, and the money-losing mail-order line F72 — down 39% — is being deliberately wound down E13. The ¥14,645M impairment F60 on the Sendai and Thailand plants is a non-cash special loss below the operating line; it's what collapsed net income to ¥3.7bn F19, but operating cash flow was actually ¥25.6bn F41. The franchise is intact. Pre-crisis this business earned ~¥20bn D4 on ~10% ROE F26; management guides FY2026 to ¥10bn F82. Normalized earning power is somewhere in that ¥10–20bn band, and the mid-cycle number is very much alive.

Now the valuation, on the back of an envelope — no Excel (P52). At ¥409bn market cap D1: on the pre-crisis ¥20bn D4 that's ~20x; on FY2026 guidance ¥10bn F82 that's ~41x; on trailing crisis EPS ¥49 F32 it's ~112x. None of these is a fifty-cent dollar (P53). Even if I hand this quality business a generous 20x multiple on fully-normalized ¥20bn earnings, I get ~¥400bn — right at today's price. Add back excess cash and investment securities of maybe ¥60–90bn and I might argue a conservative IV around ¥5,500–6,500 per share. So at ¥5,505 I'm paying roughly full freight for a normalized recovery that hasn't happened yet, with a live compensation tail hanging over it. That is not the wide, obvious discount I require. My rule is price at ~50% of conservative IV, and here I'm at ~90–100% of it.

Then the item that decides everything: the crisis-compensation tail E14, and P13 — what is the market actually afraid of? The filings record a recall provision F58 of just ¥2,176M — and it fell from ¥3,970M. But the tanshin says compensation to health-damaged customers and affected companies is ongoing E14, and the ultimate total is explicitly not fixed in the filings. A ¥2.2bn provision against a product implicated in deaths and injuries is not a number I can underwrite. This is the un-bounded liability that ruins the Dhandho asymmetry. If the tail is ¥2bn, the thesis is fine and cheap-ish. If it's ¥20bn+, it eats my entire ¥6.3bn liquid cushion D2 and several years of normalized earnings, and the "low-risk" label was a lie. I genuinely cannot bound it from the ledger. That is the definition of high uncertainty — but here it is uncertainty that can plausibly cause permanent capital destruction, which pushes it toward risk, not the benign uncertainty I hunt.

Two more flags. The ¥14,645M impairment F60 on a brand-new Sendai plant is capital-allocation carelessness — writing off a plant they just built. And the payout ratio F35 is 211%: they're paying a ¥104 dividend F34 out of ¥49 of earnings, raising it through the crisis E12 — admirable stewardship of the shareholder relationship, but it tells me management is defending the dividend, not opportunistically buying its own distressed stock.

So where does this land? The business is real, the fear is nameable, the balance sheet won't kill me — that clears the survival gate and keeps it off the too-hard pile for now. But two things stop me cold: there is no asset floor, so the whole margin of safety must come from earnings; and at ~20–40x those earnings, the discount simply isn't there at ¥5,505. Layer on an un-bounded compensation tail I can't underwrite, and this fails the fifty-cent test outright. Few bets, big bets — this isn't one I'd make today. But it's the right kind of animal, and one more filing cycle that quantifies the compensation liability would flip it decisively one way or the other. That's a watch, not a buy.

I have been thinking about Kobayashi Pharmaceutical since the red-yeast-rice crisis broke in early 2024. A consumer-health and household-goods franchise built across decades — ¥165bn in revenue from a diversified portfolio of niche category leaders E1E2 — collapsing into a product-safety scandal that ended a family dynasty at the top E5 and forced a ¥14,645M impairment of new plants that should never have been built at their stated economics F60E10. My job, as a prospective decade-owner, is not to decide whether the stock is cheap; it is to decide whether I can honestly claim to understand the business's next ten years better than almost anyone who owns it. If I cannot, the price is irrelevant.

Let me start where I always start: the knowledge bar.

What I actually know. The pre-crisis franchise was compounding quietly: net income grew from ¥19,715M in FY2021 to ¥20,338M in FY2023 F15F17, supported by steady revenue growth from ¥155bn to ¥173bn over the same interval F1F3. That is a real business with real earning power. Operating cash flow in FY2025 was ¥25,590M F41 — stronger than net income in that crisis year — which tells me the non-beni-koji segments are still generating cash. The balance sheet is genuinely conservative: ¥62,314M in cash and deposits F42 plus ¥8,300M in securities F43, against only ¥678M in lease obligations F52 and a recall provision that has been drawn down to ¥2,176M F58. The company is effectively debt-free. Retained earnings of ¥204,240M F53 reflect decades of disciplined accumulation. This is not a melting balance sheet.

What I cannot honestly know. Here the knowledge bar bites hard. The ultimate beni-koji compensation liability is not bounded with precision in the filings E14. The provision F58 is ¥2,176M, but the evidence entry itself acknowledges that compensation to health-damaged customers and affected companies is ongoing, with no fixed total. A product-safety crisis of this nature — mass-market dietary supplement, deaths reported, national regulatory response — carries a tail that Japanese corporate history suggests can extend for years. I cannot tell from the archive what the final number is, and the gap between my estimate and the real number could be material relative to a ¥409bn market cap D1. This is a genuine unknown, not a rounding error.

The second unknowable is the franchise's decade-ahead earning power through the combined shock of the crisis and the capital-allocation failures that preceded it. The ¥14,645M plant impairment F60 is not an accounting abstraction. New plants in Sendai and Thailand were built — the construction-in-progress balance remains at ¥17,656M F50, suggesting more capital is still deployed in infrastructure that has already been impaired once — and the economics apparently did not support the carrying value even before the recall made the problem acute. A management team that allocates capital this poorly on the largest investment decisions is not one I can trust to compound retained earnings efficiently going forward. The capital-efficiency pivot E11 is explicitly acknowledged in the mid-term plan, but acknowledged failure is not demonstrated repair.

What the governance response actually shows. I want to give credit where it is due. The shift from family control to professional management is real: Toyoda Kaichi is now president, while the founding-family Kobayashi moved to lead the compensation-response function E5. The board has genuine external weight — six outside directors, an outside auditor who is a KPMG New York partner, another who is a former Supreme Court official E6. The transition to an audit-committee company structure is scheduled E6. Compensation is tied to EBITDA margin and EPS, not empire metrics E8. Four directors retired after the crisis E9. These are not cosmetic changes. But governance reform and capital-allocation discipline are different things, and only one of them compounds intrinsic value.

The capital-allocation record. This is the item I weight most heavily as a decade-owner, and the record here is genuinely mixed. Retained earnings of ¥204,240M F53 against pre-crisis normalized net income averaging ¥20,025M per year D4 suggests the company retained roughly a decade's worth of earnings on the balance sheet — and yet the cash pile of ¥62,314M F42 sits at deposit rates, investment securities of ¥31,840M F47 lie semi-idle, and the largest investment decision in years — the new manufacturing plants — required a ¥14,645M impairment in the same year it was recognised F60. The dividend, maintained at ¥104 and guided to ¥106 E12F34, reflects institutional respect for shareholders, but at a 211% payout ratio in FY2025 F35 it is funded from the accumulated balance sheet rather than current earnings, which is a one-period bridge, not evidence of earning power.

The FY2026 guidance illustrates the uncertainty starkly. Revenue is guided at ¥173bn F80, recovering toward the pre-crisis peak, but operating income is guided at only ¥12,500M F81 — lower than the already-depressed FY2025 level of ¥14,923M F9. After two years of crisis, and with TV advertising fully restarted as of July 2025 E13, the company's own management does not see a V-shaped operating recovery. Net income is guided at ¥10,000M F82, implying a P/E of roughly 41x on guided earnings at the current market cap D1F82 and roughly 20x on the pre-crisis normalized rate D4. Neither multiple, with this level of uncertainty about franchise durability and liability tail, leaves the margin of safety I require.

The residual question. Is intrinsic value compounding or melting? I cannot answer this with the honesty the question demands. Pre-crisis, the business was compounding at modest but real rates F15F17. The crisis broke that trajectory in two ways: operationally, through the advertising suspension E13 and mail-order wind-down F72; and structurally, through revealed capital-allocation failures in the manufacturing buildout F60E10. The governance response E4E5E6E11 is genuine and the brand portfolio is diversified E1E2, which limits the franchise-destruction scenario. But I cannot yet distinguish between a franchise that is temporarily disrupted and recovering, and one that is recovering into a structurally lower earning-power regime — because the filing record shows only the disruption period and a guidance set that itself suggests continued impairment. I need to observe at least one more year of operating results post-advertising-restart before the answer becomes clear.

At 1.9x book D1F22 and 41x guided earnings F82, I am being asked to pay for recovery I cannot yet verify and franchise earning power I cannot yet project. The knowledge bar is not met for a buy conclusion. The governance response and the balance-sheet durability keep this out of the too-hard category. This is a watch — actively tracked, with a clear falsifier.

I reason from the ledger, not from a story about Kobayashi's brands. The first-principles question here is not "is the crisis over" — it is "what earnings are normal, and does the price pay for them." The arithmetic answers before the narrative does.

Outside view first. The reference class is a consumer brand hit by a safety scandal. This class does not have one base rate; it has two modes. In the recovery mode, the brand equity was never really impaired — customers stayed, the product was fine, the failure was a manufacturing/communication event — and earnings compound back past the pre-crisis line (the Tylenol pattern). In the never-recovered mode, the scandal exposed something structural — the category, the trust, or the distribution was the asset that broke — and earnings settle permanently lower. The ledger gives me evidence for the operational recovery: healthcare and household-goods sales are on a recovery trend E1, TV advertising resumed in full July 2025 E13, and revenue is already flat-to-recovered — FY2025 ¥165,742M F5 versus FY2024 ¥165,600M F4, within a whisker of the FY2023 pre-crisis ¥173,455M F3. Kobayashi looks like the recovery branch on the top line. The problem is that the recovery branch of the base rate is about revenue and brand; it says nothing about whether the price has already paid for a recovery that goes past the old peak. That is a separate question, and it is where the case fails.

Compute, don't assert. Market cap is ¥409,235M D1 on 74.3M shares. Three candidate "normal" EPS numbers, three P/Es: trailing ¥49.19 F32 → 112x (crisis + impairment, not normal); FY2026 guidance ¥134.52 F83 → 41x; pre-crisis ¥268.16 F30 → 20.5x. The bull will say "use the pre-crisis number, 20x is fair." Test that on a no-growth basis, because the class base rate for sustained above-peak earnings is weak and Kobayashi is not claiming a new growth engine — it is claiming a return to normal. Earnings power value, no growth, capitalized at a 6% required return (a defensible equity hurdle for a debt-free consumer-staples name in a positive-rate Japan): at the pre-crisis ¥20,025M run-rate D4 that is ¥333,750M, or ¥4,490/share; adding the deployable cash of ¥28,582M (liquid net cash D2 ¥6,294M plus after-tax investment securities F47 ≈¥22,288M, ¥384/share) gets to ~¥4,874. The stamp price is ¥5,505. So even granting the fully recovered pre-crisis peak as permanent, with no crisis tail and all surplus cash handed over, the honest owner-earnings value sits ~11% below the current price. At a 7% hurdle it is ~¥4,233 — 23% below. The reverse-DCF says the same thing more sharply: at a 6% discount and zero growth, today's price embeds ~¥22,800M of no-growth net income after crediting the deployable cash [reverse of D1/D2]. That is above the pre-crisis peak D4 of ¥20,025M. The price is not paying for recovery to normal; it is paying for recovery past the best the company ever did, permanently, with the crisis tail assumed to cost nothing. That is an outlier bet, and C3 puts the burden of proof on the buyer — a burden the ledger does not discharge.

Strongest short case (C61). Three facts, all from the ledger, none from the risk section. (1) Operating income is guided down again — FY2026 ¥12,500M F81, −16.2% on an already crisis-halved FY2025 ¥14,923M F9, itself −40% on FY2024's ¥24,860M F8. The company's own forecast does not show the recovery the price pays for; it shows a third consecutive year of operating-profit decline. (2) The ¥14,645M impairment F60 is on the Sendai new plant and the Thailand plant E10 — brand-new capacity written down almost as it came online. That is not a legacy asset cleanup; it is a capital-allocation verdict on management's most recent large investment, and it collapsed net income to ¥3,656M F19 and ROE to 1.7% F28. (3) The dividend was raised through the crisis to ¥104 E12 against ¥49 of EPS — a payout ratio of 211% F35 — funded from the balance sheet, not from earnings. The long case does not answer these; it talks past them. "Recovery trend" E1 is a revenue statement; every profit and capital line in the ledger points the other way.

Epistemic limits (C83). The load-bearing unknown is the ultimate crisis liability E14. The recorded provision is only ¥2,176M F58, down from ¥3,970M, while compensation to health-damaged customers and affected companies is described as ongoing E3E14 with no fixed total in the filings. This is genuinely unknowable from the documents — it is a live, open-ended tail, not a disclosed range — and it sits on the wrong side of the ledger for a stock already priced above its no-growth peak-earnings value. I do not need to resolve it to reach a verdict: the price fails on the disclosed numbers alone, and the unquantified tail can only make it worse. That asymmetry is the whole point.

The governance evidence is real and I credit it: a founding-family-to-professional-management transition E5, six outside directors including a KPMG partner and an ex-Supreme-Court lawyer E6, new Quality-Safety and Compensation-Response functions E7, four directors retired at the AGM E9, and a stated capital-efficiency pivot to optimize cash and sell non-business assets E11. This is a genuine reform. But it is also a demonstrated failure — the reform exists because the old structure caused deaths and a ¥14bn impairment — and reform does not change the arithmetic. A better-governed company still cannot be worth more than its earnings power plus its cash, and I have computed both.

Verdict: pass. Not because Kobayashi is a bad company or the recovery is fake — the top-line recovery is real. Pass because the price already capitalizes a permanent above-peak outcome plus a free crisis tail, and no computation from the ledger supports it. I set no buy-below: per the checklist, buy-below is derived from the downside case clearing the return hurdle, and here even the upside case (pre-crisis peak, permanent, cash returned) fails to clear it at today's price. A price is missing, not a thesis.