Daitron Co., Ltd. (7609): An Owned Moat, Smaller Than It Looks, Priced for No Growth — and Still Only a Watch

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	watch	¥2,550	B93/B42; B102
Munger	watch	¥2,600	M44; M1; M40
Pabrai	watch	¥1,600	P1; P56/P18; P53
Li Lu	too-hard	—	L1; L13/L39; L18
Claude	watch	implied ¥2,500	C34; C21; C33/C35

*Four of five lenses reached watch; Li Lu alone stopped at too-hard — a watch consensus (4-of-5).* Daitron is pick #2 of the "quality at a fair price" batch, chosen as the deliberate contrast with pick #1, EBARA JITSUGYO, which drew five watch on a rented moat. Daitron's quality edge is genuinely owned — an in-house manufacturing arm [E19] — its governance cleaner (6-of-9 independent board [E102], takeover defense abolished at the 2020 AGM [E96], a share-cancelling buyback [E82]), its ROE more durable (14.0 / 17.5 / 14.5 / 14.0 / 14.4% F47F48F49F50F51, never below 14%, beating its own 12% target [E110]), its net cash bigger (+¥21.3bn D16, 30% of market cap). And it still only drew watch — because at 2.01× book D1 / 14.57× earnings D2 there is no margin of safety (the buy-belows span ¥1,600–2,600, every one below the ¥3,385 stamp). Two forensic findings decided it. (1) The owned moat is smaller than it looks. The celebrated 27.06% 国内製造 segment margin D24 is profit ÷ external revenue only; that arm sells roughly twice as much internally as it does outside (¥8.6bn intersegment vs ~¥4.4bn external [E146]), so its standalone margin is ~9%, a real but thin edge over the 5.65% distribution core D23 — the honest whole-company net margin is ~4.8% D13. (2) FY2025 is a margin peak (国内製造 segment profit +59.3% D29) — but, and this is the twist that holds it at watch not pass, on an EV basis it is not priced for that peak: strip the ¥21.3bn net cash D16 and EV/NOPAT is only ~10.3× D19D22, implying ~0% perpetual growth. The market withholds growth credit from a ~13% compounder; the cap-based 14.6× reads "priced for growth" only until you net the cash.

The business

ダイトロン (Daitron) is an Osaka electronics house founded in June 1952, when it began life selling tape recorders as a dealer for Tokyo Tsushin Kogyo — the company that is now Sony [E11]. Seventy-four years later it does two things under one roof, and a shopkeeper can follow it in a sentence: it is a middleman that buys electronic parts and factory machines and resells them, and it is a small factory that designs and builds some of those machines itself [E16][E19]. Management calls the combination 製販融合 (manufacturing-distribution fusion) and describes itself as a "technology-driven company" [E30] — a trading company's marketing and logistics muscle, plus a maker's function bolted on to lift value-added and margin.

The group reports three segments [E16]. 国内販売 (Domestic Sales) is the M&S Company, the distribution arm: it procures electronic devices and parts — semiconductors, connectors and assembly products, embedded boards, power devices, imaging equipment, information systems — plus manufacturing equipment, both from outside makers and from its own in-house factory, and sells mostly to domestic customers [E17][E18]. This is the bulk of the enterprise: ~¥71.8bn of external revenue F426 at a thin 5.65% segment margin D23. 国内製造 (Domestic Manufacturing) is the D&P Company, the in-house maker: an Equipment Division that develops and builds optical-device, flat-panel-display and electronic-material production/inspection machines, and a Parts Division that designs harnesses, water-pressure-resistant connectors and power devices [E19]. This is the small, fat-margin arm — ~¥4.4bn of external revenue F428 at a 27.06% external segment margin D24. 海外 (Overseas) is the Overseas HQ plus eleven overseas subsidiaries — one of which, Daitron INC. in Oregon, actually manufactures for the North American market [E21] — turning over ~¥26.9bn F430 at a 7.07% margin D25. All twelve consolidated subsidiaries are 100%-owned, and no single one exceeds 10% of consolidated sales [E23][E24]; nor does any single customer [E149].

Customers keep paying because the electronics content of the world keeps rising — management points to AI, IoT and ICT demand plus automotive and robot automation as the pull [E37] — and because Daitron holds the distribution-agency rights of advanced global makers and offers what it calls a superior customer-asset base with abundant accounts, comprehensive support, and foresight [E31]. That is a genuine, relationship-plus-depth-of-line advantage. But it is not a franchise with pricing power: the filing names the loss of those distribution rights — through a supplier's M&A or a change in its sales policy dissolving an agency contract — as a specific risk [E49], and it names the cyclicality of the semiconductor / FPD / optical-device end markets, where a demand-gap adjustment or a capex decline could contract the market, as another [E48].

Here is the nuance the whole panel wrestles with, and it is the reason this study is Daitron's rather than another distributor's. Unlike EBARA JITSUGYO, whose edge is a rented agency, Daitron's differentiator is owned — the in-house D&P manufacturing arm is Daitron's own IP and its own plant [E19]. That is a structural advantage over a pure middleman. But the ledger shows the owned edge is narrower than the headline margin implies. The 27.06% is struck on external revenue D24; the same arm sold ¥8,595,441k inside the house to its own distribution channel versus ¥4,443,621k to outside customers [E146] — so most of the value-add is booked as an internal transfer that never faces a customer, and the fat margin partly reflects that transfer price. Struck on the arm's total output, the owned edge is real but modest. Owned beats borrowed as a matter of structure; it does not, by itself, make the moat wide.

The numbers

At the ¥3,385 stamp — the close on 2026-07-21, and already post the 2026-01-01 2-for-1 split [E166] — Daitron is a genuinely good, durable business offered at a full-to-fair price after a strong run.

The record is exceptionally durable — this is what earns it the quality bar. Revenue compounded ¥72,341,759k → ¥103,142,476k over five years F1F5, crossing the ¥100bn grand-target of management's own 11th Mid-Term Plan [E33]; recurring profit (経常利益) went ¥4,325,737k → ¥7,156,984k F6F10, never once a loss year; net income attributable to owners ¥2,953,131k → ¥4,923,468k F11F15, a record. Return on equity ran 14.0 / 17.5 / 14.5 / 14.0 / 14.4% across FY2021–25 F47F48F49F50F51 — never below 14% (averaging ~14.9%, recomputing cleanly to 14.37% on average ex-NCI equity D5) — and it beats the company's own 12% mid-term ROE target, the metric its executive stock plan is keyed to [E110]. Book value per share compounded from split-adjusted ¥1,011.61 to ¥1,687.18 F31F35, up ~67% in four years D35, while the company returned roughly seventy percent of earnings to owners D37.

It is earned on a fortress balance sheet, with almost no debt and almost no capital intensity. Equity ratio 44.8% F46; cash and deposits of ¥21,803,877k F277 against interest-bearing debt of just ¥478,108k (short-term borrowings + lease obligations; no bonds) — net cash of +¥21,325,769k (+¥21.3bn) D16, about ¥1,003 a share D18 and roughly 30% of the ¥71.31bn net market cap D4, atop a ¥4bn committed credit line that is entirely undrawn [E64][E156]. Capex is a rounding error — ¥413m [E68] on a ¥103bn revenue base — so owner earnings sit close to reported net income, and the return on the tangible capital the operating business truly uses is very high: on the whole book, the ROIC proxy stripped of idle cash is ~34.0% D22.

The valuation is full-to-fair — the opposite of a margin-of-safety bargain. P/B = 3,385 ÷ split-adjusted BPS 1,687.18 = 2.01× D1; P/E = 3,385 ÷ split-adjusted EPS 232.32 = 14.57× D2; dividend yield = split-adjusted DPS 95.00 ÷ 3,385 = 2.81% D15. That multiple is re-rated: the highlight-table P/E ran 8.1× → 6.0× → 7.7× → 7.0× → 10.4× at successive FY-ends F52F53F54F55F56, and reached ~14.6× at the stamp — a doubling off the five-year norm, with the FY2025 total-shareholder-return figure of 349.9% F177 showing what the re-rating plus a good cycle did to the stock. But the picture inverts once the cash is netted. Enterprise value is only ~¥49.98bn D19 because ¥21.3bn of it is net cash; EV/経常 is ~6.98× D20 and EV/NOPAT ~10.3× D19D22. So the cap-based multiple reads "priced for growth" while the EV-based multiple reveals a business priced for roughly zero perpetual growth — the market withholding the growth credit a ~13% recurring-profit compounder D33 would seem to earn.

The record year leans on the cyclical arm at a peak. The differentiator margin gap is the whole thesis — 国内製造 at 27.06% external D24 vs 国内販売 at 5.65% D23 — and it is the 国内製造 arm that drove FY2025: its segment profit rose +59.3% (¥755,037k → ¥1,202,562k) F439F440D29, on its external margin leaping from 19.42% the prior year D27 to 27.06% D24. That is capital-equipment operating leverage — communications-device machining and inspection equipment, semiconductor/FPD/optical-device demand [E57] — near a cyclical top, and management's own risk note flags exactly this end-market as cyclical [E48]. Two forensic corrections follow. First, the transfer-pricing / captive-volume finding: the 27.06% is profit ÷ external revenue D24, but the arm's real output is ~3× that (¥8,595,441k intersegment vs ¥4,443,621k external [E146]), so struck on total output the standalone margin is ~9%, and external manufacturing revenue is only ~4% of the ¥103bn group — the honest whole-company net margin is 4.77% D13. Second, the sole Key Audit Matter is revenue recognition on exactly this manufacturing-equipment line (¥27,670m, Q4-concentrated, able to "sway the achievement of forecasts") [E160], and the overseas order backlog has already rolled over to 76.8% of the prior year even as total backlog runs 109.1% and 国内製造 backlog runs 173.3% [E61].

Capital allocation has teeth, with one blemish. The dividend runs at a 45.2% payout F162 against a 40% guide [E83]; on top of it, a share-cancelling buyback — 580,700 shares bought for ¥1,586,472,400 via ToSTNeT-3 F527F528, plus a separate 522,630-share cancellation on 2025-04-01 F530[E82], a genuine cannibal shrinking the count rather than warehousing paper (there are no options and no potential shares [E5]). Total shareholder return ran ~69.7% of net income D37. The board is 6-of-9 independent, all six outside directors designated TSE-independent [E98][E102], the takeover defense was abolished at the 2020 AGM [E96], the retirement-bonus racket scrapped back in 2008 [E114], executive pay is modest (¥147,195k for three executive directors [E112]), and Deloitte has audited the accounts cleanly for 29 years [E162]. The blemish: the ¥3,065,586k of listed cross-holdings is growing — 18 names added (+¥41,332k) via the business-partner shareholding association's periodic buying, none reduced F592F594[E116] — against a stated reduction policy [E115]. And the register is insider-comfortable, with no visible activist: the 公益財団法人ダイトロン福祉財団 (Daitron Welfare Foundation) holds 9.49% F545 alongside the employee association (2.65%) and business-partner association (1.78%), a top-10 of 40.82% [E77].

The five lenses

Buffett — watch

Let me tell you what this company does in plain English, because that is where every honest decision starts. Daitron is two businesses under one roof [E16]. The big one is a middleman: it buys electronic parts and factory machines and resells them to Japanese customers [E17]. That distribution arm is about seventy percent of the ¥103bn in sales F5F426, and like most middlemen it lives on a thin slice — a 5.65% operating margin D23. The smaller one is a real factory: they design and build optical-device, flat-panel and electronic-material production equipment [E19], booking a fat 27.06% margin D24. Founded in 1952 selling Sony tape recorders [E11]; still selling and making electronic gear the same way. I can explain it to a shopkeeper, so it clears my first gate.

Now, the folks who wrote my brief want me to fall in love with that 27% factory margin. I won't, and here is why a student should watch this move closely. That 27% sits on only ¥4.4bn of external sales F428. The factory actually ships ¥8.6bn inside the house to its own distribution arm [E146] — it books a rich margin, hands the goods to the salesmen, who resell at their thin markup. So a chunk of that 27% is transfer pricing, not a moat you can take to the bank. The honest number for the whole business is the blended one: a 4.8% net margin D13. When I buy a company I buy the consolidated truth, not the prettiest segment. Call it what it is: an adequate-to-good business — a capital-light distributor with a nice manufacturing kicker — not a franchise with pricing power. It holds distribution rights it warns can be dissolved by a supplier's M&A [E49]. That fails the franchise test squarely.

But adequate-to-good is a compliment when the record and the people are this clean. Return on equity ran 14.0 / 17.5 / 14.5 / 14.0 / 14.4% F47F48F49F50F51, never once below 14%, and it beats the company's own 12% target [E110]. Recurring profit went ¥4.33bn to ¥7.16bn F6F10 with no loss year. Book value per share compounded two-thirds fatter in four years F31F35D35 while they returned roughly seventy percent of earnings D37 — and they returned it the right way: a dividend at a 45% payout F162 plus a buyback that actually cancelled 522,630 shares F530[E82] — a genuine cannibal, not a treasury-stock shuffle to soak up options (there are none [E5]). The balance sheet is a fortress: ¥21.8bn of cash F277 against ¥0.48bn of debt — net cash of ¥21.3bn, about ¥1,003 a share D16D18, near thirty percent of the whole price. Capex is a rounding error, ¥413m [E68]. The board is six-of-nine independent [E98], killed its takeover defense in 2020 [E96], ties pay to an ROE hurdle [E110], and pays its three executives a modest ¥147m [E112]. Deloitte has audited it 29 years with a clean opinion [E162]. I looked for a cockroach and found a housekeeper. There is one wart, and I'll name it: they hold ¥3.07bn of cross-shareholdings and last year they grew the pile — eighteen names up, none down [E116]. It's small, but it's pointed the wrong way, and there's no activist to object [E77].

So the business passes, the people pass — and now the only question Graham drilled into me: what do I have to pay? Here the romance ends. At ¥3,385 the stock fetches 2.0 times book D1, 14.6 times earnings D2, and about 7 times pre-tax operating profit after you hand yourself back the cash D20. The whole-business owner-earnings yield is roughly 6.9% on net income of ¥4.92bn F15 against the ¥71.3bn net cap D4. That is a fair price for a fine business. It is not a bargain, and the figures do not hit me over the head with a baseball bat. Worse, the earnings I'd be capitalizing are a peak: the record rode the factory arm jumping +59.3% D29 in an up-year for equipment [E48], and the multiple has already re-rated from 8× to nearly 15× F52D2. Price is what you pay; value is what you get, and today you're paying full value for the value you get.

That leaves me at watch. My buy-below is ¥2,550 — near the FY2025 low of ¥2,565 [E8], so it isn't a fantasy price. At that price you pay about 1.5 times book, and a ~13% owner-earnings yield on the operating business even when I normalize earnings down off the peak to a mid-cycle ~¥4.3bn (nearer the FY2023–24 line F13F14). That's a real margin of safety on numbers that don't require the good times to keep rolling. Above it, I wait — there is no called strike for standing here with the bat on my shoulder.

What a student should take from this: a clean, high-ROE, net-cash, share-cancelling business is a joy to find and still a mistake to overpay for — quality is a reason to watch, not a license to skip the margin of safety. And never let one gorgeous segment margin (here 27% on ¥4.4bn) blind you to the consolidated truth (here ~5% blended) when that margin is manufactured by selling to yourself [E146].

Munger — watch

Invert first, as always. How does buying Daitron at ¥3,385 become a mistake? The answer is sitting in plain view in the figure table: FY2025 earnings were driven by the Domestic Manufacturing segment, whose profit surged +59.3% D29 on semiconductor/FPD/optical-device equipment demand, lifting its margin from 19.42% a year prior D27 to 27.06% D24. If that is a cyclical peak — and the company's own risk disclosures make precisely that point [E48] — you are paying 14.6× earnings D2 for a trough-to-peak snapback in a notoriously lumpy equipment business, dressed up in the statistics of a compounder. That is the kill path.

Now the business, plainly. Daitron is a 製販融合 electronics group — roughly seventy percent of revenue is domestic distribution at a 5.65% segment margin D23, with a thin but high-margin wedge from an in-house maker. The distribution core is an honest, modestly-moated business: long-standing agency rights to advanced makers [E49] and "a superior customer asset base with abundant accounts" [E31], with no single customer above 10% [E149]. The moat mechanism there is relationship-plus-depth-of-line — real enough to support 5–6% margins consistently, but it will not compound wealth at rates worth paying a premium for. Distribution is the turd in this raisin box. The raisin is the D&P arm, at 27.06% D24 a genuine owned-technology differentiator — but the intersegment figure tells the story precisely: ¥8.6bn intersegment versus ¥4.4bn external [E146], meaning the arm mostly supplies the distribution channel rather than standing on its own feet, and its margin swing from 19% to 27% in one year D27D24 looks like cycle, not a widening moat.

The character screen passes, and I don't want to bury it. Five years of ROE never below 14% F47F48F49F50F51, beating the company's own 12% target [E110]; net cash ¥21.3bn against ¥0.48bn of debt D16; ROIC ex-cash ~34% D22 (though flattered by that cash pile draining invested capital). BPS grew ~13.6% per annum D35 while returning ~70% of earnings D37, and they cancelled treasury shares rather than warehousing them [E82] — honest capital allocation, not Japanese window-dressing. The board is majority independent [E98][E102], the takeover defense was abolished in 2020 [E96], and Deloitte has audited cleanly for 29 years [E162]. Genuine virtues.

But the pari-mutuel odds do not reward them. The stock traded at 6–10× earnings at year-ends through FY2024 F52F53F54F55. It now trades at ~14.6× D2, P/B 2.01× D1, after a full re-rating driven substantially by the cyclical equipment arm having a brilliant year. The market has looked at the same 14% ROE record it valued at 7–8× before and re-priced it as though the record is now demonstrably better, when the likeliest explanation is that one segment had an equipment supercycle. And the trend lines are mixed, not widening: the distribution margin actually slipped from 5.77% to 5.65% D26D23, the overseas backlog fell to 76.8% [E61], and the cross-holdings are growing via routine partner-association buying [E116] — capital leaving for relationship-maintenance equity while ¥21.8bn sits idle on the balance sheet F277. At today's price the thesis requires believing either that the manufacturing arm's elevated margin is durable, or that the distribution arm deserves a higher multiple than it has ever received; I am satisfied of neither.

Verdict: watch. Daitron is a clean, honest, well-managed business, and I cannot distinguish FY2025's manufacturing performance from a cyclical peak — and the price does not give me room to be wrong about that. My napkin case: normalized earnings ~¥185–200 a share (manufacturing reverting toward its FY2024 ¥755m level F439, distribution stable) at 9–10× plus net cash per share ~¥1,003 D18 lands intrinsic value around ¥2,850–3,400, and ¥3,385 sits at the top of that under an optimistic scenario. Buy-below ¥2,600 — below mid-case value with genuine safety, an earnings yield near 7% on normalized earnings plus net-cash backing. Revisit if a correction brings the P/E back to 9–10× or if the manufacturing segment proves its FY2025 margin was not peak-cycle.

What a student should take from this: inversion applied to a clean-looking compounder. A five-year ROE record that never dropped below 14% F47 can be a real quality signal or a distribution-plus-one-cycle-peak dressed as a compounder, depending on whether the high-margin segment's record is structural or cyclical. At 7–8× earnings you can afford to be uncertain; at 14.6× you need to be right about the durability — and the pari-mutuel odds shifted, because this was a better bet at the prices the market offered from FY2021–2024.

Pabrai — watch

Let me start where Dhandho always starts — the downside — because with this one the downside is the whole story, and it cuts both ways.

The business is easy to say in five plain sentences. Daitron is a 74-year-old Osaka electronics house that started life in 1952 selling Sony tape recorders [E11] and became a 製販融合 group — it distributes electronic devices, components and manufacturing equipment (~70% of revenue at a thin 5.65% margin D23), it builds some of that equipment in-house (a fat 27.06% margin D24), and it does both again overseas [E16][E19]. Customers stay because they hold long-standing distribution-agency rights from advanced makers [E49] and because the in-house arm mostly feeds its own distribution channel — ¥8.6bn of intersegment sales versus ¥4.4bn external [E146]. No single customer is 10% of sales [E149]. That paragraph passes the circle-of-competence gate; this is a simple, boring, understandable business.

Now the floor. The balance sheet is a fortress: ¥21.8bn cash F277 against ¥478m of interest-bearing debt — net cash of +¥21.3bn D16, about 30% of the ¥71.31bn net market cap D4, with a ¥4bn committed line entirely undrawn on top [E64]. There is no bond, no covenant tripwire, no maturity wall. Run the two-year shutdown test: if revenue vanished, working capital unwinds into cash and ¥478m of debt is a rounding error against ¥21.8bn — this company survives 24 months without touching a lender's goodwill and never dilutes at the bottom. Ruin risk here is essentially zero. That is exactly the low-risk half of "low-risk, high-uncertainty."

But P1 is not asking "can it survive" — it's asking what a pessimistic mark leaves me, and what fraction of my price it destroys. The equity cushion is not an asset floor; I've been burned enough to obey that. So I mark it down myself. Cash ¥21.8bn is money-good; listed cross-holdings ¥3.07bn F592 are Level-1 [E139], call them ¥2bn after a haircut; receivables and inventory get stressed hard against the liabilities. The hard floor — net cash plus marked securities — is roughly ¥1,003 a share D18 plus a little, call it ~¥1,100. That's about one-third of the ¥3,385 price. The other two-thirds of what I'd pay rests entirely on the operating business staying a going concern — so downside to that floor is roughly −65%, which is emphatically not "tails, I don't lose much."

The valuation confirms it. At ¥3,385 the price is 2.01× book D1 and 14.57× earnings D2; strip the cash and the operating business is EV ¥49.98bn D19 at ~7.0× 経常 D20 — a fair price for a fine business, not a fifty-cent dollar. And the normalization flag is flashing red: FY2025 was a record, and the record was manufactured by the cyclical in-house arm, whose segment profit jumped +59.3% to ¥1,203m D29F440 on its margin swinging to 27.1% D24. That is semiconductor/FPD/equipment cyclicality [E48] near a peak, not a mid-cycle number. What is the market actually afraid of? Almost nothing — and that's the problem. The 349.9% five-year total return F177 and the P/E re-rating from 8.1× to ~14.6× F52D2 tell me the uncertainty has already been priced out. This isn't a fearful-seller, distressed-industry setup; it's a well-run compounder the market has discovered and now pays up for.

The quality is real, to be clear — ROE never below 14% F47F48F49F50F51, ROIC ex-cash 34% D22, capital allocation exemplary (70% total payout D37, a buyback executed and 522,630 shares cancelled [E80][E82], no options [E5], takeover defense abolished [E96]). Cross-holdings growing [E116] are the one blemish, but small. This is a coffee-can-quality business. It is simply not on offer at a Dhandho price.

Verdict: watch. Ruin risk is capped, the business is understandable and excellent, but at 2× book / ~14.6× peak-cycle earnings the low uncertainty is fully priced and the −65% downside to the hard floor breaks the asymmetry. Few bets, big bets, infrequent bets: this one doesn't clear the bar at ¥3,385. I'd be a confident buyer near ¥1,600 — roughly 50% of a conservative ~¥3,000 IV (net cash ¥1,003/sh D18 plus the operating business at ~10× normalized ~¥200/sh owner earnings) — where net cash and securities cover most of the price and I'm buying the electronics franchise nearly for free.

What a student should take from this: net cash covering 30% of the market cap D16 makes a company nearly impossible to destroy, but it does not make the stock cheap — an equity cushion is never an asset floor, and you must mark the hard assets yourself. When the record year is driven by a cyclical segment near a peak margin, capitalizing that number at a re-rated multiple is the classic good-times mistake; normalize first. Great business and great price are two separate tests, and Dhandho only acts when both clear.

Li Lu — too-hard

I want to like this company, and that impulse is exactly what I must interrogate. Daitron is a 74-year-old Osaka electronics group that began in 1952 selling Sony tape recorders as a dealer [E11] and has become a 製販融合 — a fusion of distribution and in-house manufacturing — in three segments: 国内販売, 国内製造, and 海外 [E16]. The record is, on its face, the kind of thing I look for. Revenue compounded from ¥72.3bn to ¥103.1bn F1F5, crossing the ¥100bn mid-term-plan grand-target [E33]; recurring profit rose ¥4.33bn → ¥7.16bn F6F10; return on equity has never once dipped below 14% in five years — 14.0 / 17.5 / 14.5 / 14.0 / 14.4% F47F48F49F50F51, comfortably beating its own 12% target [E110]. The balance sheet holds +¥21.3bn of net cash against ¥0.48bn of debt D16, roughly 30% of the market cap. Book value per share compounded ~13.6% a year D35. On my own test — is intrinsic value compounding or quietly melting? — this is unambiguously compounding.

The capital allocation, too, is the work of people who think like owners. They executed a ¥1,586m buyback and then cancelled 522,630 shares outright [E80][E82], so the repurchase shrank the count rather than warehousing paper. Total payout ran ~70% of net income D37. The board is 6 of 9 independent [E98], the takeover defense was abolished in 2020 [E96], the retirement-bonus system scrapped in 2008 [E114], and the auditor is Deloitte, 29 years continuous, unqualified [E162]. My one governance reservation is that the cross-holdings are growing, not shrinking — 18 names added via a business-partner shareholding-association standing order [E116] — small, but moving the wrong way. On structure and stewardship, Daitron passes.

So why do I stop at too-hard? Because when I ask the only question that gates everything — can I honestly predict this business's earnings power ten years out better than almost anyone who owns it? — I cannot, and the reason is precise. The FY2025 record was not broad-based. It was driven by the 国内製造 arm, whose segment profit leapt +59.3% in a single year D29[E57], carrying a 27.06% margin D24 against the distribution core's 5.65% D23. That gap is the whole thesis: is the manufacturing edge a durable, knowable moat, or a cyclical bump at a peak? The filings answer against durability. That margin was 19.42% only a year earlier D27 — an eight-point swing on a tiny ¥4.4bn external-revenue base F428, most of which is internal (¥8.6bn intersegment vs ¥4.4bn external [E146]). Management itself classifies this end-market as cyclical, warning of demand-gap adjustment and capex decline [E48]. The sole Key Audit Matter is revenue recognition on exactly this equipment, where a handful of large Q4 transactions can sway the forecast [E160]. A 27% margin on a small, lumpy, cyclical, acceptance-timed revenue line is not a moat I can underwrite for a decade — it is a good year I am being asked to extrapolate.

The knowledge bar breaks a second way, on the ~70% of revenue that is thin-margin distribution D23. Its durability rests on retaining distributorships from advanced makers — and the yūhō names the risk plainly: supplier M&A or a sales-policy change can dissolve an agency contract [E49]. Whether a key principal gets acquired and pulls its line is a corporate action I cannot forecast from Daitron's own filings. And I cannot stress-test any of this: the archived five years are entirely an up-cycle, with no 2008–09 or 2020 trough to show me how revenue, margin and cash behaved under attack — the one warning sign that survives is operating cash flow turning negative in FY2022 F58 on a working-capital swing, evidence that the cash conversion is lumpy even in good times.

Even setting the knowledge bar aside, the price forecloses the case. My discipline is a dollar at fifty cents; here I am asked to pay 2.01× book D1 and 14.57× earnings D2 — after a re-rating that lifted the multiple from ~8× to ~14.6× F52D2. If the exchange closed for ten years, I would earn a ~2.8% yield D15 plus whatever per-share growth survives — but the growth I would be buying is a peak-cycle year, and the manufacturing profit that made it a record is the part most likely to revert. There is no margin of safety absorbing the unknowns. A wonderful, well-run, net-cash compounder I cannot predict at ten years, offered at a full price with no room to be wrong, is a pass — and honestly, given the unpredictable variable that carries the earnings, it is too-hard.

What a student should take from this: a clean balance sheet, durable 14% ROE, and owner-minded capital allocation are necessary but not sufficient — they get you past the structural and stewardship gates, not past the knowledge bar. When a "record year" is driven by a small, high-margin, cyclical segment sitting on top of a thin, distributorship-dependent core, the honest answer to "can I predict this in ten years?" is usually no, and no price rescues a business you cannot forecast. Too-hard is a verdict about the boundary of your competence, not a criticism of the company.

Claude — watch

My figures-blind priors, registered first, held up and sharpened. I put P(watch)≈0.45 as the single most probable landing, on two gates: H1 — is the owned manufacturing edge a durable moat or a cyclical bump at a captive peak — where I leaned "real but narrow and cycle-levered"; and H2 — is a fuller price for a better-governed cyclical business a buy/watch/pass — where I leaned watch, fearing the re-rating had left no margin of safety. The ledger confirmed and sharpened H1, and substantially overturned the pass-scenario limb of H2. (Self-distance: I hold this verdict, built the reconciled figure table all five lenses consumed, and wrote the synthesis below — read all three with that concentration of authorship in mind.)

H1 — the captive-margin finding, confirmed and worse than the headline. The ledger's showpiece is the 国内製造 segment's 27.06% margin D24 against distribution's 5.65% D23. That 27.06% is an artifact of the ratio's denominator: it divides segment profit of ¥1,202,562k F440 — earned on the arm's total output of ~¥13.0bn (external ¥4,443,621k F428 plus intersegment ¥8,595,441k [E146]) — by external revenue alone. Struck on total output, the arm's true standalone margin is ~9%, not 27%, and two-thirds of the arm sells captive to the thin-margin distributor at internal transfer prices [E18]. So the owned edge is real but ~9%, and its FY2025 lift is a cyclical peak — segment profit +59.3% D29 while overseas backlog fell to 76.8% of prior [E61] and the auditor's sole KAM is precisely the timing of manufacturing-equipment revenue [E160]. My prior stands, sharpened: a genuine owned-margin edge over the distribution core, but a narrow one.

H2 — the price is far more reasonable than "fuller price" feared, overturned toward watch. I expected the ¥3,385 stamp to capitalize the peak. The reverse-DCF says the opposite. EV is only ¥49.98bn D19 because net cash is ¥21.3bn D16 (30% of the ¥71.3bn cap D4). EV/NOPAT = ¥49.98bn ÷ NOPAT ~¥4.84bn D19D22 = **10.3×**. Solving a perpetuity for the implied growth: at r=9% the price embeds roughly −0.6% perpetual growth; even at r=8%, only ~−1.5%. The market is withholding growth credit from a business that compounded recurring profit at ~13.4% p.a. D33 and BPS at ~13.6% D35 over five years — not extrapolating a peak. That kills my pass scenario: you cannot say "the price is set on the peak" when the operating business is priced for stagnation.

The deciding work is the owner arithmetic. Bear case first, off the company's own cycle history — five-year recurring margins averaging 6.66% F6F10. Mid-cycle bear: revenue reverting to the FY2023 ¥92.2bn level F3 at a ~6.3% recurring margin gives NI ~¥4.0bn, EPS ~¥190, capitalized at 9% no-growth plus net cash ¥1,003/sh D18 ≈ ~¥3,125/sh, a −8% loss. A hard down-cycle (revenue −15%, recurring margin 5.2%, plausible given the semiconductor/FPD/optical capex cyclicality [E48] and the backlog signal [E61]) ≈ ~¥2,672/sh, −21%. The downside is real but survivable, cushioned by cash rather than by cheapness. Earnings power (EPV ~¥3,125 mid-cycle) sits far above the haircut asset floor (¥646/sh, ~19% of price) — so this is an earnings-power name, not an asset play, and the franchise carrying the value is a ~9% owned edge on a thin core: better than borrowed, not a wide moat. And the asset floor is only partly realizable — the register is insider-comfortable (Welfare Foundation 9.49% F545, employee + partner associations [E77]) with no activist, and cross-holdings growing against a stated reduction policy [E116][E115], so no mechanism compels realization. Private-owner yield, multiple-free: NI/cap = 6.9%, and on normalized peak-stripped earnings ~6.6% — against a ~7–8% hurdle for a cyclical distributor, this clears barely if at all at ¥3,385, and clears comfortably ~20% lower.

Verdict: watch. A durable ~14%-ROE F47F48F49F50F51, net-cash, near-zero-debt, clean-board (6/9 independent, pill abolished 2020 [E96][E102]), share-cancelling (522,630 shares retired F530) business — whose owned edge is genuine but ~9% and captive [E146], and whose ~70% of revenue is thin, partly-agent-net-accounted distribution [E132] — at 2.01× book D1 and ~14.6× earnings D2. The reverse-DCF shows the price is fair, not stretched; but "fair" is not a margin of safety. This is the EBARA-JITSUGYO "quality at a fair price" pattern with an owned rather than borrowed edge — a better business, still not a bargain. Implied buy-below ¥2,500 — ~20% below mid-cycle EPV and near the hard-down-cycle value where the loss goes to ~zero and the owner yield clears the hurdle with the cash intact. I would own it there, not here.

What a student should take from this: a segment margin is only as honest as its denominator — Daitron's celebrated 27% is struck on external revenue while the profit is earned on ~3× that output, so the true standalone edge is ~9% because two-thirds sells captive to its own distributor [E146]. And a "fuller price after a re-rating" is not automatically expensive — the reverse-DCF here shows ¥3,385 embeds ~0% perpetual growth once you strip ¥21.3bn of net cash D16, so the tape was pricing stagnation, not a peak. The discipline that mattered was refusing both the bull's headline margin and my own prior's fear.

Synthesis

Where the lenses agree

Four of five lenses reached watch; Li Lu alone stopped at too-hard. But the disagreement is narrower than the verdict labels suggest — every lens agrees on the same three facts, and the split is only whether those facts bar the price (watch) or bar prediction itself (too-hard). The agreement is total on three things, and they are the whole study.

First, the quality is genuine and durable. ROE ran 14.0 / 17.5 / 14.5 / 14.0 / 14.4% across five years with no loss year F47F48F49F50F51, beating the company's own 12% target [E110]; recurring profit compounded ¥4.33bn → ¥7.16bn F6F10; and it is earned on a fortress balance sheet — net cash +¥21.3bn D16, ~30% of cap, capex a rounding error [E68]. Every lens credits this. Buffett: "a very good business hiding behind a fortress balance sheet." Munger: "the character screen passes." Li Lu: "unambiguously compounding." Claude: "my priors under-, not over-, weighted the quality." And the governance is cleaner than pick #1's: 6-of-9 independent [E102], takeover defense abolished 2020 [E96], a share-cancelling buyback [E82], ~70% total payout D37.

Second, the price is full-to-fair — there is no margin of safety. At 2.01× book D1 and 14.57× earnings D2, after a re-rating from 6–10× F52F53F54F55, the panel is unanimous that the discount every deep-value name offered is simply absent. Buffett: "a fair price, not a cheap one … no baseball bat because there's no bruise." Pabrai: "an asset floor is not the same as a fifty-cent dollar." Li Lu: "priced for a good, well-run compounder — fairly, not fearfully." Claude: "fair, not cheap, and harder to beat than a bargain would need." The net cash is real but only ~30% of the price D16; the operating business is not being handed to you at a discount.

Third, the record year is a peak on the cyclical arm, and the owned edge is thinner than the headline. FY2025's record was driven almost entirely by 国内製造 segment profit +59.3% D29 on its external margin leaping to 27.06% D24 from 19.42% D27 — capital-equipment operating leverage near a top [E48][E57]. And that 27.06% is profit ÷ external revenue only D24; struck on the arm's total output (¥8.6bn intersegment vs ¥4.4bn external [E146]), the standalone edge is ~9%. Every lens normalizes the peak down, and every lens deflates the headline margin: Munger ("cycle, not a widening moat"), Pabrai ("the record was manufactured by the cyclical in-house arm"), Li Lu ("a good year I am being asked to extrapolate"), Claude ("the true standalone edge is ~9%"). This is the panel's operating-line discipline catching a flattered headline.

Where the lenses diverge

The panel agrees on the facts, but the buy-below spread is wide — ¥1,600 to ¥2,600 (with Li Lu declining entirely) — and the spread is not noise. It is four different definitions of the discount required, each driven by a specific item, plus one lens for whom no discount suffices.

Pabrai anchors to the net-cash floor (¥1,600, the lowest number). His P1 asks what he loses if wrong, and his answer is a hard asset floor — net cash plus haircut securities, ~¥1,100/share — that makes this not a pass. But P53 (the fifty-cent dollar) then demands he buy near that floor, "where net cash and securities cover most of the price and I'm buying the electronics franchise nearly for free." He will pay up only when the balance sheet does almost all the work, which drives the deepest buy-below on the panel — ~50% of his conservative ~¥3,000 IV.

Claude (implied ¥2,500) and Munger (¥2,600) bracket the top by the same normalization route — the price at which the normalized (not peak) owner yield clears a ~9% hurdle. Claude's C35 is the sharpest statement of the shared crux: the owner yield "clears barely if at all at ¥3,385, and clears comfortably ~20% lower." Munger reaches essentially the same number from a napkin case — normalized EPS ~¥185–200 at 9–10× plus net cash per share ~¥1,003 D18 — landing intrinsic value at ¥2,850–3,400 and a buy-below below it. Buffett's ¥2,550 sits with them, but by a different protection: not the floor and not the normalized-yield hurdle but a demanded ~13% owner-earnings yield on the operating business at ~1.5× book, "a real margin of safety on numbers that don't require the good times to keep rolling."

Li Lu prices the knowledge bar — and finds no price clears it (too-hard). His L1 gate is neither the floor nor the normalized-yield price but predictability: he can map the two decisive variables (the cyclical 国内製造 margin [E48], the retention of thin-margin distributorships [E49]) to disclosures, but cannot answer them "with the confidence my standard demands." Where the other four convert their doubt into a discount (a buy-below), Li Lu converts it into a boundary: a business whose carrying earnings ride an unpredictable, un-stress-tested cyclical variable is one he declines to underwrite at any price — and at 2× book on a peak year, the price gives him no reason to strain the boundary. That is the whole difference between his verdict and the other four: same facts, but for him the earnings are unknowable rather than merely cyclical-and-fully-priced.

The gap across the panel, then, is not whether the business is good (all five agree it is) but what protects you — the asset floor (Pabrai), the owner-earnings-yield / normalized-yield hurdle (Buffett, Munger, Claude), or nothing short of predictability itself (Li Lu). All five agree ¥3,385 — 14.57× peak earnings D2 — is not a buy; four would own it ~25–50% lower, and one would not underwrite it until the cyclical variable that carries the earnings becomes knowable.

The red team, engaged

Because four lenses converged on watch (a consensus non-decline), a fresh adversary (ledger only) argued the stronger PASS case — that ¥3,385 (2.01× book D1) is not a "wait for cheaper" but a "drop it from the list," because the quality on display is borrowed-in-spirit, peaked, and flattered, so no future price reliably converts this into a sound minority holding. Its claim is deliberately worse for the name than "it's too dear": that the thing you would be buying at any price is a thin, mostly-captive, cyclical-peak sliver of quality bolted onto a moat-light distribution book, with no forcing agent to convert the cash into a return you capture. The synthesis must meet its strongest points by name. Each sharpens the crux; none breaks watch down to pass, and here is why.

• "The owned moat is a transfer-pricing artifact, and it is small." The bear's central point, and it is true as stated — indeed it is the panel's own finding. The 27.06% D24 is profit ÷ external revenue, but the arm sold ¥8,595,441k internally vs ¥4,443,621k external [E146], so the standalone margin is ~9% and external manufacturing revenue is only ~4% of the ¥103bn group F428 — a 4.77% net-margin whole D13. This is exactly the crux the panel already priced; it is why the buy-belows sit 25–50% below the stamp. The rebuttal to pass is that a small owned edge is a reason to demand a discount and to refuse to celebrate the headline margin — not a reason to declare the business un-ownable. A ~9% owned edge still beats the 5.65% distribution core D23, is Daitron's own IP and plant [E19] rather than a rented agency, and has helped produce a 15%+-through-cycle ROE with no loss year F47F48F49F50F51. The bear proves the edge is narrow; it does not prove it is worthless.

• "Peak earnings × peak multiple." Fully conceded, and the panel conceded it first. The record is one lever — 国内製造 profit +59.3% D29 on a margin that leapt to 27.06% D24 from 19.42% D27 — priced at 14.57× D2, re-rated from the 6–8× this stock carried at every FY-end through FY2024 F52F53F54F55. This is exactly why every lens normalizes down and none pays 14.6× the peak. But a peak-year print is a reason to value on the cycle, not to conclude the through-cycle business is worthless — and the through-cycle record here is genuinely good (no loss year, ROE trough at 14.0% F47F50). The guide-down caps the price you pay; it does not zero the franchise.

• "No forcing agent — the cash can sit for years." Conceded on the facts. The register is engineered for insider comfort — Welfare Foundation 9.49% F545, employee and partner associations, top-10 40.82%, no visible activist [E77] — and against a stated reduction policy [E115] the cross-holdings grew: 18 names added, zero reduced [E116]. This is the bear's real challenge to watch itself, and it is the honest tension. The rebuttal: the net cash is a genuine downside cushion whether or not it is activated (~30% of cap D16), and the capital-return trajectory is executing without an activist forcing it — a 45.2% payout F162, a buyback of 580,700 shares F527 with 522,630 cancelled F530[E82], ~70% total payout D37. The bear is right that a lower price does not force the cash out; the panel never claimed it would — the buy-belows assume the bear case (normalized earnings, no re-rating) and simply demand to buy the good-through-cycle business where the normalized yield clears.

• "No margin of safety at the stamp or at the watch price." The bear presses that even at ¥1,600–2,600 you have re-priced, not removed, the defects. True — and the panel agrees the defects persist at any price; that is why they are reasons for the discount, not reasons the discount cures them. But the decisive rebuttal that holds it at watch not pass is the C34 EV finding: the market is not capitalizing the peak. Strip the ¥21.3bn net cash D16 and EV/NOPAT is ~10.3× D19D22, implying ~0% perpetual growth — so ¥3,385 is a fair, not a full, price for a ~13% compounder D33 once the cash is netted, and a ~25% lower price is a genuine margin of safety, not merely a cheaper hold of the same problem. The bear values the name on its cap-based 14.6× D2; the panel values it on its EV, and the two multiples tell opposite stories from the same ledger.

• The bear's honest concession — backlog momentum — cuts against a simple peak. The red team concedes, and the panel weights, that order backlog genuinely continues: total 109.1% YoY, with 国内製造 backlog at 173.3% [E61]. The bear holds this is a volume signal not a margin signal, which is fair. But it is the single best fact against a clean peak call: it plausibly holds FY2026 revenue up even if the mix normalizes, and it is precisely the observable the panel is waiting on. The bear's own falsifier — standalone 国内製造 margin holding ≥9% on total output for two more years, or segment profit holding above ~¥1.0bn F440 — is the same test every lens registered.

The red team did not move any verdict off its landing, but it earned its keep: it is why the synthesis states the crux as sharply as it does — is the owned edge wide enough, and is FY2025 a peak? The honest resolution is that both concerns are true (the edge is 9% and captive; the year is a peak), and both are reasons for the ~25% discount the panel demands rather than reasons to declare the business uninvestable. The net cash (+¥21.3bn D16), the EV-basis finding (10.3× / ~0% growth D19D22), and the genuine no-loss-year durable-ROE history (floor 14.0% F47F50) keep it a watch, not a pass: a good business, correctly priced as fair, worth owning ~25% cheaper — not a value trap to discard.

Self-distance note. The Claude lens holds one of the five verdicts compared above (watch) and wrote this synthesis; it also built the dual-blind reconciled figure table and evidence ledger all five lenses consumed, and the red-team ran on the same model family. That is an unusual concentration of authorship — the answerer, the ledger-builder, one of the five voters, and the adversary are the same system. Read the synthesis with that in mind.

Prediction-vs-actual: VOID. This was an autonomous headless cycle; the human blind prediction is voided (void: no-human-prediction, never forged). No prediction-vs-actual scoring applies.

Verdict accounting (fixed ex-ante)

• A buy-below-¥X verdict is price-falsifiable against the unadjusted stamp. Three lenses issued explicit buy-belows — Buffett ¥2,550, Munger ¥2,600, Pabrai ¥1,600; Claude publishes an implied buy-below of ¥2,500; Li Lu issued no buy-below (too-hard).
• pass / watch / too-hard are recorded but unscored in any future review. Four verdicts here are watch and one is too-hard; the buy-belows sit below the ¥3,385 stamp and are the price at which each watch lens would revisit toward buy.
• The original verdict counts at its original stamp regardless of later corrections.
• On a stock split, reverse split, or consolidation, the buy-below threshold restates mechanically by the announced ratio (corporate-action disclosure cited); the stamp itself never restates. (The ¥3,385 stamp already reflects the 2026-01-01 2-for-1 split [E166]; all buy-belows are stated on the post-split basis.)

Red team

A consensus red-team (four of five lenses converged on watch) was dispatched to argue PASS — that ¥3,385 / 2.01× book is not a "wait for cheaper" but a name to drop, because the FY2025 quality is captive and small (a 27% margin that is really ~9% on total output [E146]), peaked (国内製造 profit +59.3% D29 near a cyclical top), and un-forced (an insider-comfortable register with no activist and cross-holdings growing [E77][E116]), so no future price reliably makes it a sound minority holding. Its strongest points, verbatim-faithful, and the synthesis's engagement with each, are in "The red team, engaged" above. In brief, the adversary's three ranked points were: (1) the owned moat is a transfer-pricing artifact and small — profit ÷ external revenue inflates 9.22% into 27.06% D24[E146], on ~4% of group revenue F428; (2) peak earnings × peak multiple — a +59.3% one-segment print D29 priced at 14.57× D2, re-rated from 6–8× F52F53F54F55; (3) no forcing agent on a balance-sheet-heavy thesis — ~30% of cap is net cash D16 but the register is foundation-plus-associations with no activist [E77] and cross-holdings were added to [E116] against a written reduction policy [E115]. The synthesis concedes every factual point — the edge is ~9% and captive, the year is a peak, the cash is un-forced — and meets the load-bearing claim (that these make the business un-ownable at any price) with the decisive C34 EV finding: the market is not capitalizing the peak — strip the ¥21.3bn net cash D16 and EV/NOPAT is ~10.3× D19D22, implying ~0% growth — so ¥3,385 is a fair, not full, price for a ~13% compounder D33, and each defect is a reason for the ~25% discount the panel demands, not a reason to drop a no-loss-year, net-cash, durable-ROE business. The backlog momentum [E61] cuts against a simple peak, and the bear's own falsifier (standalone 国内製造 margin holding ≥9% on total output, or segment profit above ~¥1.0bn F440, for two more years) is the same observable the panel is waiting on. A consensus that faced its strongest opponent and emerged intact — sharpened on the crux (owned-but-narrow moat; peak vs plateau), unmoved on the verdicts.

What would change our minds

Pre-registered falsifiers, per lens issuing a buy-below or holding the watch (and, for Li Lu, what would move too-hard). The verdicts converge on a single observable: does the 国内製造 standalone margin (~9% on total output) and segment profit hold through FY2026–27 — proving the owned edge is durable, not a one-year peak — and does the ¥21.3bn cash get forced out — versus the price simply falling into the ¥1,600–2,600 range.

• Buffett (watch, buy-below ¥2,550). Toward buy: a price near ¥2,550 (P/B ~1.5×, ~13% owner-earnings yield on the operating business off mid-cycle earnings) — near the FY2025 low ¥2,565 [E8] — clears the margin-of-safety bar. Toward pass/too-hard: if durable ROE breaks below its own 12% MTP floor [E110] on a down-cycle for two consecutive years, OR the share-cancelling capital return reverses (buybacks stop F527 while cross-holdings keep growing [E116]) — showing the discipline was a fair-weather habit.
• Munger (watch, buy-below ¥2,600). The falsifier is durability: 国内製造 segment profit (FY2025 ¥1,202m F440) holds above ¥900m for two consecutive fiscal years AND net cash remains above ¥15bn — demonstrating the +59.3% FY2025 surge D29 was durable, not a peak-cycle bump; conversely, a reversion of that segment profit below ~¥900m confirms the peak.
• Pabrai (watch, buy-below ¥1,600). Toward buy: a quotation drop to ≈¥1,600 (≈50% of a conservative ~¥3,000 IV) converts watch to buy — OR the 国内製造 cyclical peak proves durable (segment margin holds ≥20% D24 and segment profit does not mean-revert off the +59.3% spike D29 across the next two prints), raising the sustainable base enough to justify entry nearer today's price. Toward too-hard: a lost distribution agency [E49] or a supplier M&A visibly breaks the thin-margin distribution core.
• Li Lu (too-hard). Toward watch/buy: the two unknowable variables become knowable — the 国内製造 standalone margin holds through a demonstrated down-cycle (not just an up-year), and the distributorship base proves durable through a supplier-M&A event [E49] — converting assumptions into verified facts. As long as the carrying earnings ride an un-stress-tested cyclical variable [E48] on an all-up-cycle archive F58, it stays too-hard regardless of price.
• Claude (watch, implied buy-below ¥2,500). A close ≤ ¥2,500 with the FY2026 yūhō confirming recurring margin ≥6% and net cash preserved flips watch → buy; OR overseas backlog worsening [E61] into a 国内製造 external-margin collapse below the FY2024 19.4% D27 with recurring margin <5.5% flips watch → pass. The registered probabilities: FY2026 recurring profit below FY2025's peak (0.55), the 国内製造 external margin reverting below 27.06% (0.70), ROE still ≥12% (0.80), and another share-cancelling buyback (0.60).

The single observable most lenses converge on is whether the 国内製造 arm's ~9%-standalone margin and its segment profit hold on a NORMAL (non-peak) year — the difference between a durable owned edge and a one-year cyclical peak F439F440 — together with whether the ¥21.3bn net cash D16 is forced out or returned — versus the price falling to the ¥1,600–2,600 range where each watch lens's discount is satisfied.

What this taught the checklists

Queued for the next study (F2 revision proposals; see docs/process/evolution.md), attributed per lens:

• Buffett — add a transfer-pricing flag for 製販融合 / vertically-integrated groups to B24/B26. When a high segment margin is struck on a segment whose internal sales exceed its external sales (here 国内製造: ¥8.6bn internal vs ¥4.4bn external [E146]), the item should require stating the consolidated blended margin alongside the segment margin, so a transfer-price artifact can't be mistaken for a moat. Also: B43 should record an explicit "archive-depth" sub-check — where only five fiscal years are archived, mark pass (5-yr, shallow) rather than a clean pass, so a shallow record can't silently earn full ten-year credit.

• Munger — three sharpenings. M19 (volume cost curve) needs a segment branch for distributor-plus-manufacturer hybrids — the relevant scale metric for the distribution arm is revenue-per-employee, not blended gross margin (which is mix-driven D7D8). M40 (raisins and turds) needs a companion condition for distribution-dominant businesses: if the thin-margin segment generates >60% of revenue at a 5.65% margin D23 and sets the ceiling for blended returns, trigger a fail regardless of capital-employed concentration. M85 (reason test) should explicitly require distinguishing cycle from structure in the third "why" layer for cyclical-component businesses — here the "why" chain for the 27% margin broke at "equipment demand surged with the communications-device capex cycle" [E57].

• Pabrai — two. Add a "quality-compounder trap" note to P1: a thick net-cash cushion caps ruin risk but not price risk; when net cash is a minority of the price, the majority still rides on a going-concern operating business and must be marked at a going-concern-failure scenario, not just liquidation — the exact error this name invites. And P56 needs a segment-mix normalization prompt: for 製販融合 groups the consolidated earnings number can be lifted by a small high-margin cyclical segment (here 国内製造, 27% margin D24, driving +59.3% D29 on ~4% of external revenue F428); re-strike normalized earnings on that segment's mid-cycle margin before valuing.

• Li Lu — two. L3/L38 archive-depth interaction: several too-hard-forcing items require history a single 5-year yūhō window structurally cannot supply — a checklist note should distinguish "too-hard because the business is genuinely unknowable" from "too-hard because the archive is too shallow to know it," because only the first is a permanent verdict. And a cyclical-peak margin sub-test under L39/L41: no single item currently asks "is the highest-margin segment carrying the result also the most cyclical, and is it at a peak?" — the exact pattern (国内製造 27% D24 driving +59.3% D29 while flagged cyclical [E48]) that was decisive here.

• Claude — two checklist sharpenings plus the library note. (1) C21 — a "segment margin denominator" sub-check: for any segment with material intersegment sales, compute segment margin on total output (external + intersegment), not external alone, and show both — the 27.06%-vs-9% gap D24[E146] is exactly the trap the current wording lets a bull set, and the fix changed this verdict's core finding. (2) C40/C34 — a "net-cash-adjusted reverse-DCF" instruction: for net-cash names, run the reverse-DCF on EV net of deployable cash, not market cap, or the implied-growth read is badly distorted — here the cap-based P/E (14.6× D2) reads "priced for growth" while EV/NOPAT (10.3× D19D22) reveals "priced for ~0% growth," opposite conclusions from the same ledger.

  Library (Class-level) note. Daitron extends the EBARA-JITSUGYO "durable-return, net-cash, quality at a fair price" corner as its OWNED-moat counterpart — it does not seed a distinct corner. N=2. The corner's constructive axis is the moat's provenance: EBARA JITSUGYO rents its edge (agency rights withdrawable on supplier M&A — the exact risk Daitron also discloses [E49]); Daitron owns its edge (in-house 製販融合 manufacturing [E19]). But the Daitron datapoint teaches that an owned edge can still be narrow, captive, and cyclical — a ~9% real margin over the distribution core, two-thirds sold internally [E146], at a one-year peak D29 — so "owned > borrowed" is a structure advantage that does not by itself widen the moat or clear the price. Both names landed at watch on a fair price with no margin of safety (Daitron adding one too-hard), which is the corner's emerging regularity: quality-at-a-fair-price Japanese electronics distributor-makers screen as late, not cheap, and the fair-price entry — not the business quality — is what fails the buy bar for both. The open question the corner should now answer: does the owned-vs-borrowed moat distinction ever produce a different verdict, or does the fair-price entry dominate both? — and its own base rate: for good-but-narrow-moat compounders at a fair price, how often does the market later offer the ~25–30% discount at which the owner yield clears?

---

Corrections