三晃金属工業株式会社（Sanko Metal Industrial Co., Ltd.）

研究原文：Sanko Metal Industrial Co., Ltd. (JPX official English name SANKO METAL INDUSTRIAL CO.,LTD., code 1972, TSE Standard, non-consolidated reporting, 19,800,000 shares after the October 2025 5-for-1 split) is a roofing contractor. Roofing is 92.9% of FY3/26 sales and about 97% of segment profit; Building Materials earns a 1.9% segment margin and power sales are immaterial. Construction work (recognised over time) is 84.1% of revenue (JPY 39,584m of 47,058m), product sales 15.7%. The model integrates development, design, manufacture and installation of metal roofs and walls at about 3,000,000 sq m a year, so one project earns both a construction margin and a product margin.

内部研究类型：business_identity_and_model

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Shareholders own a focused domestic contractor with no conglomerate buffer; a long-span roofing downturn cannot be offset inside the portfolio.

不确定性（原始记录）：Line-level gross margins and the manufacturing/installation split of construction margin are not disclosed.

研究原文：The moat is execution capability rather than a product: in-house development-to-installation integration, design-stage specification, plants, a national site organisation and a landmark reference list (National Stadium, Tokyo Big Sight, Haneda cargo) prequalify the company for large-span projects. The industry-leading and 'largest roofing contractor' claims rest on the company's own 3,000,000 sq m volume; no third-party share, win-rate or pricing series exists to verify them.

内部研究类型：moat_judgment

内部来源标记：COVERAGE35

普通股东含义（原始记录）：The moat defends the roughly 21-23% construction margin band where nearly all profit is made; if the leadership claim is overstated, FY3/26-H1 FY3/27 order resilience would be cycle rather than share gain.

不确定性（原始记录）：Unverifiable in share terms; margin behaviour through the next downturn is the real moat test.

研究原文：FY3/26 closed with revenue up 3.7% to JPY 47,058m but operating profit down 8.1% to JPY 3,778m (ordinary profit -7.1% to JPY 3,843m). Construction gross margin fell 1.2pt to 21.6% (gross profit -JPY 293m, about JPY 475m of margin loss at that volume) on construction cost increases and manufacturing/construction-strengthening spend, and headquarters-relocation costs lifted SG&A (salaries +JPY 190m, R&D to JPY 330m). The decline was a profit-quality event, not a demand decline.

内部研究类型：profit_quality_finding

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Shows the margin, not volume, decides profit direction; one point of construction margin is about JPY 400m a year.

不确定性（原始记录）：Project-level margin dispersion not disclosed; relocation costs treated as non-recurring by management narrative only.

研究原文：H1 FY3/27 (30 October 2026): revenue fell 4.2% to JPY 21,425m on slipped completions (completed construction -6.5% on predecessor-process delays and weather), yet ordinary profit held at JPY 1,636m (-0.3%) because construction gross margin improved 0.9pt to 21.9% on price pass-through and cost improvement, SG&A fell JPY 75m to JPY 2,892m and interest income rose to JPY 56m from JPY 24m. The profit defence, not revenue, is the period's story; note part of the defence is interest income on idle cash.

内部研究类型：profit_quality_finding

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Evidence that the pass-through lesson of FY3/26 was learned; revenue shortfall is timing, not lost demand.

不确定性（原始记录）：Repair is half-year evidence; the heavy second half is the real test.

研究原文：Demand and orders have diverged: national non-residential steel-frame starts fell 4.8% in the FY3/26 window and 10.4% in April-August 2026 (factory/warehouse -11.9% then -8.4%), yet FY3/26 orders held flat at JPY 47,933m and H1 FY3/27 orders rose 5.0% to JPY 24,359m on large new projects (production facilities, logistics warehouses, sports and railway facilities), lifting the backlog to a record JPY 39,395m (+8.2%). Long-span roofing orders rose 12.1% and high-tough 22.2%, while formed-product orders fell 39.6% and painting orders were net negative on cancellations. The order engine is mix (renovation, share gains, large projects), not market beta; share gain is inferred, not measured.

内部研究类型：demand_mechanism

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Order resilience against a shrinking pool supports the backlog-based earnings case, but only if the divergence is share and renovation capture rather than lagged cycle.

不确定性（原始记录）：Renovation share of orders and customer concentration are not disclosed; share-gain interpretation is inference.

研究原文：The carried backlog of JPY 39,395m is about 0.85 times annual sales and is concentrated in long-span roofing (JPY 29,904m). It substantially pre-sells the next two to three quarters of revenue, so the binding constraint on FY3/27 delivery is site execution and predecessor trades, not order intake. It converts the investment question from whether demand will come to whether execution and margin hold. The share scheduled for H2 FY3/27 and its locked-in margin are not disclosed.

内部研究类型：visibility_mechanism

内部来源标记：OPEN_DISCOVERY

普通股东含义（原始记录）：Near-term earnings surprises are more likely to be timing than demand in origin; backlog is the model's inventory of future profit.

不确定性（原始记录）：Open question: what share of the JPY 39,395m backlog completes in H2 FY3/27 and at what locked-in margin (UNKNOWN; would be resolved by project-level scheduling or briefing materials).

研究原文：Backlog quality improved as it grew: at September 2026 growth sits in long-span roofing (+10.9%) and high-tough (+64.4%), R-T +9.3%, the core large-project lines, while the shrinking lines are formed products (-26.6%) and solar (-29.3%), the smaller and shorter-cycle ones.

内部研究类型：quality_finding

内部来源标记：COVERAGE35

普通股东含义（原始记录）：The backlog increase is in the lines that carry the construction margin and longest visibility.

不确定性（原始记录）：Backlog margin profile and half-by-half completion schedule not disclosed.

研究原文：On 30 October 2026 FY3/27 sales were cut JPY 500m to JPY 46,500m on construction slippage while operating profit was raised to JPY 3,570m, ordinary profit to JPY 3,700m (+2.8%) and net income to JPY 2,490m (EPS JPY 129.16). The implied second half is sales JPY 25,075m (+1.5%) and ordinary profit JPY 2,064m, 6.3% below H2 FY3/26's JPY 2,202m: margin give-back in the heavy half despite a record backlog. It is either prudence or an early warning; prior guidance behaviour (Q3 FY3/26 held forecast while nine-month operating profit was -14.1%) suggests revenue guidance is credible but margin guidance less informative. How much of the H2 figure depends on a few large-project completion dates versus broad margin defence is unresolved.

内部研究类型：guidance_and_expectation_bridge

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：A low bar against a record backlog gives room to beat, but the conversion check arrives with the Q3 release.

不确定性（原始记录）：Project-level H2 bridge not disclosed (UNKNOWN); Q3 release date inferred from the company's calendar.

研究原文：Earnings are completion-weighted: Q1 contributes under a tenth of annual profit (Q1 FY3/26 net income JPY 222m of JPY 2,645m), nine months reached 65% of net income and about 35% landed in the fourth quarter. The H1 ordinary-profit progress of about 44% against the revised forecast is consistent with the seasonal pattern rather than a shortfall.

内部研究类型：earnings_rhythm

内部来源标记：OPEN_DISCOVERY

普通股东含义（原始记录）：Interim price moves between January and April prints carry little information about the case.

不确定性（原始记录）：Progress ratio is an inference from the prior-year pattern.

研究原文：Product gross margin fell 1.7pt to 14.7% in H1 FY3/27 while product sales rose 9.6% to JPY 3,536m (FY3/26 product margin 15.7%): growth came from formed products, the volume line, not higher-margin work, so mix rather than pricing appears to be the swing factor. The opposite movement of construction margin (+0.9pt) and product margin (-1.7pt) is the period's key quality signal. Attribution among mix, pricing and cost, and recoverability of the prior roughly 16% level, is unresolved; formed-product orders fell 39.6%, so the line's volume support is itself weakening.

内部研究类型：margin_anomaly

内部来源标记：OPEN_DISCOVERY

普通股东含义（原始记录）：Product line is only 15.7% of revenue, so effect on group profit is limited, but it is a mix-driven drag.

不确定性（原始记录）：Cause of the margin fall (mix vs price vs cost) not disclosed.

研究原文：Construction margin is a race between input cost (steel and materials bought at market-linked prices that remain high; labour) and project-by-project negotiated pass-through. FY3/26 showed the failure mode (-1.2pt) and H1 FY3/27 the partial repair (+0.9pt to 21.9%). Labour is the cost line most likely to compress margin next: management built hiring-driven labour increases and quality-strengthening costs into the FY3/27 plan, predecessor-trade delays (labour shortage) already shifted JPY 500m of FY3/27 revenue, and Middle-East-driven materials risk is explicitly excluded from the forecast as unquantifiable. Steel supply comes substantially through the Nippon Steel group, which is also the anchor shareholder.

内部研究类型：causal_mechanism

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Each point of construction margin is roughly JPY 400m of gross profit a year; margin path, not revenue, decided profit direction in both FY3/26 and H1 FY3/27.

不确定性（原始记录）：No quantified steel- or labour-price sensitivity; supplier purchase shares and slide-clause coverage not disclosed.

研究原文：Cash and deposits were JPY 13,774m at September 2026 (about 59% of the JPY 23,463m market capitalisation) with no bank borrowings (lease obligations of about JPY 70m only) and an equity ratio of 71.5%. FY3/26's apparent investing outflow of JPY 10,438m was 91% a shift into time deposits (JPY 9,500m), with only JPY 797m of tangible fixed-asset purchases, and a further JPY 500m went into time deposits in H1; headline cash-flow reads that miss this invert the liquidity picture. Land is carried at JPY 3,908m with a JPY 2,284m revaluation reserve; investment securities are negligible (JPY 11m), so there is no cross-shareholding portfolio to unlock.

内部研究类型：balance_sheet_structure

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Downside protection is as high as Japanese listed contractors offer, removing refinancing and covenant risk; the price is a capped ROE.

不确定性（原始记录）：Cash is consolidated-equivalent and unencumbered as far as disclosed; interim balances are point-in-time.

研究原文：Profit converts to cash with working-capital noise: FY3/26 operating cash flow JPY 2,614m covered net income JPY 2,645m; H1 FY3/27 swung to +JPY 653m from -JPY 1,558m on receivables collection (receivables and contract assets JPY 10,288m, electronic receivables JPY 3,203m against payables of about JPY 6,135m). Capex runs near depreciation (JPY 940m of asset additions vs JPY 780m of depreciation), leaving free cash flow before dividends of roughly JPY 1.7bn; the productive asset is the organisation and order book, not the plant. Working capital explains quarterly cash noise and cannot decide the case.

内部研究类型：cash_economics

内部来源标记：COVERAGE35

普通股东含义（原始记录）：The dividend is paid from cash flow, not from the balance sheet; growth does not consume capital.

不确定性（原始记录）：Customer payment terms and plant ages/remaining capacity not disclosed.

研究原文：ROE has held a 9.7-11.3% band across FY3/23-FY3/26 (8.0% in FY3/22; 9.7% in FY3/26) while the equity ratio rose from about 62-63% to 68.1% (71.5% at September 2026). Segment assets are only JPY 25,772m of JPY 40,543m total assets, so the operating business earns its margin on a much smaller base than the balance sheet suggests and the cash pile dilutes returns. Book value per share compounded from JPY 1,095.5 to JPY 1,433.2 in four years (about 7% a year) on roughly 50% retention with a fixed share count. Returns are maintained despite, not because of, the cash weight; valuation therefore turns on capital allocation rather than operations. Management has published no cost-of-capital estimate, so whether ROE clears cost of capital is borderline by construction.

内部研究类型：capital_efficiency_judgment

内部来源标记：COVERAGE35

普通股东含义（原始记录）：Shareholders earn a bond-plus return on book while the operating business earns equity returns; the gap is the prize of any capital-policy change.

不确定性（原始记录）：Operating-base ROIC not formally computed; management cost-of-capital estimate not public.

研究原文：Shareholder return is a dividend-only formula at about 50% payout (49.8% FY3/25, 50.3% FY3/26, about 50% FY3/27E), dividend-on-equity 4.9-5.6%, split-adjusted JPY 69 for FY3/26 and JPY 65 forecast for FY3/27 (interim raised JPY 30 to 31 mechanically with the profit revision), an interim dividend introduced in FY3/26 and a 5-for-1 split in October 2025. Buybacks are nil and treasury stock sits static at 522,315 shares (2.6%); there is no dilution machinery (19,800,000 shares, no potential shares). Facing sub-1.0 price-to-book and a large cash pile, management revealed a preference for liquidity and income measures over buybacks. The payout is sustainable through a downturn: even at the FY3/22 trough (net income about JPY 1.65bn) the cash position covers the dividend many times over. FY3/26 cash dividends of JPY 2,104m exceeded declared dividends (JPY 1,330m) purely on timing of the FY3/25 final and first interim.

内部研究类型：capital_policy_behaviour

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Dilution risk is zero and income is reliable; the real exposure is that the formula never changes and total-return holders wait on an unsignalled policy move.

不确定性（原始记录）：Whether any unused buyback authorisation exists is not stated in the sources read.

研究原文：The single disclosure that would most change valuation structure is a capital-policy move beyond the roughly 50% payout (buyback, higher payout or a published cost-of-capital framework). The TSE cost-of-capital and share-price awareness reform targets companies persistently below 1.0 times book, a group this stock (about 0.8x) belongs to, and is the most plausible external trigger; yet no framework has been announced, which is itself information about management's posture. Unresolved: whether management would ever deploy the JPY 13,774m cash into buybacks or higher payout, and whether FY3/28 guidance (April 2027) introduces any shareholder-return or capital-cost framework.

内部研究类型：research_debt_catalyst_uncertainty

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Moves the downside floor and the ROE ceiling together; decides whether the stock is merely defended or re-rated.

不确定性（原始记录）：Undated, low observed probability per year; evidence that would resolve it is the April 2027 full-year release or any standalone capital-policy notice.

研究原文：The register is anchored: Nippon Steel holds 32.31% and Nippon Steel Bussan 6.66% (about 39% combined; trust-bank custody accounts about 3.6%), individuals hold 42.8% of units, other corporations 41.34%, financial institutions 3.53%, and foreign ownership has fallen from 14.0% (FY3/24) through 11.6% (FY3/25) to 8.94% (FY3/26), with 10,758 holders of record. The anchor is also the core of the steel supply chain, so supply and register stability are one relationship. Stability removes takeover optionality and abrupt register change, embeds a related-party discount and makes the dividend the binding contract with minorities. The institutional bid left during the period of the price re-rating; why, and whether the shift toward domestic individuals changes pressure for capital-policy reform, is unresolved. President Eiichi Aoki's disclosed behaviour is consistent (conservative guidance, quality investment, no financial engineering); board composition and executive pay were not verified.

内部研究类型：ownership_governance_structure

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Minorities are price-takers in a controlled register; return is dividend plus whatever re-rating patience buys, and governance pressure to change capital policy is low.

不确定性（原始记录）：Commercial terms with the group, identity of large individual holders, board composition and reasons for foreign exit are not disclosed.

研究原文：Normalised earnings power is ordinary profit of roughly JPY 3.7-4.1bn (FY3/25 JPY 4,139m, FY3/26 JPY 3,843m, FY3/27 forecast JPY 3,700m) and net income of roughly JPY 2.4-2.9bn on revenue of JPY 46-48bn; net income implied by split-adjusted EPS rose in a staircase from about JPY 1.65bn (FY3/22) to 2.37, 2.59, 2.94 (peak FY3/25) and 2.65bn. The band's lower edge (about JPY 3.7bn) is the prudent anchor; earnings are backlog volume times a defended construction margin, both near observed highs. A starts-driven order fall would lower the band with a year's lag.

内部研究类型：normalized_economics

内部来源标记：COVERAGE35

普通股东含义（原始记录）：At the lower edge the stock still trades under ten times earnings with the cash floor beneath it.

不确定性（原始记录）：Ordinary profit for FY3/22-FY3/24 not separately verified; EPS series only bounds it.

研究原文：Valuation at the 6 November 2026 close (research cut-off 9 November 2026): share price JPY 1,185; market capitalisation JPY 23.46bn (x 19,800,000 issued shares); forecast PER 9.2x on revised FY3/27 EPS JPY 129.16 (about 8.6x on the quote page); actual PBR 0.83x on FY3/26-end net assets per share JPY 1,433.20 (0.81x on the aggregator); FY3/26 ROE 9.7% as reported (non-consolidated); forecast dividend yield 5.49% on JPY 65.00 (5.82% on the FY3/26 dividend of JPY 69); EV/EBITDA 2.1x (EV JPY 9.69bn after net cash JPY 13.77bn at 30 September 2026; EBITDA JPY 4.56bn = FY3/26 operating profit plus JPY 780m depreciation). Look-through: the market pays about JPY 9.7bn for the operating business against ordinary-profit power of JPY 3.7-4.1bn, so the discount is a governance and capital-allocation discount, not an earnings discount. Basis note: PBR uses year-end book against a September price, PER uses company forecast EPS, and EV treats all deposits as available cash.

内部研究类型：valuation_state

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Priced as a low-growth contractor despite a record backlog; 59% of the price is cash.

不确定性（原始记录）：Mixed period bases (September balance sheet cash vs March book); land market value unknown and excluded from the cash floor.

研究原文：The price embeds low expectations: 8.6-9.2x forecast earnings with net cash at 59% of market value assumes either that the backlog disappoints or that the cash is never returned; failure of either assumption is upside. No analyst consensus was found in the sources read, and retail flow, not institutions, sets the marginal price: the 30 October 2026 package (record backlog, forecast raise, interim dividend raise) produced no upward response, the close drifting to JPY 1,185.

内部研究类型：expectation_mismatch

内部来源标记：COVERAGE35

普通股东含义（原始记录）：Contrarian positioning exists because sponsorship is absent, which is also why the discount can persist for years.

不确定性（原始记录）：Expectations inferred from price, register drift and margin data, not consensus.

研究原文：Principal fundamental risk: the shrinking new-build pool eventually reaches the order book. Non-residential steel-frame starts are down 10.4% and structurally pressured by labour, cost and demographics; formed-product orders already fell 39.6% in H1 and painting orders were net negative. With backlog cover of about 0.85 years, a continued double-digit decline would turn the order question into a revenue question by FY3/28; the FY3/27 year-end backlog print is the earliest hard evidence. Planned cancellations and postponements on cost inflation are named by management.

内部研究类型：principal_risk

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Cycle effects reach the income statement a year late and diluted; FY3/28 orders, not FY3/27 revenue, carry the verdict.

不确定性（原始记录）：Unresolved: whether year-end backlog holds near record (UNKNOWN until the April 2027 release); industry-wide order and backlog aggregates unavailable.

研究原文：Renovation of buildings completed 20 or more years ago is the structural offset to shrinking new-build starts: the installed base of metal roofs generates demand as it ages, and the FY3/26 release explicitly credits renovation capture and formed-product orders for holding orders flat. Its absolute size, its share of orders and backlog, and the pace at which it must grow to offset new-build decline are not disclosed. Renovation likely prices on condition and urgency rather than tender competition, which would also support margin (inference).

内部研究类型：structural_offset_uncertainty

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：A stock-versus-flow race: long-run revenue depends on renovation overtaking the new-build decline.

不确定性（原始记录）：Share of renovation in orders unmeasured (UNKNOWN); order-mix table in each release is the best available evidence.

研究原文：Accounting quality is clean on the evidence read: Japanese GAAP, non-consolidated, no policy or estimate changes or restatements in FY3/26 or H1 FY3/27, negligible extraordinary items (net -JPY 3m), a negligible construction-loss provision (JPY 1m) and a falling warranty provision (JPY 122m); uncompleted construction spending and materials fell modestly. Revenue is recognised over time by progress. Profit can be taken at face value for valuation purposes.

内部研究类型：accounting_quality

内部来源标记：COVERAGE35

普通股东含义（原始记录）：Reported profit is a reliable base for the normalised earnings and valuation bridge.

不确定性（原始记录）：Detailed audit notes (retirement-benefit assumptions, tax reconciliation) were not examined.

研究原文：Dated catalysts: the Q3 FY3/27 release (expected late January 2027, inferred from the company's calendar) tests H2 conversion progress; the April 2027 full-year release brings FY3/28 guidance, final dividend and any capital-policy statement. Each resolves one valuation doubt in sequence: conversion first, then guidance and capital. The company's calendar gives roughly two checkpoint windows a year.

内部研究类型：catalyst_calendar

内部来源标记：COVERAGE35

普通股东含义（原始记录）：Patience is the cost of the position; interim price moves carry little evidence.

不确定性（原始记录）：Q3 release date is inferred, not announced; no other dated corporate events.

研究原文：Thesis: a debt-free, backlog-rich specialist in large metal roofs with defensible ordinary-profit power of JPY 3.7-4.1bn, priced at about 0.8 times book with 59% of the price covered by cash. It works if backlog conversion holds and is re-rated, not merely defended, if capital policy moves beyond the 50% payout. Two clocks decide: the backlog clock (resolved within two quarters) and the capital clock (resolved only by announcement). Falsification: a year-end backlog decline together with a construction-margin relapse below 21%, or evidence the renovation offset has stalled, which would turn the discount from an anomaly into a fair price; the thesis fails by sequence (backlog decline, margin slip under 21% and continued capital silence together), each alone being absorbable.

内部研究类型：thesis_and_falsification

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Shareholders are paid about 5.5% to wait, with asymmetry from the cash floor below and two falsifiable doubts above.

不确定性（原始记录）：Capital clock has no schedule; falsifiers are checkable on the release calendar.

研究原文：The current market value of land carried at JPY 3,908m (with a JPY 2,284m revaluation reserve in equity, reflecting a past upward revaluation) is not disclosed, and whether any sites are surplus to operating needs is unknown. It is therefore excluded from the cash-based downside floor.

内部研究类型：hidden_asset_unknown

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Potential undisclosed asset value, but not countable without evidence; monetisation would also face tax leakage.

不确定性（原始记录）：UNKNOWN; resolvable only through disposal or property valuation disclosure.

研究原文：Cross-lane reading, built outside the protected Technical structure and without altering it: the September 2025 spike to JPY 1,734 (record monthly volume 4,115,400 shares) has flow-event fingerprints (margin-buy imbalance above 300 times, 463,700 bought vs 1,400 sold at the July 2026 reference; a 32.9% nine-month retracement) laid over a genuine fundamental re-rating (six years of revenue growth, split-adjusted EPS from about 85.5 to a 152.6 peak); the two must not be conflated. The price staircase of annual highs tracks the earnings staircase year by year. The stock sits near the 95th percentile of 310 lifetime monthly closes (October close JPY 1,245; only about 18 closes at or above JPY 1,200) while valuation sits at about 0.8x book with cash at 59% of market value: the tape prices a finished re-rating and the statements price a distressed contractor. The fundamental long leg holds above the 60-month average (JPY 957; June 2026 low JPY 1,164 about 21.6% above it), and the 30 October good-news package drew no price response. The Technical State remains as concluded in the protected structure (altitude unconfirmed); positive fundamentals do not upgrade it.

内部研究类型：cross_lane_semantic

内部来源标记：OPEN_DISCOVERY / TECHNICAL

普通股东含义（原始记录）：Distinguishes how much of the price altitude is capitalised fundamentals versus retail flow; informs the risk that margin-long overhang amplifies any disappointment.

不确定性（原始记录）：Margin balances are a July 2026 reference and may have normalised; how much of the spike was flow versus fundamental is not quantified.

研究原文：Durability rests on three legs with different evidence grades. (1) Backlog conversion, evidenced: a position of site-execution capacity, design-stage relationships and a landmark record turns a shrinking pool into record orders and schedules most of the next year before it starts, so demand shocks reach the income statement a year late and diluted. (2) Margin defence, evidenced but contested: one bad year in five, a full point lost in FY3/26 and 0.9pt repaired in H1. (3) Capital evolution, unevidenced: balance-sheet durability (no debt, cash 59% of market value) was bought with shareholders' returns and the formula (50% payout, no buybacks, no framework) shows no sign of evolving on its own. Not durable: the new-build pool is in structural decline and the renovation offset is real but unmeasured. Failure is a sequence (year-end backlog decline, construction margin under 21%, another year of capital silence) that would reprice the company as a melting niche contractor with a cash museum attached.

内部研究类型：structural_durability_judgment

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Defines what protects and what can break the earnings and the balance-sheet floor.

不确定性（原始记录）：Leg three depends on management choice; leg one depends on the unmeasured renovation share.

研究原文：Backlog growth concentrated in long-span roofing and high-tough raises the quality of the visibility that n06 describes.

内部研究类型：QUALIFIES

内部来源标记：COVERAGE35

研究原文：Conversion of the pre-sold backlog at defended margin decides whether the implied H2 ordinary profit of JPY 2,064m is beaten; slippage is timing, not demand loss.

内部研究类型：CONDITIONS

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：Cost increases outrunning pass-through compressed FY3/26 construction margin by 1.2pt.

内部研究类型：CAUSED

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：Price pass-through and cost improvement restored 0.9pt of construction margin in H1 FY3/27, though labour build-up remains a forward margin risk.

内部研究类型：PARTIALLY_REPAIRED_BY

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：The cash pile is simultaneously the downside floor and the ROE ceiling: it lowers return on equity from what the operating asset base earns.

内部研究类型：DILUTES

内部来源标记：OPEN_DISCOVERY

研究原文：Unchanged dividend-only capital policy with the cash retained explains the sub-book valuation; the discount is a capital-allocation discount rather than an earnings discount.

内部研究类型：DETERMINES_DISCOUNT

内部来源标记：COVERAGE35

研究原文：The roughly 39% anchor and a retail-heavy, institution-light register reduce internal pressure for capital-policy change; the TSE reform is the main external lever.

内部研究类型：CONSTRAINS_SPEED_OF

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：A shrinking new-build pool does not impair the current backlog but would reduce its replenishment, turning an order problem into a revenue problem by FY3/28.

内部研究类型：THREATENS_FUTURE_REPLENISHMENT_OF

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：Renovation demand from the ageing roof base is the mechanism that may offset new-build decline, but its scale is unmeasured, so the offset is a hypothesis supported by order behaviour rather than a quantified fact.

内部研究类型：MAY_OFFSET

内部来源标记：OPEN_DISCOVERY / COVERAGE35

研究原文：The cross-lane reading takes the protected Technical Synthesis as given and does not redefine it; the cross-lane layer only compares price behaviour with fundamental timing.

内部研究类型：REFERENCED_AS_WHOLE_BY_CROSS_LANE

内部来源标记：TECHNICAL / OPEN_DISCOVERY

研究原文：A price at the 95th percentile of lifetime closes coexists with 0.8x book and cash at 59% of market value; the shelf is where the contradiction between the tape and the statements is being settled.

内部研究类型：IN_TENSION_WITH

内部来源标记：OPEN_DISCOVERY / TECHNICAL

研究原文：A record pre-sold backlog and a price that has not fallen back into its old base both say the next year is spoken for; both resolve on conversion at defended margin through the January and April windows.

内部研究类型：PARALLEL_EVIDENCE_REGISTER

内部来源标记：OPEN_DISCOVERY / TECHNICAL

研究原文：The Q3 and full-year releases supply the checkable evidence for the thesis's backlog and capital clocks.

内部研究类型：TESTS

内部来源标记：COVERAGE35

研究原文：The absence of accounting distortion allows reported profit to anchor the normalised earnings band.

内部研究类型：SUPPORTS_RELIABILITY_OF

内部来源标记：COVERAGE35

研究原文：Cash-backed earnings, light capex and 50% retention add to the deposit pile each year unless payout rises.

内部研究类型：FEEDS

内部来源标记：COVERAGE35

研究原文：Completion-weighted earnings make the January Q3 and April full-year prints the only hard checkpoints each year.

内部研究类型：DETERMINES_TIMING_OF

内部来源标记：COVERAGE35

研究原文：Formed-product orders (-39.6%) and product margin compression may be an early signal of the shrinking pool, though attribution is unresolved.

内部研究类型：POSSIBLE_EARLY_SIGNAL_OF

内部来源标记：OPEN_DISCOVERY

研究原文：Protected Technical Synthesis (RESEARCH_TRUTH.technical_state)

内部研究类型：protected_technical_synthesis

内部来源标记：TECHNICAL

不确定性（原始记录）：受保护的 Technical Research 结构；公开边界以其正文路由为准。

公开边界：受保护 Technical Research 仅保留可追踪身份，本索引不重写其内部语义。

研究原文：Durable Business Portrait: integrated roofing contractor with a pre-sold book, no debt and an unchanged capital formula｜Durable foundation: an execution-based niche position converts a shrinking pool into record orders; the backlog times the defended construction margin gives earnings; cash-backed conversion at light capex and a debt-free balance sheet give resilience; a permanent anchor holder and an unchanged 50% payout formula fix the capital outcome. Current change (H1 margin repair, record backlog, forecast raise) is an increment to this foundation.

内部研究类型：durable_business_portrait

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：A high-floor, low-optionality holding whose returns are income plus whatever capital-policy change or conversion beats eventually unlock.

不确定性（原始记录）：Moat unverifiable in share terms; renovation offset unmeasured; capital evolution unevidenced.

研究原文：Order mix to backlog to second-half conversion to margin: how demand reaches shareholders｜Starts decline, but order mix (large projects, renovation, share gain) fills the backlog; the backlog schedules revenue 0.85 years ahead, so the P&L impact of the cycle is delayed; execution and pass-through determine margin; the pool's decline reaches orders and then revenue by FY3/28 unless renovation offsets it.

内部研究类型：main_causal_chain

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：FY3/27 is substantially pre-sold; the year-end backlog and margin prints determine whether the discount is an anomaly or a fair price.

不确定性（原始记录）：H2 scheduling, backlog margin profile and renovation share are undisclosed.

研究原文：Cash as downside floor and ROE ceiling in one balance｜The JPY 13.8bn deposit pile is simultaneously the downside floor (59% of market value) and the ROE drag (about 10% ROE while operations earn far more on a smaller base); any capital-policy change moves floor and ceiling together, making it the highest-leverage disclosure the company could make.

内部研究类型：cross_domain_mechanism

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Re-rating depends on capital allocation, not operations; the floor is real but already counted by the market.

不确定性（原始记录）：No cost-of-capital framework; land value unknown and excluded from the floor.

研究原文：Nippon Steel anchor: stability, supply and the permanent discount are one position｜The roughly 39% holding stabilises register and steel supply in the same act, removes takeover optionality from the valuation and sets a practical ceiling on how fast capital policy can change.

内部研究类型：cross_domain_mechanism

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Minorities hold alongside a permanent anchor: stability is high, optionality low; discount and stability are not separable.

不确定性（原始记录）：Commercial terms with the group and reasons for foreign exit are not disclosed.

研究原文：Seasonality as verification calendar: two checkpoints a year｜With Q1 under a tenth of annual profit and Q4 about 35%, the January and April prints are the only hard tests; interim price moves carry little information.

内部研究类型：cross_domain_mechanism

内部来源标记：COVERAGE35 / OPEN_DISCOVERY

普通股东含义（原始记录）：The case is testable on a fixed, sparse schedule.

不确定性（原始记录）：Q3 date inferred from the calendar.

研究原文：Renovation mix links order durability to construction margin｜The ageing-base shift that defends order volume plausibly also defends construction margin because renovation and cover-method work price on condition and urgency rather than new-build tender competition; if the mix shift stalls, volume and margin weaken together. This is an inference; the order-mix table in each release is the most informative evidence.

内部研究类型：cross_domain_mechanism

内部来源标记：COVERAGE35 / OPEN_DISCOVERY

普通股东含义（原始记录）：Renovation share is the single most informative variable for both orders and margin.

不确定性（原始记录）：Margin treatment of renovation work is not disclosed (inference).

研究原文：Normalised earnings to look-through valuation to embedded expectation｜Ordinary-profit power of JPY 3.7-4.1bn on clean accounts, against an operating business priced at about JPY 9.7bn after cash, at 8.6-9.2x earnings and 0.8x book, means the price assumes backlog disappointment or permanent cash immobility; either failing is upside.

内部研究类型：valuation_bridge

内部来源标记：COVERAGE35 / OPEN_DISCOVERY

普通股东含义（原始记录）：Downside cushioned by cash; upside requires one of two doubts to fail.

不确定性（原始记录）：Mixed period bases; land value excluded.

研究原文：Thesis: two clocks, with falsification by sequence｜The backlog clock resolves within two quarters; the capital clock only by announcement. Failure requires a combination: year-end backlog decline, construction margin below 21% and continued capital silence.

内部研究类型：thesis_and_falsification_structure

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：A 5.5% yield is paid for waiting; asymmetry comes from the cash floor and two falsifiable doubts.

不确定性（原始记录）：Capital clock unscheduled.

研究原文：Material unresolved questions carried forward｜Highest-value debt: H2 backlog schedule and locked-in margin; year-end backlog versus the shrinking pool; renovation share; capital-policy framework. Lower-value: product-margin cause, land value, foreign exit reasons, flow versus fundamental share of the 2025 spike. Each is resolvable mainly through the January and April 2027 releases or by disclosure the company has not yet made; no reasonable current public path was identified.

内部研究类型：research_debt

内部来源标记：OPEN_DISCOVERY / COVERAGE35

普通股东含义（原始记录）：Open items condition confidence in conversion and the re-rating path rather than the floor.

不确定性（原始记录）：Unresolved items stay UNKNOWN; no reopen is warranted now.