How to Read a Kessan Tanshin

Section 1What a kessan tanshin is

A kessan tanshin (決算短信, literally a settlement-of-accounts brief report) is the earnings summary every company listed on a Japanese exchange publishes at the end of each quarter and each fiscal year. It is the first and fastest official financial disclosure a Japanese listed company makes, and for most of the gap between one filing and the next it is the only complete set of financial statements available.

Four properties define it.

It is an exchange rule, not a securities law filing. The tanshin exists because the exchange requires it under its timely disclosure rules, not because the securities statute mandates it. That distinction drives everything else about the document.

It is unaudited. The tanshin carries an explicit checkbox stating whether the attached statements have been reviewed by an accountant. On a quarterly filing the box normally reads 無, no review. That is the standard condition of the document rather than a signal about the issuer.

It is fast. Companies target publication within roughly six weeks of period end, which is why it arrives long before the audited statutory filing.

It is a summary with statements attached. The first page is a standardised cover sheet: revenue, operating profit, ordinary profit, net profit attributable to owners of the parent, per-share figures, total assets, net assets, equity ratio, dividend status, and full-year guidance. Behind it sits an attachment (添付資料) carrying management's discussion, the balance sheet, the income statement, the comprehensive income statement, and a limited set of notes.

Section 2Cadence, and where it publishes

Four filings per fiscal year: three quarterly and one full-year. The full-year document is titled simply 決算短信 with no quarter marker, while quarterly ones carry 第１四半期, 第２四半期 or 第３四半期. A company's cadence follows its own fiscal year end, and a fiscal year end can change. Any historical series that runs back far enough will cross such a change, and the periods either side are not comparable in length.

The tanshin is transmitted through TDnet (適時開示情報伝達システム, the timely disclosure network) operated by the exchange group. The moment of transmission is the moment the information becomes public.

Section 3How it differs from the securities report

Japanese listed companies also file a yūka shōken hōkokusho (有価証券報告書, annual securities report) and, at the half-year mark, a hanki hōkokusho (半期報告書). These are statutory filings submitted to EDINET, the regulator's electronic disclosure system, which is a different system from the exchange's.

One timing note that catches readers who learned the Japanese system some years ago: the separate quarterly securities report for the first and third quarters was abolished, leaving the tanshin as the primary quarterly disclosure vehicle with expanded content requirements, and the semi-annual report occupying the second-quarter slot.

Section 4Where the bitcoin lives

It does not sit in current assets. It sits at the bottom of the non-current section, inside investments and other assets.

固定資産            Non-current assets
  有形固定資産        Property, plant and equipment
  無形固定資産        Intangible assets
  投資その他の資産      Investments and other assets
    ビットコイン        Bitcoin        <- here
    繰延税金資産
    その他

Two things follow immediately.

The line is a single number and it is the whole position. There is no split between pledged and unpledged, none between directly held and subsidiary held, and no cost against market reconciliation on the face. Everything needed to separate collateralised coin from free coin is absent from the quarterly document.

Revaluation runs through non-operating expense, not operating profit. A valuation loss on the stack (ビットコイン評価損) appears under 営業外費用, which places it below operating profit and above ordinary profit. The result is the headline shape that confuses first-time readers: operating profit can rise sharply while the same quarter reports a very large net loss, from the same document. Operating profit is the operating business. The loss is the mark on the stack. Neither number contradicts the other, and neither is a per-share coin input.

Section 5Between filings: the timely-disclosure layer

Holdings move continuously and the tanshin reports quarterly. The gap is filled by timely-disclosure announcements, and those are the documents carrying the numbers between filings.

Cadence changes are invisible unless you look for them

An issuer can move from a notice per acquisition, sometimes weekly, to a single consolidated notice announcing the completion of a quarter's accumulation. Nothing announces the change. It shows up as a gap in the date column of the historical holdings table carried inside every notice: frequent entries, then a long silence, then one entry per quarter.

A series built on the assumption of per-purchase disclosure will show a spurious flat line from the change onwards, and it will date the change to the first notice that looks consolidated rather than to the first consolidated entry, which is usually a quarter earlier. The fix is in the same document: the historical table inside the latest notice carries the full dated series in one place, on one basis.

Section 6Bonds payable, and a balance sheet that describes one instant

Under Japanese presentation, bonds payable split across two lines by remaining term at the balance sheet date:

The split moves without any transaction occurring. A bond issued at three years crosses from non-current to current on its own, two years in. If you track the mix as a proxy for refinancing pressure, that reclassification is a real signal, and it is not an event and will not be announced.

A short-dated bond never touches non-current at all. An instrument issued just inside a year books as current on every balance sheet it ever appears on. Two implications: the 社債 line can read nil while current liabilities carry the whole bond programme, so no 社債 is not no bonds and both lines have to be read. And for anyone scoring permanence, a bond structured just under a year is rollover-dependent by construction. Its current classification is a permanent feature of the instrument rather than a signal of approaching maturity pressure.

The subsequent-events note does not rescue that. 重要な後発事象 can read 該当事項はありません, not applicable, while a substantial issuance sits between the balance sheet date and the filing date, because the note carries a materiality threshold that such an issuance need not clear. The subsequent-events note is not a reliable bridge from the balance sheet date to today. The timely-disclosure stream is.

Read the redemption clause, not the maturity date

Two terms commonly sit in the redemption clause of these instruments, and neither is visible from a summary table.

The bondholder can hold a put. On a few business days' written notice the holder may demand early redemption at par, with no lockout and no make-whole. Whatever the maturity date says, that is demand paper at the holder's election.

Redemption can be triggered automatically by the holder's own subsequent funding. Where the same counterparty holds warrants, the terms can require the company to retire bonds in fixed increments each time cumulative amounts paid in by that holder cross a threshold. The bond is then self-liquidating against the warrant exercises, with no separate decision by either party. Where a prior series in the same programme was retired that way, the pattern is documented in primary filings and worth reading before treating a stated maturity as the instrument's life.

Two facts that must not be bridged

Bonds are not the only debt layer. Alongside them can sit collateralised borrowing, which does carry interest and does show up in non-operating expense.

Where the balance sheet states short-term borrowings in yen at the period end, and the narrative states a facility drawdown in dollars as of the filing date, those are two different dates, two different currencies and two different measurement moments. There is no conversion that turns one into the other, and any figure produced by applying a rate to one and comparing it to the other is fabricated. The narrative figure is a subsequent-period liquidity disclosure. The balance sheet figure is the period-end obligation. Use the balance sheet figure for the period-end claim.

Section 7The claim that is not on the balance sheet

This is the section the page exists for. Read a Japanese net-assets section, and look for the preferred stock.

It is not there. There is no line for it. An issuer can have a class of preferred shares outstanding, described in the same document's narrative as 永久資本, perpetual capital, and paying a dividend that quarter, with no caption of its own anywhere on the face.

Compare a US filer, where preferred stock takes its own line in stockholders' equity with par value and liquidation preference disclosed on the face or immediately in the notes. There is no Japanese equivalent of that presentation in the quarterly document.

Where the class shares do appear

So the tanshin gives you the existence and the dividend cost of the class. It does not give you the claim size. The claim size is the liquidation preference, and the liquidation preference is not a tanshin figure.

Two term-sheet features that change the instrument

The preference can be a formula rather than a number. A per-share base amount, plus accumulated unpaid dividends, plus an accrued dividend equivalent to the distribution date pro-rated by days. Where the base amount per share exceeds the cash raised per share, the claim was never the money, and a model carrying the proceeds figure as the claim understates it from day one.

Arrears can compound. Where a quarter's dividend falls short, the shortfall can accumulate and accrue interest at a stated rate until paid, with accumulated arrears payable in full before any current dividend. That is a contractual claim-growth mechanism, it scores as one, and a model holding the claim flat understates it in exactly the state where the claim binds.

And read the word perpetual against the term text. A class described as perpetual capital can carry a holder-side resolution date, at which point neither the debt layer nor the preferred layer is permanent on the holder's side, whatever the framing suggests.

Section 8Share counts, and three lines with two bases

Share counts appear in a small block on the second cover page, headed 発行済株式数（普通株式）. Three lines, each measuring something different.

The denominator is line one minus line two. Line one includes treasury stock, so using it raw overstates the count. Where a treasury block is trivially small the error is invisible, which is precisely how it becomes a habit that breaks on a company running a real buyback. Subtract every time.

Line three is a weighted average for the income statement. It is an earnings-per-share denominator, not a balance-sheet denominator.

Preferred shares are in none of these lines. 発行済株式数（普通株式） means common stock. A class share count appears only in the key-metrics table in the management discussion.

The company's diluted figure answers a different question

A tanshin key-metrics table can report a 完全希薄化後発行済株式数, a fully diluted issued share count, and use it as the denominator for the company's own per-share coin metric. That number is constructed to the company's own definition: issued common shares, plus shares from conversion of instruments whose proceeds have already been received, plus options, while excluding whole series of stock acquisition rights and moving-strike warrants until exercise and payment.

Two consequences. It is not a basic count, and substituting it into a per-share coin calculation understates the result by whatever the gap between the two denominators happens to be. And the excluded warrant overhang can be large and is disclosed elsewhere in the same document, so the exclusion is visible to anyone who looks for it.

Share counts move between filings through the monthly warrant exercise and treasury purchase notices, both typically published on the first business day of the following month. The quarterly purchase notice also restates issued and diluted counts at each quarter end.

Section 9The yen layer, and the price that rebases

A yen reporting company states every consolidated statement figure in yen and only in yen. Any dollar equivalent you have seen for such a company's financials was produced by someone applying a rate the filer did not supply.

The exception matters more than the rule. A tanshin narrative can state a dollar denominated facility in dollars. That is not a conversion. It is the obligation as contracted, stated in its own currency inside a yen filing. Converting it into yen to sit alongside the balance sheet is the reader's construction, and it carries every basis requirement that any other conversion carries. Both directions need a stated basis, and neither direction is safe to eyeball.

Two patterns follow, and both travel past Japan. A crossable pair beats a quoted rate found elsewhere, because the implied rate is the issuer's own and is reproducible from the filing alone. And a filer's own conversion is document scoped. Using a rate as stated means as stated in that document. It does not license carrying that conversion into a series across documents.

The two-prices trap

A single quarter's disclosures can contain several different coin prices in yen, and none of them is labelled as the odd one out. The carrying value divided by the coin count implies one price. The key-metrics table states a reference price. The narrative's market value as of the filing date implies a third.

The middle one is the trap. A reference price defined in a footnote as the latest daily closing price on a named venue is applied across every column of the table, including columns headed by dates a year earlier. It is not a period-end price. It is a constant. The definition of the price and the reason it is held constant commonly live in two different footnotes, so a reader who reads only the one attached to the price learns what it is and not that it is deliberately frozen.

The rebasing, which silently restates history

The reference price is constant within a document. It is not constant across documents. When a later notice adopts a new reference, every historical yen gain figure in the table changes with it, including figures for quarters that closed long before and cannot have moved. Every row scales by exactly the ratio of the two reference prices. The underlying coin gains are unchanged. The yen series was restated wholesale by a change of measurement unit.

The lesson generalises: a figure labelled "for comparability" is comparable within the document that defines it and nowhere else. A comparability note is a scope statement, not a guarantee.

The translation layer inside the carrying value

A balance sheet carrying a material 為替換算調整勘定, a foreign currency translation adjustment, is telling you that assets sit in subsidiaries whose functional currency is not the yen. The quarterly document generally does not say which subsidiaries, and the argument does not need it: the adjustment is sufficient evidence on its own.

The consequence is that the consolidated coin line is a blend: coin held directly in yen, plus coin held in non-yen functional subsidiaries translated at the period-end rate. You cannot back a clean market price out of it, because part of it already has a currency translation baked in. The movement in the translation account is a currency effect on the position and it flows through other comprehensive income rather than through the valuation loss in the income statement, which is why comprehensive loss and net loss differ by exactly that amount. A model reading only the income statement misses it.

Section 10The split, as a worked trap

Japanese presentation requires that per-share figures and share counts in comparative periods be restated as if a split had occurred at the beginning of the earliest period presented. Every filing after the effective date shows pre-split periods on a post-split basis. The document does not shout about it. It is a one-line note at most.

Why that is a trap and not a footnote: the same period yields two different per-share figures depending on the basis, and only one of them is printed in the document you are holding. The loss for the period never changes. The denominator does. Divide the same result by the pre-split and post-split weighted averages and you get two per-share numbers an order of magnitude apart, both correct on their own basis.

How it breaks a series

The failure mode is not using the wrong number. It is building a series by pulling each period from the filing that first reported it, which is the natural research workflow, and which produces a series with a silent tenfold discontinuity at the split. The chart shows a collapse in per-share coin on the effective date. There was no collapse. There was a split.

The same failure runs in reverse on share counts. A key-metrics table shows a retroactively adjusted count at a date on which the actual register read the unadjusted figure. Both are "the share count" at that date.

Two further hazards sit around the edges. The count in a split announcement is not the count the split applied to, because shares are issued between announcement and record date, so take split reference counts from the record date. And a forward split can be preceded by a reverse split in the opposite direction; the two do not commute with intervening issuances, so a series crossing both needs each period's adjustment factor derived explicitly rather than assumed to cancel.

Section 11Reading order

Then, and only then, assemble. A capital structure built from this filing in any other order will be missing the preferred stack, because the preferred stack is the one thing the balance sheet does not carry.