The Real Cost of Manual Seikyusho Processing
for Japanese SMEs

The Subscription That Hides Thirty-Eight Hours of Labour

Every cost analysis of manual invoice processing begins with the same mistake: measuring the visible subscription fee and declaring that the problem costs ¥12,000 a year. A Japanese accounting app — freee (¥12,936 per year, tax-inclusive), Yayoi (弥生, ¥11,800), or MoneyForward Cloud (マネーフォワード クラウド, ¥11,880, as listed on each provider's pricing page) — appears on the credit card statement as a single line item. The statement does not itemise the thirty-eight hours a month the app's automatic journal entry (自動仕訳) feature cannot handle, because the bank feed never sees the paper seikyusho that arrived by mail, or the PDF attachment with the scanned bank stamp (銀行印), or the supplier whose invoice layout changed this month and the OCR engine the app bundles has never seen before.

The accounting app automates the journal entry for transactions it can match to a bank feed. It does not read a paper seikyusho, does not know which line items require gensen choshu withholding, and cannot verify whether a T+13-digit invoice registration number is still valid on the NTA's Qualified Invoice Issuer public database. Every one of those steps lands on a person — and that person's time is the cost the ¥1,000 monthly subscription was priced not to cover.

The problem is structural. A Japanese seikyusho contains fields that no Western invoice format includes: the payment terms expressed as a shimebi (締め日, the monthly cut-off date) plus a payment window (翌月末払い meaning payment by the end of the following month), the furikomi bank details (振込先 — bank name, branch name, account type, account number, account holder name) that must be transcribed character-for-character to avoid a failed transfer, the shohizei (消費税) broken out by tax rate (10% standard vs. 8% reduced), and since October 2023, a T+13-digit invoice registration number (インボイス制度登録番号) that the buyer must verify to claim the input tax credit. An accounting app automates none of these checks. The monthly subscription reduces the writing of a journal entry to a click, not the reading of a seikyusho to a glance — and the reading is where the thirty-eight hours go.

Four Cost Layers That No Global Benchmark Captures

Industry studies put manual invoice processing at $15 to $26 per document globally — a figure cited in Ardent Partners and IOFM research. That number folds labour, error correction, and overhead into one average, and its frame of reference is the US and European AP department. Drop it into a Japanese SME — where the invoice is a seikyusho with furikomi details that can trigger a failed bank transfer, where certain line items trigger a 10.21% withholding obligation that, if missed, becomes the company's tax liability plus penalties, and where every invoice since October 2023 carries a registration number the buyer must verify or lose the input tax credit — and the cost structure looks different. Four cost layers that no global benchmark captures:

To keep the arithmetic grounded, the calculations that follow use a single example: a Japanese SME (中小企業) with ¥500 million annual revenue, an AP team of two people, and 300 supplier seikyusho per month arriving as paper by mail and as PDF attachments by email. Substitute your own volume, labour rate, and error experience at each step.

Layer One — What an Hour of Manual Seikyusho Entry Costs in Japan

The reason manual data entry labour is consistently underestimated in AP cost analyses is that no invoice arrives for your own staff's time. But an hour spent opening a seikyusho, reading each field, typing it into the accounting system, and cross-checking the furikomi details against the supplier master data is an hour not spent on higher-value work — and that hour has a market price in Japan that is well-documented.

Japanese bookkeeping and AP staff earn ¥1,800 to ¥2,800 per hour as dispatch workers (派遣社員). Experienced AP staff in Tokyo and Osaka command ¥2,500 to ¥3,500 per hour with full burden. A bookkeeping outsourcing service (記帳代行) charges ¥30 to ¥80 per journal entry on a volume basis — which, for a seikyusho with five line items each requiring separate journal entries for the purchase amount, consumption tax, and withholding, translates to ¥150 to ¥400 per invoice in outsourced data entry alone. A tax accountant (税理士) on a monthly advisory contract (顧問契約) costs ¥30,000 to ¥40,000 per month for a small business — and that typically includes reviewing the books, not entering the invoices from scratch. If the accountant has to re-enter or correct manually entered seikyusho data before closing the month, the fee reflects it.

The time-per-invoice number varies by complexity, but a usable range exists. A simple seikyusho — single line item, standard 10% consumption tax, no withholding, furikomi details that match the supplier master — takes three to four minutes to enter and verify. A complex seikyusho — multiple line items with mixed tax rates (10% and 8%), a withholding-eligible service component, furikomi details that differ from the previous month's, and an invoice registration number that needs verification — takes eight to twelve minutes. At 300 invoices per month split 70% simple and 30% complex: 210 simple invoices at four minutes each (fourteen hours) plus 90 complex invoices at ten minutes each (fifteen hours) equals twenty-nine hours. Add the month-end reconciliation — checking that the total entered matches the supplier statement (請求書一覧表), chasing missing invoices, correcting mis-entered shimebi dates — and the monthly total lands at thirty-two to thirty-eight hours.

Layer Two — The Cost of One Wrong Digit in a Furikomi Block

A Japanese seikyusho carries a payment instruction block that has no equivalent in Western invoice formats. The furikomi block (振込先) specifies a bank name (銀行名), branch name (支店名), account type (普通/当座), a seven-digit account number (口座番号), and an account holder name (口座名義) that may use katakana (カタカナ) or kanji. A manual AP process means a person reads these fields off the seikyusho and types them into the internet banking interface — character by character, with no cross-validation against the supplier master data unless the company has built one. A mistyped branch code — 銀座支店 (Ginza branch) entered as branch code 001 instead of 002 — does not prevent the transfer from executing. It sends the payment to a different account.

When the supplier calls ten days later asking why the payment has not arrived, the AP team enters the furikomi error chain: locate the original transfer confirmation in the internet banking history, identify which digit was wrong, determine whether the funds landed in a closed or active account, and if the latter, initiate a kumimodoshi (組戻し).

The kumimodoshi procedure varies by bank but follows a consistent structure. Mizuho Bank's published FAQ lists the fee at ¥880 (tax-inclusive) and requires the caller to provide the transfer date, receiving bank and branch, recipient name, and amount. SBI Shinsei Bank warns that the procedure may take two weeks to several months and that the receiving account holder must consent — if the recipient refuses, the funds are not returned, and the ¥880 fee is still charged. The Rakuten Bank FAQ adds that kumimodoshi is not guaranteed even after the request is filed. These are not edge cases. A Japanese AP team processing 300 seikyusho per month with a conservative 1% furikomi entry error rate — three incorrect transfers per month — faces ¥2,640 in direct kumimodoshi fees per month, plus the staff time spent on each recovery: telephoning the bank, documenting the error, following up with the supplier, and processing a corrective payment. At one hour of staff time per incident and ¥3,000 per hour, the labour component adds ¥9,000. The total monthly furikomi error cost for a 1% error rate: approximately ¥11,640, or ¥139,680 per year.

The supplier relationship cost — a vendor who now asks for payment upfront or shortens the shimebi window — has no line item and matters more than the ¥880 fee. It is also the cost most visible to the procurement team and least visible to the accounting app vendor who priced their product at ¥980 a month.

Layer Three — Gensen Choshu: When a Misclassified Line Item Becomes a Tax Liability

Japan's withholding tax system (源泉徴収制度) requires that payers of certain categories of service income — professional fees to lawyers, certified public tax accountants (税理士), judicial scriveners (司法書士), social insurance labour consultants (社会保険労務士), designers, writers, and speakers — withhold income tax at source. The withholding rate is 10.21% of the payment amount (20.42% on the portion exceeding ¥1 million for a single payment to the same recipient). The withheld amount must be remitted to the tax office by the 10th of the month following the payment.

In a manual AP process, the decision of whether a seikyusho line item triggers withholding rests on a person reading the payment description and knowing the rules. The seikyusho from a web design firm for ¥480,000 — does it include a design service component that is withholding-eligible? The seikyusho from a consulting sole proprietor (個人事業主) — is the entire amount subject to withholding, or only the portion above expenses? An AP clerk who has processed 200 invoices this week, half of them paper seikyusho from suppliers who do not digitally label withholding status, makes a call — and when the call is wrong, the mistake is invisible until the tax office's reconciliation or audit uncovers it.

The consequence of missed withholding flows through a defined penalty structure. The unremitted withholding tax itself becomes the company's liability — ¥48,998 on a ¥480,000 payment at 10.21%. The non-payment penalty (不納付加算税) is 10% of the unremitted amount (5% if the company discovers and corrects the error before the tax office issues a notice). Late payment interest (延滞税) accrues daily from the original statutory due date at the rate set under the General Law of National Taxes (国税通則法), currently 3% per annum (the statutory rate under Article 404 of the Civil Code, effective April 2026 under Ministry of Justice Notification No. 73 of 2025) for the first two months and potentially higher thereafter. On a single missed withholding of ¥48,998 discovered six months late, the company owes: ¥48,998 (principal) + ¥4,900 (10% penalty) + roughly ¥735 (six months at 3% annualised) = ¥54,633. That is 11.4% of the original payment — and the penalty multiplies across every withheld-eligible invoice the AP clerk misclassified over a year.

A separate exposure concerns the contractual delay penalty rate. Many Japanese commercial contracts specify a delay damage rate (遅延損害金) of 14.6% — commonly derived from the rate in the National Tax Collection Act — and while the Civil Code's default statutory rate was lowered from 5% (and the former Commercial Code's 6% 商事法定利率 was abolished in April 2020), a contractually specified rate overrides the statutory default. An SME that consistently misses its own payment deadlines because manual furikomi entry delays the batch — and whose contracts specify a 14.6% delay rate — faces a compounding exposure on top of the direct penalties from the tax office. The legal framework for 2026 delay damages under Japanese law explains the interaction in full.

Layer Four — The Compliance Tax: Verifying Registration Numbers That Were Valid Last Month

The Qualified Invoice System introduced a cost that did not exist in Japanese AP before October 2023: the per-invoice compliance check. To claim an input tax credit on a purchase, the buyer must hold a qualified invoice (適格請求書) bearing a T+13-digit registration number (登録番号) — and that number must belong to a currently registered qualified invoice issuer, not one that was cancelled or expired. The NTA's Qualified Invoice Issuer Public Database (適格請求書発行事業者公表サイト) is the authoritative source, and it requires per-query lookup: enter the registration number, see if the issuer is still active. The database accepts up to ten numbers per batch query.

For a Japanese SME with 200 active suppliers, the compliance check is not a one-time setup. A supplier's registration can be cancelled at any time — voluntarily, or automatically when the business closes or its tax status changes. The Tokyo Chamber of Commerce and Industry (東京商工会議所) advises buyers to verify registration status at each transaction. That means every seikyusho from a supplier whose registration was verified three months ago needs re-verification. A quarterly batch verification for 200 suppliers, at ten suppliers per query and roughly two minutes per query to enter numbers and record results, takes forty minutes. Per year: 160 minutes. If the AP team does it monthly — as recommended — that is 480 minutes or eight hours. At ¥3,000 per hour, ¥24,000 annually.

The cost is small per invoice but large in consequence when missed. Receiving a seikyusho without a valid registration number and claiming the input tax credit anyway means the credit is disallowed on audit — and the disallowed amount becomes the company's additional consumption tax liability plus penalties. On a single ¥1,500,000 seikyusho with ¥150,000 in consumption tax that was claimed but the supplier's registration had expired, the company owes ¥150,000 in back tax plus a potential understatement penalty (過少申告加算税) of 10% to 15% under the General Law of National Taxes. The per-transaction cost of verification is ¥40 to ¥80. The per-transaction cost of not verifying is, at minimum, the input tax credit itself.

Your Total: A Four-Layer Calculation Framework Per Monthly Settlement Cycle

With the four layers separated, the annual cost of manual seikyusho processing becomes a function of your own invoice volume, labour rate, error frequency, and compliance requirements rather than a global benchmark that averages out every market's differences. The table below runs the calculation for the example SME — 300 seikyusho per month, a two-person AP team, ¥3,000 hourly labour rate, 1% furikomi error rate, 2% gensen choshu misclassification rate, and monthly invoice compliance verification:

The accounting app subscription — the only cost most SMEs count — is roughly 1% of the total. To adapt the framework to your own operation, substitute into five expressions:

• Layer One = (monthly invoice count × average entry time per invoice in hours) × your AP staff hourly rate × 12 months
• Layer Two = monthly invoice count × your furikomi error rate × (kumimodoshi fee at your bank + staff recovery time × hourly rate) × 12 months
• Layer Three = monthly invoice count × your gensen-eligible proportion × estimated misclassification rate × average withheld amount × penalty rate (5%–10%), annualised with audit probability
• Layer Four = (number of active suppliers ÷ 10 suppliers per batch query) × minutes per batch ÷ 60 × hourly rate × queries per year
• App subscription = your current accounting software annual fee (¥12,000–¥17,000 for SME tier)

The same cost structure — invisible downstream costs dwarfing visible subscription fees — appears across markets that depend on manual document data entry. The companion article on what manual BAS entry costs Australian small business owners found the same pattern: the labour you see is less than half the total cost. The analysis of manual passbook transcription costs for Japanese sole proprietors identified three cost lines hiding behind a ¥1,000 monthly subscription. The jurisdiction and the document type change, but the cost architecture — visible labour plus error correction plus compliance friction plus penalty exposure — is remarkably consistent.

What Happens When the Typing Step Disappears

Every one of the four cost layers traces back to the same operation: a person reading a seikyusho and typing its fields into a screen. Layer One is that operation multiplied by 300 invoices every month. Layer Two is what happens when the furikomi block is mistyped. Layer Three is what happens when the withholding decision is made by a person under time pressure rather than by a consistent rule applied to every invoice. Layer Four is what happens when compliance verification is a manual batch query rather than an automatic lookup. Remove the manual reading-and-typing, and all four layers compress.

The mechanism is Custom Column Extraction: you define the columns you want — "Invoice Date (請求日)," "Supplier Name (請求元)," "Line Item Description (品名)," "Quantity (数量)," "Unit Price (単価)," "Line Amount (金額)," "Subtotal (小計)," "Consumption Tax (消費税)," "Total Amount (合計)," "Bank Name (振込先銀行名)," "Branch Name (支店名)," "Account Number (口座番号)," "Account Holder (口座名義)," "Invoice Registration Number (インボイス登録番号)," and "Shimebi (締め日)" — and the AI reads every seikyusho and populates each column by understanding what the labels mean, regardless of each supplier's layout or whether the invoice arrived as a PDF, a scan, or a photograph. Because it is batch-first, 300 seikyusho become one structured table in minutes rather than the thirty-five hours of manual entry that Layer One measures. The step-by-step version of that workflow for a single seikyusho is covered in the guide to extracting Japanese invoice data to Excel with payment fields and consumption tax. The guide to batch-processing Japanese supplier invoices into an AP spreadsheet with furikomi bank details covers how the same approach scales to a full monthly settlement cycle.

Two capabilities directly attack Layers Two and Three. An inferred column can flag whether a supplier is a withholding-eligible entity: add a column named Withholding Required (options: Yes/No/Check) and the tool makes an initial categorisation you verify rather than build from zero — collapsing the decision that currently rests on an AP clerk's memory of which suppliers are gensen-eligible. A computed column can verify the arithmetic: define a column that sums all line item amounts plus consumption tax and compares the result to the stated total (合計), and every seikyusho with a calculation error is flagged before it enters the payment queue. What was a manual cross-check becomes an automated validation.

The cost comparison for the example SME:

At roughly ten seconds of AI processing per seikyusho versus five to ten minutes of manual reading, typing, categorising, and verifying by hand, extraction does not reduce Layer One — it eliminates the reading-and-typing step and replaces it with a verification glance. The furikomi error risk on Layer Two collapses because the bank details are extracted from the document text, not typed from memory. The gensen choshu exposure on Layer Three shrinks because the withholding decision is flagged consistently rather than made under time pressure. The compliance cost on Layer Four disappears because the invoice registration number is captured alongside every other field and can be batch-verified. The arithmetic is not gradual improvement. It is a structural change in the cost base.

FAQ — Manual Seikyusho Processing Costs for Japanese SMEs