How to Extract Data from Japanese Payslips
Into Excel

What Makes a Japanese Payslip Structurally Different for Data Extraction

A Japanese payslip is not a single list of deductions. It splits employer-mandated subtractions across four legal categories — each with its own calculation base, its own rate structure, and its own government agency destination.

Under Article 24 of the Labor Standards Act (労働基準法第24条), wages must be paid at least once a month on a fixed date. Under Article 231 of the Income Tax Act, a written payslip must be delivered with each payment. Neither law prescribes a specific payslip layout — which means the same data can appear in entirely different positions depending on the payroll software that produced it.

The four deduction categories printed on every Japanese payslip are:

1. Social insurance (社会保険, shakai hoken). Health insurance (健康保険, kenko hoken) and welfare pension (厚生年金, kosei nenkin). Both are calculated on the employee's standard monthly remuneration (標準報酬月額, hyojun hoshu getsugaku) — a graded amount set by the Japan Pension Service based on the employee's average salary, not the actual monthly pay. The standard remuneration system has 39 grades (as of 2026), each with a fixed contribution amount. This means the social insurance deduction on two different payslips for the same employee in the same year is identical month to month — a useful extraction sanity check. For employees aged 40 to 64, long-term care insurance (介護保険, kaigo hoken) is added on top.
2. Labor insurance (労働保険, rodo hoken). Employment insurance (雇用保険, koyou hoken) — paid by both employee and employer — and workers' accident compensation insurance (労災保険, rosai hoken) — paid by the employer only. Unlike social insurance, employment insurance is calculated on actual gross pay, not standard remuneration. The rate is 0.55% for the employee and 0.90% for the employer as of April 2026 for general businesses, and higher for construction and agriculture sectors.
3. Income tax (所得税, shotoku zei). Withheld at source by the employer under Japan's progressive tax system, with rates ranging from 5% to 45% across seven brackets. An additional 2.1% reconstruction surtax (復興特別所得税, fukko tokubetsu shotoku zei) applies on the income tax amount. The withholding rate depends on the employee's dependents (扶養家族, fuyo kazoku), declared on the dependents declaration form submitted to the employer.
4. Resident tax (住民税, jumin zei). A flat 10% levy (4% prefectural + 6% municipal) — and this is where Japan diverges from every other country's payslip. Resident tax is calculated on the previous year's income, not the current year's. An employee who started working in Japan in January 2026 will see zero resident tax deductions between January and May, followed by a sudden monthly deduction starting in June. An extraction tool that does not anticipate absent values will flag these rows as incomplete when they are, in fact, correct.

Resident tax sits in a different tax year than income tax. A generic extraction that pools both into one "Tax" column creates a number that corresponds to no actual tax obligation — it is the sum of a current-year income tax liability and a prior-year resident tax liability, two numbers that should never be added together.

If you have worked with payslip extraction in other markets, the contrast is clear. An Spanish nomina separates Seguridad Social into five sub-lines on one payslip block. A Korean payslip (급여명세서) itemizes four national insurances (national pension, health, employment, industrial accident) plus income tax and local income tax — structurally the closest analogue to Japan's system. A German Lohnabrechnung separates Steuer from Sozialversicherung with a Steuer-ID. Japan is the only one of these where two tax types on the same payslip — income tax and resident tax — operate on income from different tax years.

Social Insurance Rates — The Numbers That Determine Every Employee Row

Japanese social insurance is not one rate for the whole country. Health insurance varies by prefecture — and the difference between the lowest and highest prefecture is 1.47 percentage points, meaning two employees earning the same salary in Saga and Niigata have materially different take-home pay.

As of FY2025, the combined (employee + employer) health insurance rate under the Japan Health Insurance Association (協会けんぽ, Kyokai Kenpo) — which covers most SME employees — ranges from 9.31% in Niigata to 10.78% in Saga, split 50/50. Tokyo is 9.91%. These rates are published annually by prefecture. Welfare pension is locked at 18.30% combined (9.15% each) under the 2017 reform, and it stays there until at least the next statutory review. Long-term care insurance, applicable from age 40 to 64, adds approximately 1.59% combined (split 50/50).

The standard monthly remuneration (標準報酬月額) system matters for extraction because it means social insurance deductions are not proportional to the exact salary on a given payslip. A ¥310,000 monthly salary falls into grade 21 (¥300,000–¥330,000 range), and social insurance is calculated on the midpoint — ¥320,000. If the employee's actual gross pay fluctuates month to month — overtime in June, fewer hours in August — the social insurance deduction stays flat. An extraction that applies a fixed rate to the actual gross pay will produce numbers that do not match the payslip, and the error compounds across every social insurance line. This is why extracting the printed deduction, not calculating it, is the correct approach.

Mandatory Payslip Fields — Mapping Each One to Your Spreadsheet Columns

Every Japanese payslip contains the fields below, mandated by tax law and payroll practice. A missing field is not a formatting oversight — for social insurance and tax items, it is a statutory obligation. Here is the extraction column map for a standard Japanese payslip:

Nineteen columns is not excessive — it is what the payslip's statutory structure demands. A generic extraction that outputs Employee Name, Gross Pay, Net Pay, and one "Tax" column will fail to capture the employer's side of the journal entry (the employer health insurance and pension contributions, which are not printed on the employee's payslip but are calculable from the standard remuneration and rates) and will mix income tax with resident tax in a way that makes year-over-year compensation analysis unreliable.

Resident Tax vs. Income Tax — The Two-Year Extraction Trap

The most common error in Japanese payslip extraction is treating resident tax (住民税) and income tax (所得税) as one number. They belong to different tax years, different agencies, and different accounting entries.

Income tax is withheld on the current month's income, based on the employee's dependents and the National Tax Agency's monthly withholding table. The employer remits it to the tax office by the 10th of the following month. In December, the employer performs the year-end adjustment (年末調整, nenmatsu chosei) — recalculating the total annual tax liability against actual income and deductions, and refunding or collecting the difference in the December payslip. This means the December income tax line can be negative (a refund), which looks like an extraction error but is correct.

Resident tax follows a completely different timeline. It is based on the employee's total income from the previous calendar year, reported to the municipal office by the employer in January via the payment report (給与支払報告書, kyuyo shiharai hokokusho). The municipality calculates the annual resident tax amount and sends a tax levy notice to the employer in May. The employer then withholds the total in 12 equal installments from June of the current year through May of the following year.

What this means for extraction:

• A new employee hired in Japan in February 2026 has zero resident tax on every payslip from February 2026 through May 2027. From June 2027, a resident tax deduction appears — based on the very little income earned in February–December 2026. An extraction tool that expects a non-zero value in the Resident Tax column for every row will flag 16 consecutive months as missing data when none is missing.
• An employee leaving Japan in August 2026 will receive a final payslip with a lump-sum resident tax deduction. The employer collects the remaining installments (September through May) in one deduction from the final payslip. A computed column that checks "Resident Tax should equal 1/12 of the annual amount" will flag this row as an outlier when it is, in fact, the correct statutory treatment.
• The December year-end adjustment can make income tax negative for that month. If you are building a month-over-month tax trend analysis, a negative income tax cell in December followed by a positive cell in January is the expected pattern — not an extraction anomaly.

The extraction rule is straightforward: always extract the printed number on the payslip, not a calculated number. For verification, use Computed Columns to cross-check the printed numbers against expected ranges — but use them to flag rows for review, not to auto-correct values you think are wrong.

How to Extract Japanese Payslip Data to Excel — Step-by-Step Workflow

The workflow is the same whether you are processing one payslip for a visa renewal document check or 80 for month-end payroll reconciliation at a multinational HR shared service center. The difference is in column design and verification steps.

Verification Checks — Gross-to-Net Reconciliation Using Computed Columns

A Japanese payslip carries a built-in audit trail: Total Earnings minus Total Deductions must equal Net Pay. If these three columns are extracted, one formula in a fourth column flags every discrepancy before the data enters the payroll journal.

This is not a post-extraction step in Excel. It is built into the extraction itself using Computed Columns — columns whose names describe calculations the AI performs during extraction. Define a column named "Reconciliation Check (Gross Pay minus Total Deductions minus Net Pay)" and the AI computes the difference for every row while extracting. Any non-zero result flags a row that needs manual review — before the data reaches your general ledger.

Three additional verification checks worth embedding in your extraction columns:

For a detailed walkthrough of using computed columns in payslip extraction, see the guide on AI-computed net pay verification across gross-to-net calculations. The verification concepts apply across all payslip formats — US, EU, and Asian — but Japan's dual-tax-year structure makes them particularly necessary.

Handling Payslips from Different Payroll Software in One Batch

Japanese payroll software produces payslips that look nothing alike, even though they carry identical mandatory data. Knowing what to expect from each platform prevents the most common extraction failure: building your column list around one software's layout, then processing a batch that includes payslips from a different platform and wondering why half the rows have missing data.

Yayoi Kyuyo (弥生給与) — approximately 3 million users, the most established SME payroll platform in Japan. Yayoi payslips typically use a two-column layout: earnings (支給) on the left, deductions (控除) on the right, with the deduction section itemizing each social insurance line separately. The bottom of the payslip shows the standard monthly remuneration and dependents count as reference data. Yayoi is desktop-first; many companies generate payslips as printed PDFs, which means the text is digital but the layout is not tagged — requiring the AI to read the document visually rather than from structured text fields.

freee HR (freee人事労務) — cloud-based, all-in-one HR and payroll platform with over 399,000 companies. freee payslips use a cleaner, more modern layout with earnings and deductions arranged in a single vertical flow, often with clear section headers in Japanese. The year-end adjustment (年末調整) result appears as a separate line item in the December payslip, marked clearly. freee payslips are typically exported as digital PDFs with selectable text, making them easier for text-based extraction but equally readable by visual AI.

Money Forward Cloud Payroll (マネーフォワード クラウド給与) — 400,000+ users, known for accounting software integration. Money Forward payslips follow a card-style layout where each earnings and deduction item appears in its own block. The health insurance rate is sometimes printed alongside the deduction amount, providing a secondary verification source. The excel output from Money Forward's own payroll module is already structured — but if you are receiving PDF payslips forwarded from clients or subsidiaries that use Money Forward, the extraction workflow handles them the same way as any other PDF.

SmartHR and jinjer — SmartHR focuses on labor management and does not have a built-in payroll engine; companies using SmartHR for HR management typically pair it with an external payroll system (often Money Forward Payroll or Yayoi Kyuyo), and the payslip format follows the payroll system used. jinjer produces its own payslips with a clean, single-page layout.

The key to batch-extracting payslips from multiple payroll platforms is not knowing every software's layout. It is using extraction that reads fields by meaning — the AI recognizes "健康保険" or "¥25,300" in a deduction context regardless of whether that number sits in a right-hand column, a vertical list, or a card-style block.

For companies that need to collect payslips from employees across multiple subsidiaries — each potentially using different payroll software — the batch workflow is identical: upload all files, define the column names once, and process. The same concept extends to the year-end reconciliation process. When the withholding tax certificate (源泉徴収票, gensen choshu hyo) arrives in January, the annual totals it reports should tie back to the twelve monthly payslip extractions. See the guide on preparing Japanese year-end payroll withholding data for the reconciliation workflow between monthly payslip data and annual tax certificate totals.

FAQ

Does an AI extraction tool handle Japanese-language payslips correctly, including kanji field names?

Yes. The AI reads Japanese text — including kanji, hiragana, katakana, and alphanumeric characters — as it appears on the document. Column names you define can include Japanese terms as context for the AI (e.g., "Health Insurance (健康保険)"), which helps it locate the correct fields regardless of whether the payslip uses Japanese labels, English labels, or both. The extracted values preserve the original text from the payslip — amounts as numbers, names as Japanese characters.

What happens if the employee's health insurance rate changes mid-year?

Health insurance rates under Kyokai Kenpo (協会けんぽ) are set per prefecture per fiscal year and updated annually in March. A rate change between February and March payslips is normal and expected. The extraction tool reads the printed deduction amount — it does not calculate it — so a rate change manifests as a different deduction value on the March payslip compared to February. Your reconciliation process should verify that the new deduction amount aligns with the employee's standard remuneration grade at the new rate, but the extraction itself is unaffected.

Can the extraction handle the December year-end adjustment where income tax becomes a refund?

Yes — with a verification step. The AI extracts the printed income tax value, which may be negative in December after the year-end adjustment (年末調整). If you use a Computed Column like "Dec Tax Sign (if Pay Period is December and Income Tax is negative, output 'NTA Refund — verify'; otherwise output 'OK')", the output flags negative December income tax for a reviewer to confirm. The tool does not alter the extracted value — it extracts what is on the payslip and lets you decide how to handle the result.

Does the tool handle payslips from all Japanese payroll software?

Because the AI uses semantic understanding — locating values by what they mean rather than by pixel coordinates — the same column definition works across Yayoi Kyuyo (弥生), freee HR, Money Forward Cloud Payroll (マネーフォワード), SmartHR outputs, and jinjer payslips. A layout difference (vertical list vs. horizontal table vs. card layout) does not affect extraction accuracy as long as the field labels and values are legible. For a batch containing payslips from multiple payroll platforms, use the same column list for all files.

What about employees with zero resident tax for the first year of employment?

Employees in their first year of employment in Japan (or those who had no income in Japan in the prior year) will have a resident tax (住民税) line showing ¥0 on every payslip from June of the current year through May of the following year. The extraction tool reads and outputs ¥0 — it does not flag a zero as missing data. If you are doing year-over-year compensation analysis, the distinction between an employee with ¥0 resident tax (first year) and one with a standard ¥20,000–¥40,000/month deduction (subsequent years) is visible in the data — no manual annotation needed.

Can I extract only the net pay and skip the individual deduction breakdown?

Yes — define only the columns you need. If your use case is a simple income register for personal records, extracting Employee Name, Pay Period, Gross Pay, Net Pay, and a single Total Deductions column is sufficient. But if you are using the data for payroll journal entries, social insurance reconciliation, or year-over-year tax analysis, extract each deduction into its own column. The five-minute investment in defining separate columns saves the hour it would take to manually split one "Deductions" total back into its component parts when the quarterly social insurance payment notice (納付書) does not tie to your extracted data.