Methodology

Methodology

How this site is built, how sources are handled, and how data quality is maintained. Every claim links back to an official document.

How the site works

No fake crosswalks, no hidden source layer.

Data Pipeline

The site follows a strict data pipeline to ensure accuracy and traceability:

  1. Source acquisition — Official PDF documents and structured data files are downloaded from Soumu and e-Stat. No third-party data sources are used.
  2. Extraction — The build scripts parse official PDFs and JSON files to extract the complete JSIC hierarchy: 20 sections, 98 major groups, 519 groups, and 1411 industry codes.
  3. Validation — Each extracted code is validated against the official structure. Parent-child relationships are verified. Missing or orphaned codes are flagged.
  4. Enrichment — Official annotations (overview, inclusions, exclusions, notes) are linked to each code. Crosswalks to ISIC, NACE, and NAICS are applied at the section level.
  5. Generation — The build script generates static HTML pages for every code, section, and classification level. Each page includes source references and hierarchy navigation.

Source Verification

Every piece of data on this site is traceable to an official source:

  • Primary sources — Ministry of Internal Affairs and Communications (Soumu) publishes the official JSIC Rev. 14 documents. These are the authoritative reference.
  • Structured data — e-Stat provides machine-readable versions of the classification hierarchy used for programmatic access.
  • Cross-references — International classifications (ISIC Rev.4, NACE Rev.2, NAICS 2022) are referenced from their respective official bodies (UN, Eurostat, US OMB).
  • No fabrication — If a crosswalk or mapping does not exist in the official data, it is not created here. The site does not invent relationships between classifications.

See the Sources page for direct links to all official documents.

Translation Approach

English translations on this site follow these principles:

  • Official translations first — Where the Japanese government publishes official English equivalents, those are used.
  • Direct translation second — For codes without official English names, translations are made directly from the Japanese text, preserving the technical meaning.
  • Consistency — Related codes use consistent terminology (e.g., all "大分類" → "Section", all "中分類" → "Major group").
  • No interpretation — Translations reflect the literal Japanese meaning. Editorial interpretation is avoided.

Risk Score Methodology

Each industry code receives a compliance risk score from 0 to 10 based on seven deterministic factors. The score reflects data coverage, not business risk.

See the dedicated Risk Score Methodology page for the full formula, factor descriptions, and implementation details.

Quality Assurance

The following checks are run at build time:

  • Hierarchy integrity — Every code must have a valid parent chain (section → major group → group → industry). Orphaned codes are rejected.
  • Duplicate detection — Identical codes at the same hierarchy level are flagged.
  • Crosswalk validation — Mapped codes must exist in both source and target classifications.
  • Link verification — Internal links are checked for correct routes. External links to official sources are verified.
  • Page generation — The build process generates 1411+ static pages. Any template errors or missing data cause the build to fail.

Build Technology

  • FrameworkAstro static site generation.
  • Data format — JSON files for hierarchy, crosswalks, and metadata.
  • Build scripts — Node.js scripts for data extraction, validation, and page generation.
  • Hosting — Static files served via CDN. No server-side processing.
  • Analytics — Google Analytics (GA4) for usage statistics only. No personal data collection.

Limitations

  • This site covers JSIC Rev. 14 only. Earlier revisions are referenced but not fully browseable.
  • Crosswalks are section-level only. Detailed 4-digit to 4-digit mappings between JSIC and other classifications are not available in the official data.
  • English translations are best-effort. For regulatory purposes, refer to the official Japanese documents.
  • Risk scores reflect data coverage, not compliance advice. They should not be used as the sole basis for business decisions.

Contact

For questions about methodology, data sources, or technical implementation, visit the Contact page.