Methodology

1. 1 · Raw sourceDone

   Extraction of the 1-AJs_portal_24-08-2026.xlsx workbook (snapshot 24/08/2026) into CSV, preserving the raw name, aggregate counts, source note and the original Excel row. Nothing is rewritten at this stage.

2. 2 · NormalisationPartial

   Typing and deterministic validation only (integer counts, ISO dates, triage enums). Accents, casing and company names stay as in the source; no name rewrite is persisted.

3. 3 · Entity resolutionNot started

   Deterministic similarity produces alias candidates for human review. No canonical record exists. Without an official identifier there is no possible canonicalisation.

4. 4 · Case-level ingestionBlocked

   Depends on an export with case_source_id/case number, project_id, aj_source_id, lineage and extracted_at. Until it exists, Atlas shows no cases from this seed.

5. 5 · ReleaseRestricted

   Internal use only for this seed. No publication, no external exposure and no integration with the institutional site or the fictional demo.

Name comparison produces review candidates only. No grouping changes data, no name is elected canonical and no count is recalculated from groupings. Every transformation records its origin and rule, and is reversible.

• The sum of case–administrator links must reconcile to 2,779.
• The sum of administrator–project links must reconcile to 3,007.
• The distinct project count must reconcile to 2,949.
• The 423 raw rows must remain traceable to the original Excel row.
• Every entity merge requires recorded human approval and is reversible.
• Placeholders and court experts must not be counted as insolvency administrators.
• Any published metric must cite snapshot, source and coverage limitation.

• דEvery Brazilian judicial reorganisation case is in Atlas.”
• דComplete case base.” — this seed ingested no proceedings.
• דOfficial ranking of insolvency administrators.”
• דX confirmed insolvency administrators.” — these are raw names pending review.
• ×Any mention of clients, traction, production use, revenue or testimonials.
• ×Attributing a specific case to an administrator based on this aggregate.

• It does not prove that each raw name matches a real, unique insolvency administrator.
• It does not prove how many distinct cases exist — only link counts declared by the source.
• It does not prove geographic, court, division or period coverage.
• It does not prove market share, real concentration or the performance of any administrator.
• It does not prove the current status of any proceeding: the data is a single snapshot of 24/08/2026.
• It does not prove that placeholders and experts were removed — they remain in the seed, flagged.