A sample package

manifest.json

{
  "business_id": "DEMO-0001",
  "synthetic": true,
  "industry": "residential construction",
  "standard_version": "redaction-1.0",
  "years": [2018, 2025],
  "participants": ["PARTICIPANT-01", "PARTICIPANT-02"],
  "records": {
    "annual_layer": 8, "payroll_year": 112, "payroll_period": 1867,
    "revenue_group_year": 40, "entities": 21,
    "decision_narratives": 63, "findings": 41
  },
  "coverage": { "declared_documents": 14, "read": 14, "exported": 14 }
}

decision_narratives.jsonl — one record

{
  "id": "dn_0142",
  "participant": "PARTICIPANT-01",
  "conversation": "people",
  "question_shape": "why", "shape": "why", "retrospective": true,
  "question": "In 2023 you brought on a salaried heating-and-cooling tech, then two helpers the next year. How did that come about?",
  "answer": "I was writing checks to [ORG-003] for close to three hundred grand a year and waiting on his schedule. His lead tech was the one I actually trusted. When he said he was thinking about leaving, I ran what a licensed tech, a van, and a supply account would cost against what I was paying out. It penciled if we kept him busy on our own jobs.",
  "evidence_refs": ["finding:payroll:2023:new_category", "payroll_year:2023:ROLE-006", "doc:payroll_register_2023"],
  "source_doc": "payroll_register_2023", "consent_id": "consent_v1_0032",
  "collected_at": "2026-10-03", "redaction_version": "1.0"
}

Follow evidence_refs from the sentence to the finding, the payroll rows, and the document. The tag ORG-003 resolves in entities.jsonl.

entities.jsonl — one record

{ "tag": "ORG-003", "type": "business", "descriptor": "the heating-and-cooling subcontractor (through 2022)", "years": [2015, 2022], "confirmed_by": "PARTICIPANT-01" }