Invoice & procurement document set

The full procure-to-pay paper trail of a wound-down manufacturer: purchase orders, electronic invoices (NF-e), delivery receipts, and supplier terms, linked across the cycle. Documents with real layout variety and matching structured records, ideal for document AI that must reconcile paper against ledgers.

Specs

Country

Brazil

Industry

Manufacturing

Span

2018-2024 (7 yrs)

Volume

~62k documents

Formats

PDF, XML, CSV

Language

Portuguese (BR)

Origin

Wound-down manufacturer, 2018-2024

Composition

POs, NF-e invoices, receipts, terms

Refinement

Suppliers masked, totals preserved

Clearance

• Seller verified (KYB) before intake.
• Anonymization report available under NDA.
• Source records destroyed within 72 hours after processing.

Use cases

• Train invoice extraction on real NF-e layouts with linked structured truth.
• Model three-way matching: PO to receipt to invoice, with genuine mismatches.
• Stress-test document pipelines against layout drift across seven years.