Data lanes: public USAspending vs synthetic

Public lane — USAspending reference sample

*Curated, non-probability public reference sample of 250 USAspending awards. Not total CBP spend and not a confirmed active-contract inventory.* Activity window 1 Oct 2023 – 8 Oct 2026.

• 250 awards and orders: 166 delivery orders, 36 BPA calls, 36 definitive contracts, 12 purchase orders.
• 131 direct parent vehicles — context records only, excluded from totals.
• 143 vendors, 19 funding offices, 9 awarding offices, 32 canonical organizations.
• 2,997 award actions (full available history); all 250 ledgers reconcile to source-reported cumulative obligations.
• 147 awards whose published current POP overlaps the snapshot date and 103 outside it — overlap is not confirmed active.

Synthetic lane — fictional workflow cases

*SYNTHETIC WORKFLOW — fictional people, amounts, approvals and documents. No legal effect. Never aggregated with public records.* Each of the 250 scenarios (SYN-CBP-001…) links to a public award by public_reference_award_key for context only, and carries its own separate fictional award (SYN-CBP-001-AWD).

• 12 full-lifecycle (deep) cases with ~20 evidence documents each — readiness is scored.
• 238 baseline cases with 3 documents each — intentionally incomplete training fixtures, never scored as deficient.
• 954 synthetic PDFs (257 editable official-form examples: 122 SF1449, 103 OF347, 8 SF1442, 24 SF30); every record has a hash, version, state, extraction and page/field anchors.

Baseline lane — DHS Ghost fixture

The Baseline pages use scenario GHO-PDCC-DHS-CSO-001 (classification SYNTHETIC-TEST, seed PEEK-PDCC-2026-08-18), locked to the clock 18 Aug 2026 16:00 ET. Its golden contract is CTR-0042 (PIID 70DHS-SYN-26-D-0042, task order TO-1054, vendor *Delta Solutions Inc. (SYNTHETIC)*).