How do paid supply-chain data vendors obtain their data — split by filing data, customs data, survey data, direct company outreach, and estimation — and for each method, which primary sources are free and public and which are not?
The evidence is 50 selected rows, 10 per method family, drawn from 17 source hosts (no single URL backs more than 6% of rows): patent disclosures, government methodology pages, SEC guidance and academic studies describing vendor practice and original-source access conditions. “Free and public” below always describes the original source, never the vendor's normalized database.
Each cell is one evidence row; hover or tap a block for the organization, source and link. Colour means access to the original primary source, not to the vendor product. Several customs and estimation rows are Panjiva patent family mirrors on different hosts — technical illustrations, not independent confirmation of current commercial coverage.
FactSet/Revere, EDGAR disseminators, registries
Vendors ingest corporate regulatory filings, annual reports, investor presentations, press releases and websites, then extract named relationships, cross-validate and build reverse links. SEC EDGAR and many issuer sites are free and public; live EDGAR feeds to disseminators are fee-based, some registries and bill-of-lading microdata are restricted, and the consolidated relationship database is proprietary. Disclosure thresholds mean filings capture major or voluntarily disclosed relationships only.
Panjiva, Dun & Bradstreet, U.S. Census Bureau
Shipment-level customs records, bills of lading and official trade statistics are resolved to entities, normalized to HS codes, deduplicated and error-corrected. Patents characterize U.S. customs records as public but fragmented; yet entity names and origins can be redacted, sensitive records missing, weights erroneous, and importer/exporter microdata confidential. Aggregate public statistics are not named shipment records.
EcoVadis, BEA, World Bank LPI, Commerce
Supplier questionnaires, buyer feedback and uploaded evidence are scored and validated by analysts or rules. Methodology documents and benchmark indices can be free and public, but respondent-level answers, uploaded documents, buyer feedback and the resulting scorecards are private, permissioned or subscription-only.
Resilinc, FactSet reverse links, supplier directories, Panjiva enrollment
Vendors invite suppliers to identify sites, parts, sub-tier suppliers and continuity plans, and verify or correct model-generated links. Open directories and self-disclosures can be free, but responses to vendor invitations, customer master data, bills of materials, site confirmations and continuity plans are confidential or customer-controlled.
Panjiva nowcasting, EcoVadis country indices
Vendors infer missing shipment volume or value, entity identity, air freight, risk and macro trade measures from observed records plus auxiliary data, using rules, statistics or machine learning. Inputs may be free (public filings, customs fragments, commodity prices, public country indices) or licensed and confidential — but the estimated field, confidence score and trained model are vendor-created outputs either way.
Collection → entity resolution → normalization → cross-validation → inference
No method stands alone: patents describe triangulating public customs records with private third-party records, directories, filings and surveys. The critical distinction is between original input and compiled product — a public manifest in, a proprietary relationship graph, historical linkage, alerts and confidence scores out.
Anyone can read the original record at no charge — EDGAR filings, issuer websites, published methodology pages, some customs record fragments.
The record is public in principle but costs money or effort to obtain in usable form — live EDGAR dissemination feeds, paid registry documents, bulk extracts.
Access is limited by law or contract — importer/exporter microdata, respondent-level survey answers, buyer feedback, unredacted shipment identities.
Created by the vendor — the resolved entity graph, normalized codes, historical linkage, alerts, confidence scores and estimates. Public inputs do not make these public.
| Method | Organization | Access of original source | Primary source | Vendor process / restriction | Source |
|---|
Method: 50 selected evidence rows (10 per acquisition-method family) across 17 source hosts, compiled from patents, government methodology pages, SEC guidance and academic papers; no URL backs more than 6% of rows. Each row records one described vendor practice or one original-source access condition; access class refers to the original primary source, not the vendor product. Patent rows (notably mirrored Panjiva families) are technical illustrations, not confirmation of current commercial coverage. Long text fields truncated for space.