Freehand

Freehand — Shipment-level trade-compliance decision corpus: HS classification assignments for 100% of audited cross-border shipments with the product-spec evidence behind each code, per-shipment duty calculations by HS code, origin and trade agreement, FTA qualification verdicts with country-of-origin documentation, drawback claims linking duty payments to qualifying exports, broker invoice audit findings, and the flagged-error records (misclassifications, duty overpayments, missed FTA savings, unclaimed drawback) caught before payment cleared — unified in a Context Graph of product specifica…

Freehand offers (Alternative, Supply Chain, Reference) — Volume is driven by the 100% shipment classification coverage mechanic — every audited shipment line is classified, priced and origin-checked — so for a large shipper portfolio this is order of tens of millions of classified shipment lines per year across the customer base. The page cites $4.1M average annual duty recovery per customer, implying a duty base on the order of tens to hundreds of millions of dollars per account and therefore a very large shipment-line count. No shipment or record count is published; request further research for classified lines per year per customer..

Training and evaluating HS-classification models from free-text product descriptions (the hardest problem in customs automation, and this corpus is expert-verified rather than broker-default); tariff-shock exposure models mapping a named tariff action onto an audited product-level shipment base to size cost impact; FTA utilisation and sourcing-shift models for which lanes and origins actually qualify; landed-cost and margin-at-risk models for retailers and manufacturers; duty-overpayment and invoice-fraud detection; commodity-by-lane trade-flow inference from classified shipment volumes.

The data is with 3 years of history.

Coverage spans US, Europe, UK; Transportation Services, Application Software; alternative, supply_chain, reference; equities, commodities.