What is the Commodity Flow Survey?

The Commodity Flow Survey (CFS) is a sample survey of US establishments that asks businesses what they shipped, where it went, how it travelled and what it weighed and was worth. The U.S. Census Bureau collects the responses; the Bureau of Transportation Statistics (BTS) co-sponsors the program and shapes the freight content. It runs every five years — the 2017 and 2012 cycles are still published alongside 2022 — and its output feeds the Freight Analysis Framework, the federal government's origin-destination freight database.

The survey's frame is US establishments, not carriers or border crossings. Businesses receive the questionnaire, and their answers describe shipments originating inside the United States — which is why imported goods that move internationally before arriving never appear, while an import that travels from a port warehouse to a distribution center does. Each record also carries SECTOR, the industry sector of the shipping establishment (code 31-33 covers manufacturing), so analysts can separate factory gate flows from wholesaler flows.

That establishment-based design explains both the survey's authority and its blind spots. It captures the tonnage, value and distance of the domestic economy at shipment level — no other dataset in Datadory's Cargo Ground Transportation slice exposes individual shipment value, weight and distance together — but it does not count vehicle movements, highway speeds or spot prices. Those jobs belong to FHWA's probe program and USDA's rate tables, which sit further down this page.

The PUMS data dictionary documents 18 variables per row, keyed by SHIPMT_ID running from 00000001 to 37576546. Geography arrives as three nested origin-destination pairs: ORIG_STATE and DEST_STATE hold FIPS state codes, ORIG_MA and DEST_MA identify metro areas, and ORIG_CFS_AREA and DEST_CFS_AREA carry CFS areas, the finest published geography. Commodity is SCTG, the two-digit Standard Classification of Transported Goods code, and MODE separates for-hire truck from private truck — the split that matters most in ground cargo work.

The measures are concrete: SHIPMT_VALUE in dollars (documentation example 125,000), SHIPMT_WGHT in pounds (example 42,000) and SHIPMT_DIST_GC in great-circle miles (example 1,745). Four flags complete the record — TEMP_CNTRL_YN for temperature-controlled freight, EXPORT_YN and EXPORT_CNTRY for exports, and HAZMAT — plus WGT_FACTOR, the tabulation weight that inflates the surveyed sample into population estimates. Apply WGT_FACTOR before totalling: raw row counts describe the sample, not the economy.

Those variables compose into the analyses ground cargo teams actually run. Grouping SHIPMT_WGHT-weighted rows by SCTG and MODE yields commodity-by-mode tonnage comparable to the published tables; binning SHIPMT_DIST_GC produces distance-band profiles that show how far freight travels before it changes mode; filtering TEMP_CNTRL_YN isolates the reefer share of domestic shipments, the same population USDA's weekly rate table prices lane by lane.

Choose by question shape. If your question matches a published tabulation — tonnage by mode, value by commodity group, flows between states — take the summary table, because roughly 30 of them arrive pre-cut with title, table ID, notes, sources and release date attached. If you need a cut nobody publishes, such as temperature-controlled shipments over 500 miles between specific metro areas, the PUMS is the only route, since its 37,576,546 records support any filter the 18 variables allow.

Context around the shipments comes cheaply too. FHWA's Freight Performance Measurement Program refreshes NPMRDS-based travel-time reliability and bottleneck measures monthly at TMC-segment level across the National Highway System, and the Data.gov Freight Catalog searches 552,271 federal records through a JSON API, surfacing roughly 130 titles matching cargo, freight, truck and motor-carrier terms. Internationally, the CFS has no direct equivalent in the slice: Eurostat's road_go_* family reports 1,886 billion tonne-kilometres moved in the EU in 2025 under commercial delivery terms 4.0, while the UK Department for Transport publishes about 40 RFS OpenDocument tables quarterly under OGL v3.0 — comparable national aggregates rather than shipment microdata.

Because trucks dominate the flows it measures, and the survey is the only public source measuring them at shipment level. ATA's economics hub reports trucks moved roughly 72.7% of US freight by weight in 2024 — about 11.27 billion tons — so the MODE column separating for-hire from private truck slices the largest share of American freight activity into analyzable rows. A modeler weighting SHIPMT_WGHT by WGT_FACTOR and grouping by SCTG and distance band reproduces, cell by cell, the same shipment population federal freight planners start from.

Its influence compounds through reuse. FAF5 integrates the 2017 CFS with Census trade data and sector sources, meaning downstream users inherit CFS methodology even when they only touch FAF5 tables. For teams building demand models, siting distribution capacity or sizing corridor investment cases, the practical starting point is the dataset record itself, then the workflows collected for data scientists working with cargo ground transportation data. Quality-wise, Datadory scores the source 10 out of 10 — the highest tier in the slice, shared only with FAF5.