Data.gov - Food-Tagged structured datasets Collection

An index card for every dataset the American government publishes on food. Data.gov - Food-Tagged structured datasets Collection is the food-keyword slice of catalog.data.gov, the General Services Administration's federated harvest catalog that advertised 552,271 datasets at the August 2026 research pass. Cursor-paging the catalog's search to exhaustion confirmed 208 records carrying the exact keyword 'food', with 30 more matching 'meat' - a deliberately small shelf cut from an enormous warehouse.

The shelf holds named releases most food operators know by reputation: FSIS FoodKeeper Data in XLS and JSON across English, Spanish and Portuguese; the Meat, Poultry and Egg Product Inspection Directory by Establishment Number and by Establishment Name on a monthly declared cycle; FSIS Establishment Demographic Data and Quarterly Enforcement Reports; the USDA National Nutrient Database for Standard Reference Legacy release and the Nutrient Data Set for Retail Meat Cuts; ERS U.S. Food Imports and Food Availability products; EPA's U.S. Food Waste Flows Between Sectors 2018 v1.3.2; and municipal files such as Chicago Food Inspections, NYC restaurant inspection results and grocery store location lists.

What turns a search page into a dataset is uniformity: every hit arrives as the same DCAT-US metadata record with twelve verified fields, whichever of the federal, state, county, city, university, tribal or non-profit publishers produced it. Within Datadory's other-specialty-retail shelf - ten datasets averaging a 7.2 quality score - this record scores 7. Get a sample of this dataset scoped to the slice you follow.

One row per catalog record - who published what, when they last touched it, how often they say they refresh it, under what reuse terms. Five verified rows from the August 2026 pass, exactly as they land:

title                : FSIS - FoodKeeper Data
publisher            : Food Safety and Inspection Service
identifier           : usda-fsis-006
modified             : 2025-01-22          accrualPeriodicity : irregular
formats              : [xls, json]

title                : FSIS MPI - Meat, Poultry, and Egg Inspection Directory by Establishment Number
publisher            : Food Safety and Inspection Service
identifier           : USDA-FSIS-02246
modified             : 2025-01-22          accrualPeriodicity : R/P1M

title                : USDA National Nutrient Database for Standard Reference, Legacy Release
publisher            : Agricultural Research Service
identifier           : 10.15482/USDA.ADC/1529216
modified             : 2025-11-22          formats : [zip]

title                : U.S. Food Imports
publisher            : Economic Research Service, Department of Agriculture
identifier           : USDA-ERS-00093
modified             : 2020-02-28          accrualPeriodicity : R/P1Y

title                : U.S. Food Waste Flows Between Sectors, 2018 v1.3.2
publisher            : U.S. EPA Office of Research and Development (ORD)

Read them as the collection's range in miniature: a consumer-facing storage guide refreshed on its own schedule, an inspection directory enumerating federally inspected establishments monthly, a legacy nutrition archive distributed as one zip, an annual import table untouched since February 2020, and a waste-flow model from EPA's research arm. Five publishers, five rhythms, one schema.

Twelve documented fields define every record, each definition verified against the source during the August 2026 research pass rather than inferred from column names. They split into four jobs. Identity: title, description, publisher, identifier - the identifier doubling as DOI on records such as the SR Legacy nutrient release. Dating: issued (present on roughly half of records) and modified, which is the column freshness filters actually run on. Classification: the keyword array the food filter resolves against, plus the ISO 8601 accrualPeriodicity code where a publisher declares one at all. Resolution: the distributions array naming each file with its format and media type, the contactPoint naming the accountable human, the landingPage, and the reuse statement carried on about 91% of sampled records.

Two gaps deserve respect before modeling: issued is missing on roughly half the corpus and accrualPeriodicity on more than half, so absence of a declared cadence is the normal case rather than an anomaly - treat modified as the only universal recency signal, and expect roughly 9% of records to declare no reuse terms at all, worth checking per record before redistribution plans are made. Everything beyond the spine folds under additional fields on request: raw harvested DCAT payloads keyed by record UUID, expanded distribution detail with media types, and any pivot - one row per keyword match ranked by relevance or modified date. Name the shape when you request the sample.

Geography - the United States, layered by government tier. Federal agencies dominate the food slice, but state, county, city, university, tribal and non-profit publishers harvest into the same catalog, so one query returns an FSIS national inspection directory beside Chicago Food Inspections and NYC restaurant inspection results, each row labeling its publisher explicitly.

Temporal - varies per record, and says so honestly. Harvest dates run to the present, with some records modified daily while others have sat static since publication: the ERS import table's last modified date is February 2020 and the EPA waste-flow model still carries its 2018 version stamp. The modified column is what separates live series from archival ones - filter on evidence instead of trusting any blanket currency claim.

Granularity - one metadata record per dataset, 208 of them on the food keyword plus 30 on meat, inside a 552,271-record catalog. There are no measurements in this domain: anyone promising per-establishment inspection outcomes out of a catalog row is describing one level down, in the publisher-hosted files each distribution entry points toward. Datadory routes that underlying material into your delivery once the sample confirms which records you need.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the twelve-field dictionary travels unchanged through all three. Catalog pages never reach your pipeline - records arrive as flat normalized rows keyed on the record identifier, so consecutive pulls diff cleanly and newly harvested food datasets surface as inserts rather than surprises. Cadence changes are a settings conversation, not a re-integration project, and a sample filtered to your keyword comes first either way.

• Supplier vetting before the first purchase order - the Meat, Poultry and Egg Inspection Directory enumerates federally inspected establishments by name and number, so a specialty buyer can check who actually appears on the federal register before terms are extended; see supply chain mapping.
• Regulatory-watch automation - FSIS Quarterly Enforcement Reports and establishment demographic files give compliance teams named records to diff between pulls instead of press-release reading; see competitor tracking.
• Nutrition and formulation reference work - the SR Legacy nutrient release and the Nutrient Data Set for Retail Meat Cuts anchor product claims to a published reference table; see market sizing for the demand side.
• Waste and sustainability baselines - EPA's sector-to-sector food waste flows give circularity reporting a citable starting matrix.
• Dataset-directory and discovery products - normalized rows behind food-data catalogs, one schema covering USDA, FDA-adjacent, EPA and municipal publishers.
• Citation-grade sourcing for published work - every figure traces to a named agency with a revision date attached; see citation grade research.

Compliance and QA leads at specialty retailers rank first - an inspection directory that names establishments and an enforcement-report trail turn vendor claims into checkable records. Category managers and buyers come next, screening prospective suppliers against the federal register before the first order. Market researchers and consultants get the fastest route to whatever the US government actually publishes on food - imports, availability, waste flows, inspection coverage - before sizing anything; see market researchers use cases. E-commerce and grocery operators pair product-level nutrition references against their assortment; see e commerce operators use cases. Data scientists and developers get a uniformly schemed corpus for retrieval, deduplication and directory products through one ingestion path; see data scientists use cases and developers builders use cases. Everything above ships daily, weekly, or hourly.

Against its sibling on the same shelf, Data.gov Retail Catalog (552k+ Datasets Federated Search) walks the 'retail' keyword to 264 records scattered across taxable sales, food-environment indices, tobacco licensing and fuel prices - the landscape lens. This collection trades breadth for density around one supply chain, packing inspection directories, enforcement reports and nutrient releases into 208 food-keyworded records. The scored head-to-head is worked through on the retail-catalog comparison: sample both, pick by fit.

The specialists beat it on depth once you know the series. FAOSTAT Food and Agriculture Statistics carries country-year production and trade panels; FDA Recalls, Market Withdrawals & Safety Alerts serves event-level recall notices with hazard reasons; Edamam Food Database API and Nutritionix Natural Language & Grocery API hold item-level grocery nutrition at application grain. The practical split: come here first to learn what exists and who publishes it, go to those once the series is named.

Three things deserve respect in these rows. First, this is an index, not the measurements: each record describes a dataset owned by another publisher, and the row-level depth lives one level down in the files each distribution entry names - Datadory routes that underlying material into your delivery rather than leaving you to chase it portal by portal. Second, uniformity holds at the spine, not beneath it: individual records vary widely in upkeep, which is why the modified date rides on every row - filter staleness on evidence, and expect roughly 9% of sampled records to declare no reuse terms at all, worth checking per record before redistribution plans are made. Third, the count is query-dependent: 208 is the exact-keyword 'food' universe confirmed by cursor-paging, not a published figure, and broader full-text matching sweeps in far more - pin your query definition against a scoped sample before building anything that assumes a fixed universe.

Because the raw artifact is a federated metadata catalog wearing hundreds of publishers' conventions, and most questions want one clean table. Datadory normalizes the twelve-field spine across every publisher tier, joins distribution detail and contact points as readable dimensions, pairs the discovered records with the actual measurement feeds they point toward - recall events, trade panels, item-level nutrition - and scopes everything to the topics, publishers and geography you actually work in. Start with a sample of this dataset, then browse the rest of the shelf on the Other Specialty Retail data hub, the best Other Specialty Retail datasets ranking, or how the custodian operates on the Data.gov source profile.