Data.gov U.S. Government Open Data Catalog

It is the statutory front door to American government data, and it is enormous. Operated by GSA's Technology Transformation Services, the catalog held roughly 552,000 datasets at the August 2026 review - counts observed between 552,259 and 552,276 as agencies published and retired entries - and registration is not optional: the OPEN Government Data Act (Title II of the Foundations for Evidence-Based Policymaking Act of 2018, Public Law 115-435) requires federal agencies to publish information as machine-readable data with metadata registered in the catalog.

For health care facilities work the reach matters more than the raw count. One query surface touches CMS provider tables, Veterans Affairs county statistics, HUD property records and Census demographics side by side with state hospital licensure lists and city clinic registries - publishers who would otherwise be hunted portal by portal. The organization facet alone sizes them: 293,639 datasets at the Census Bureau, 87,209 at NOAA. Get a sample of this dataset cut to your facilities question before anything else.

Two records from the facilities slice of the catalog, shown with the metadata each carries:

title       : FY 2025 Disability Compensation Recipients by County
publisher   : Department of Veterans Affairs
identifier  : 95an-zhhy
periodicity : R/P1Y (annual)
resources   : csv | json | xml
contactpoint: named agency contact with direct email
lastUpdated : 2026-06-19

title       : COVID-19 Hospital Capacity by Facility
publisher   : U.S. Department of Health and Human Services
grain       : one row per hospital per collection week
resources   : csv

Read together they show the range. The Veterans Affairs entry arrives fully dressed - identifier, annual cadence code, resource listings in three formats, a named contact and a June 2026 publication stamp. The HHS capacity series wears a thinner wrapper around one of the richest facility-week tables in American health data, which is the pattern to exploit here: modest catalog records sitting on top of deep underlying tables. Every result also carries a government-level tag, a description snippet and views counted over the trailing month, so popularity is legible before any commitment gets made.

Eleven verified Common Core fields, the schema every registered dataset answers to. They identify the dataset, qualify its publisher, pin its cadence and location, and attach a human contact to every entry. Record-level extras - keyword tags, landing pages, spatial classification, monthly view counts, harvest lineage - are real but fold under additional fields on request rather than padding every delivery.

Geography - every American administrative level in one result list: federal agencies, states, counties, cities, tribal governments, plus universities and non-profits. Map-based selection filters by bounding box with within-box and intersecting-box matching, and international portals sit indexed for reference beyond the US core.

Temporal - there is no single time span, because each record carries its own issued and modified stamps. Per-dataset coverage runs from near-real-time series to annual tables, and sorting by last published date puts the freshest registrations first.

Granularity - dataset-level records describing whole datasets, with resource-level format listings beneath them. The underlying grain is whatever the publisher chose - a county-year table in one entry, a facility-week panel in the next - so plan analysis around the specific datasets a search surfaces, not the catalog layer itself.

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

Pick the channel your team already runs and set the cadence - change either when the project changes. Deliveries arrive typed and keyed rather than as raw catalog dumps: the field dictionary attached, dates parsed, and a sample cut to your facilities scope shipped first so validation takes minutes instead of days.

Discovery earns its keep on specific jobs:

• Facility-landscape discovery - enumerate every publisher of hospital, clinic, nursing-home and behavioral-health tables before building a facility intelligence product; see market researchers use cases.
• Access-desert research - pair shortage-area and health-center filings with the Veterans Affairs, Census and HUD tables the same query surfaces alongside them.
• Vendor territory mapping - learn which states and cities publish machine-readable facility data before committing coverage plans; written up on our sales growth teams use cases page.
• Pipeline scoping - turn a search into an ingestion roadmap: what exists, which agency owns it, which format it lands in; see developers builders use cases.
• Citation-grade sourcing - anchor stories, theses and filings to named agency publications with contacts attached; see journalists academics use cases.
• Diligence breadth checks - confirm what facility records exist in a market before trusting any single vendor's version of it; see investors quants use cases.

Market researchers and consultants get the widest aperture in the industry: every publisher level searchable at once instead of portal by portal. Developers building data products get a stable index to scope ingestion roadmaps against. Journalists, academics and students get named agencies, named contacts and citable provenance on every number. Sales and growth teams get a jurisdiction map of who publishes usable facility data. Investors and quant researchers get a diligence layer showing where a thesis can be checked against primary records, and competitive intelligence and product teams watch which agencies register new tables first - a read on where the category is moving.

Four things worth knowing upfront.

First, the headline count drifts. Between 552,259 and 552,276 datasets were observed across the August 2026 review, and the number moves daily as agencies publish and retire entries - size a publisher through the organization facet rather than quoting the total.

Second, this is an index, not a warehouse. Records describe datasets living on their original publishers' portals, so structure and completeness vary by publisher; the catalog guarantees you can find them, not that any two will match.

Third, metadata completeness is uneven by design. A rich entry arrives with identifiers, contacts, cadence codes and format listings; a lean one may carry little beyond a title and a modification stamp. Plan validation around the thin rows, not just the rich ones.

Fourth, catalog-level pulls were validated record by record during the review rather than end-to-end, because the catalog's own machinery moved mid-review. Name your facilities scope when requesting a sample and the extract arrives cut to exactly that shape.