Data.gov Catalog - IT Services structured datasets

Data.gov Catalog - IT Services structured datasets sits on Datadory's IT consulting & other services shelf as the map layer beneath everything else in the industry. The parent catalog is run by the U.S. General Services Administration and held 552,271 harvested dataset records at research time (August 2026), drawn from federal agencies plus state, city and county publishers. Queried for 'information technology services', it returns a relevance-ranked band led by USDA Farm Service Agency's Information Technology Internal Services and Support (USDA-FSA-00055) and HHS ONC's Health IT Catalog, with the City of Chicago's Public Technology Resources, NCO NITRD's software-engineering catalog and a state CIO study filling out the captured set.

Every hit normalizes to one DCAT-US record: title, abstract, identifier, publishing organization, availability designation, keywords, modification date, distribution count and harvest timestamp. That uniformity is the product - the catalog itself is a discovery layer pointing onward to agency systems, and this slice turns it into rows. Get a sample of this dataset scoped to the query terms your work actually runs.

Five records exactly as captured during the August 2026 research pass, kept in their raw shape:

Read them as one argument about shape. The lead record is a federal agency documenting its own internal IT support function - identifier USDA-FSA-00055, keywords 'administrative, information technology', untouched since November 2014. Records two and three bracket the distribution range: HHS ONC's Health IT Catalog attaches no files at all while the City of Chicago's Public Technology Resources carries six, the fullest capture in the set. Record four is a research-program catalog from NCO NITRD with one distribution, and the fifth is a 2012 state CIO study still indexed and findable thirteen years on. The accessLevel column does real work across all five: two public, one non-public, and that difference decides integrability before anyone writes a line against a record.

Nine verified fields ride on every record, and the dictionary below documents each with its literal example value from this slice. Three deserve attention before anything gets modeled.

First, accessLevel is a working filter, not decoration: public, restricted public or non-public, declared by the publishing agency itself. The lead record in our capture - USDA-FSA-00055 - reads non-public, which tells you before any integration work that its underlying files were never published. Second, distribution_count separates file-bearing records from metadata-only ones; the range in our capture runs from zero (both USDA and HHS ONC) to Chicago's six. Third, last_harvested_date gives every row an audit stamp - August 01, 2026 in our example - so 'how fresh is this' resolves as a column read.

Everything beyond the nine folds under additional fields on request: the continuation token used to page deep result sets, federal bureau and program classification codes, named contact points, and the per-record detail flags that distinguish downloadable entries. Name what you need and the delivered schema extends to match.

• Geography · United States, spanning federal agencies plus state, city and county publishers. The IT-services band skews federal - four of five captured records come from federal organizations - with Chicago's portal entry showing the municipal tier one query away.
• Temporal · Two clocks travel separately on every row: the agency's own modified date (2014-11-24 on the USDA lead; a 2012 vintage on the state CIO study) and the catalog's last_harvested_date (August 01, 2026 in our capture). A recent sweep says nothing about the underlying numbers; track both.
• Granularity · One catalog record per matching dataset, with distribution counts at the file level. Beneath that, structure follows the publisher - which is precisely what the delivered normalization layer resolves.

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

You pick the channel and set the cadence; the catalog's own harvesting rhythm belongs to GSA, but the delivery rhythm belongs to you. Most teams take the slice as a standing pull and let new matches rotate into place on whatever schedule their models expect - hourly mirrors for monitoring workflows, weekly snapshots for analysis, daily drops for pipeline hygiene. The nine-field schema stays stable across all three channels, so switching delivery modes never means rewriting downstream consumers. Start with a sample of the rows above; keep the feed if the dictionary fits the job.

• Federal IT-market mapping - publisher names plus keyword lists enumerate which agencies document IT spending, procurement and workforce activity before anyone commissions primary research.
• Availability screening - accessLevel beside distribution_count separates publishable records from metadata-only or restricted ones in a single read, which is the difference between a pipeline that works and one that stalls on a non-public designation.
• Provenance for client decks - identifier, publisher and harvest timestamp on every row mean a cited figure traces to a named agency in two reads; see market researchers use cases.
• Spend-model features - federal IT-spending and procurement records arrive normalized enough to join into models without bespoke extraction per agency; see data scientists use cases.
• Citation-grade sourcing - stories and papers anchor to records whose provenance prints on their face; see journalists academics use cases.

Developers and builders hold the strongest match at relevance 3: one normalized record shape across every agency beats maintaining an integration apiece, and the enumeration-first workflow suits how they already build. Journalists, academics and students also match at relevance 3 - a federated official catalog where every record carries provenance metadata is exactly what citation discipline requires. Market researchers and consultants and data scientists and ML engineers hold relevance 2, the former anchoring analyses in discoverable official sources, the latter harvesting spending and procurement records for models. Sales and growth teams and investors and quant researchers hold relevance 1 and use it honestly - an indirect prospecting angle onto federal IT work, an occasional diligence table - while knowing nothing here reaches contact level or trading cadence. Persona workflows sit at sales growth teams use cases.

Within the four primary datasets Datadory catalogs for this industry, this is the router - everything else is a destination. Census Annual Integrated Economic Survey (AIES) scores 10 by publishing annual revenue, payroll and employment by six-digit NAICS - including NAICS 5415 computer systems design - as bulk public-domain files reaching back through archived tables to 1992; the catalog scores 6 because it points at records like those rather than being one. H1BData.info - H-1B Disclosure Database holds more than 4.8 million Department of Labor wage disclosures from October 2013 through September 2025 - compensation texture no catalog carries. Clutch.co Top IT Services Directories maps private vendor landscapes with ratings and rate bands. Come here first to learn which agency owns what; go to those when you need the numbers themselves. The full ranking sits on the best it consulting other services datasets list.

Four honest caveats.

First, the yield is a subset, not a count. Roughly 552,271 records sit in the full catalog, and the IT-services query returns a ranked slice of them - the catalog does not expose a total for the subset, so treat sample results as the honest measure of depth rather than a denominator.

Second, availability varies by design. The accessLevel enum includes non-public for a reason: the lead record in our capture carries it, meaning its underlying files were never published. Screening on that column first saves everyone the surprise.

Third, two clocks per row. An agency's modified date and the catalog's last_harvested_date travel separately - a 2014 revision harvested in August 2026 is still a 2014 dataset. Date claims accordingly.

Fourth, depth sits one level down. This is a discovery layer; most records point onward to the agency systems where files actually live. Pairing the routing with the destination source keeps citations defensible - which is exactly why it ranks second on the industry shelf despite scoring 6 against a 7.81 catalog mean across 1,744 datasets.

Because a federated catalog only earns its keep once someone normalizes it, watches it and routes it. Datadory flattens every record to the nine-field dictionary above, tracks harvest stamps so you learn when an agency's record moved, screens accessLevel and distribution_count before a pipeline commits to a record that cannot deliver, and joins each row through to the deeper source it describes when you ask. Start with a sample scoped to the query terms and fields you name, then browse the rest of the industry on the IT consulting & other services data hub or the free IT consulting datasets rundown.