UK National Data Library / Find Open Data (data.gov.uk)

The UK government's own answer to a question every analyst eventually asks alone in front of a search box: does the state already publish this? The National Data Library, fronted at data.gov.uk, describes itself as the home of UK public data to inform decisions and build services, and holds tens of thousands of records from UK government and public sector bodies. Discovery runs through a Directory search; a migration announced in March 2026 is moving the estate onto the Library architecture, with the original directory still searchable alongside while the work completes.

The broadcasting slice proves the catalogue is more than an index of PDFs. A single broadcasting query returns 23 records, nearly all published by Ofcom: Radio broadcasting carrying community radio revenues and expenses beside counts of commercial radio services broadcasting, Broadcast radio, the Public Service Broadcasting Tracker with respondent-level files from the 2013-2015 vintages, Broadcast radio transmitters: technical parameters, and Broadcast media: hours, revenue and spend. Every record documents its own abstract, publishing body, thematic classification, coverage window where stated, and a typed manifest of attached files. Inside Datadory's catalog - 1,744 datasets averaging 7.81 - this one scores 8/10, definitions verified rather than inferred. Get a sample of this dataset and the descent from catalogue row to usable columns is done for you.

The unit of observation is the catalogue record - one row per dataset, with the file manifest folded beneath it:

# one row = one catalogue record returned by a 'broadcasting' query, captured Aug 2026

title            : Radio broadcasting
publisher        : Ofcom
attached_files   : 4
files            : Community radio revenues and expenses (CSV)
                   Number of commercial radio services broadcasting (CSV)

title            : Broadcast radio
publisher        : Ofcom
abstract         : Number of commercial radio services broadcasting, commercial radio
                   revenues, and community radio revenues and expenses

title            : Public Service Broadcasting Tracker
publisher        : Ofcom
files            : psb_tracker_2014_csv | psb_tracker_2015_tab
                   psb_tracker_2013_tables (PDF)

title            : Broadcast radio transmitters: technical parameters
publisher        : Ofcom

# search-level shape
query 'broadcasting' -> 23 matching records, nearly all Ofcom-published

That shape is the point. A record tells you what exists, who publishes it, how many files hang off it and in what formats - enough to plan a load before a single file arrives, and honest enough to expose records whose coverage window is only implied by file names.

Fourteen documented structures, defined below with their role stated plainly. Record-level fields identify and classify (title, notes, organization.title, theme-primary) and date-stamp the shelf's own history (metadata_created, metadata_modified). Coverage fields state the period a record's data covers - explicitly optional per record, which is why the dictionary flags their presence rather than promising it. The per-file layer (resources[].format, resources[].mimetype, resources[].size, resources[].last_modified, counted up in num_resources) types every attachment, so mixed-format records stop being a surprise at load time.

Deliveries extend the dictionary with derived cuts - release-cadence reconciliation, stacked tracker vintages, canonical publisher registers, cross-source join keys - folded under additional fields on request and locked against live rows before pipeline work begins.

Three chips, drawn straight from the record:

• Geography - the United Kingdom throughout. The catalogue indexes UK government and public sector bodies, and the broadcasting holdings are Ofcom's UK market series.
• Temporal - the catalogue reaches back to the 2000s. Within the broadcast slice, the Public Service Broadcasting Tracker files run from the 2013-2015 vintages, and radio revenue tables begin in 2019.
• Granularity - two layers by construction. One catalogue row per dataset, a per-file manifest beneath it, and underneath both an underlying grain that varies by series: UK-wide aggregates in the revenue and station-count tables, individual respondents in the tracker microdata, individual masts in the transmitter parameters.

The layered grains are the feature worth designing around. The same query that surfaces a national aggregate hands you the address of respondent-level microdata - catalogue and contents in one frame.

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

The catalogue sets its own rhythm; your delivery cadence gets set separately, to match the decision you are feeding. Take flat files sized for overnight loads, live lookups for products, or a direct pipe into Snowflake, BigQuery or Redshift - the rows arrive identical either way, normalised to one row contract across the record layer and the file layer alike.

Every delivery ships the field dictionary unchanged plus sample rows for validation, so the first join happens against evidence rather than optimism. Derived cuts ride along on request: release-cadence reconciliation for records that omit it, tracker vintages stacked into one panel, join keys onto the Ofcom research shelf or BARB viewing rows. Name the records and layers you want, and your sample comes back shaped to them before any commitment.

Ranked by how directly one row settles the day job:

1. Market researchers sizing the UK audio market. Commercial radio service counts, community radio revenues and expenses, and broadcast media hours-revenue-and-spend come from the regulator that collects them - attribution included.
2. Strategy teams checking whether the state got there first. Twenty-three broadcasting records settle in minutes what a commissioned study would spend weeks rediscovering.
3. PSB and policy analysts. Respondent-level tracker files support custom attitude segments rather than borrowed top lines.
4. Broadcast engineers and coverage modellers. Transmitter technical parameters at mast level, beside the market series that give those masts their listeners.
5. Data engineers building publisher-intelligence panels. Record and file timestamps make catalogue maintenance itself measurable.
6. Quant and investment teams. Regulator-collected revenue series as ground truth under UK audio theses.

Market researchers and consultants get the cleanest fit: official UK radio and audio denominators with publishing bodies named, ready to sit beside BARB's demand-side audience estimates. Competitive intelligence and product teams follow closely - station counts and revenue tables chart who broadcasts where.

Journalists, academics and students gain citable public-sector records plus microdata for replication. Data scientists and ML engineers get typed, keyed layers that land without transformation theatre. Developers and builders power station lookups and coverage tools off transmitter parameters. Investors and quants read the revenue series as the floor beneath private-market valuations.

This record answers what does the UK state publish about broadcasting; its neighbours answer other questions. Ofcom Research and Data - Broadcasting and Media carries the publisher's full shelf in depth - 471 publications, Media Nations editions, Made Outside London registers - while this catalogue proves the holdings and types their files. BARB Weekly Top 50 Shows and Viewing Data supplies the demand side, fifty rows a week of what was actually watched. The Internet Archive TV News Archive makes 4.37 million news broadcasts caption-searchable, and the EBU Media Intelligence Service widens the frame to European public service media. Against entertainment-metadata references such as TMDB's TV section, the division of labour is simple: they describe programmes, this describes the UK broadcast market.

Four things, stated plainly.

First, it is a catalogue with contents inside it. A record row and the series it describes behave differently in a warehouse; say which you want and the sample ships exactly those layers, not a blend of both.

Second, some coverage fields are optional by design. Temporal windows and release-cadence labels appear only on some records, so where a record omits them the delivery flags the gap instead of papering over it with inference.

Third, the illustrative values above are shapes, not records. Titles, publishers and file lists were captured during the August 2026 research pass; live rows get pinned when your sample is cut, which is also where formatting locks for your pipeline.

Fourth, the migration is real and worth respecting. Since March 2026 the estate has been moving onto the National Data Library architecture; deliveries pin the state behind your rows with timestamps, so whatever shifts upstream stays diffable downstream. Get a sample before any commitment.

Because a catalogue deserves better than being browsed one tab at a time. Datadory resolves the whole stack into typed rows: record fields joined to their file manifests, publishers normalised, formats typed before anything lands, and the underlying series scoped to whatever your project actually touches - radio economics today, tracker microdata next quarter. Start with a sample cut to your brief, then browse the rest of the slice on the broadcasting data hub or the best broadcasting datasets ranking.