Eurostat Structural Business Statistics (NACE J62/J63)

A census-shaped answer to one question: how big is the European digital-economy sector, country by country, year by year? Structural Business Statistics "describe the detailed structure, economic activity, and performance of businesses over time" and sit inside European Business Statistics - the statistical office's own claim to "the most comprehensive picture of the European economy, both at country and EU level." The flagship table for this industry, annual enterprise statistics for special aggregates of NACE Rev.2 activities, holds 1,149,716 observations across 16 reference years.

Every observation is dimensioned four ways: NACE activity (J62 computer programming, consultancy and related activities; J63 information service activities - the division containing data processing, hosting and related activities), structural indicator, country, and reference year, finished with the value and a per-cell quality flag. A verified pull for reference year 2020 returned 728,484 J62 enterprises in the EU-27 turning over EUR 600 billion, and 88,751 J63 enterprises - with Germany alone hosting 10,847 J63 enterprises worth EUR 18,003.4 million.

That combination - one classification, 27 countries, three decades of comparability discipline behind it - is what makes this the anchor production-side measure for outsourcing work. Aggregate only: no firm names, no contacts, just structure. Get a sample of this dataset and read the rows against your own market definition before anything ships.

One row per country x activity x indicator x reference year, flat enough to enter a model without reshaping. Five verified observations for 2020:

geo      : EU27_2020
nace_r2  : J62 - computer programming, consultancy and related activities
indic_sb : V11110 Enterprises - number
time     : 2020
value    : 728484

geo      : EU27_2020
nace_r2  : J63 - information service activities
indic_sb : V11110 Enterprises - number
time     : 2020
value    : 88751

geo      : EU27_2020
nace_r2  : J62
indic_sb : V12110 Turnover or gross premiums written - million euro
time     : 2020
value    : 600000        (million euro; cell flag: 'e' estimated)

geo      : DE
nace_r2  : J63
indic_sb : V11110 Enterprises - number
time     : 2020
value    : 10847

geo      : DE
nace_r2  : J63
indic_sb : V12110 Turnover or gross premiums written - million euro
time     : 2020
value    : 18003.4       (million euro)

Read the first row as a sentence: nearly three-quarters of a million European firms sell computer programming and consultancy for a living. Read the fourth against it and you have Germany's share of the information-services base without touching a spreadsheet. And read the third row's flag for what it is - a cell that arrives marked estimated stays marked estimated in every delivery, so confidence intervals survive the pipeline intact.

Seven fields define the spine of every observation: which activity, which indicator, which country, which year, the value itself, and the quality flag attached to it. The dictionary below types all seven with worked examples. Definitions were verified against live output during research, so the table describes what actually arrives rather than what a brochure promises.

Two structural details reward attention. First, status rides on every cell - estimated, low-reliability and confidential cells are labelled at source, which turns disclosure control from a parsing headache into a filterable column. Second, the indicator axis is deliberately wide: 33 structural measures per cell means enterprise counts, turnover, persons employed and their companions arrive pre-paired on identical dimension keys.

Geography - the whole EU-27, each member state reporting beside the EU27_2020 aggregate, so a bloc-wide benchmark and a single-country cut come from the same grid. EFTA and candidate countries surface in the wider European business statistics family rather than this flagship table.

Temporal - sixteen reference years, 2005 through 2020, in the flagship special-aggregates table: the long-memory backbone for trend work. Later reference years live in the wider annual SBS family, so current-year sizing pairs this table's history with a successor's recency - a sequencing question, not a gap.

Granularity - country x NACE division x structural indicator x reference year. There is no firm identifier anywhere in the grain: these are sector structures, not company registers. J62 and J63 are divisions; how far the classification splits beneath J63 - toward the data processing and hosting classes this industry is named for - is confirmed against the classification list in your sample rather than assumed. Aggregate cells join cleanly onto registers and flow series keyed on geography and period.

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

You pick the channel and the cadence; extraction, cleaning and schema stability are our problem. When a revision lands on a historical cell or an indicator list grows, the feed you consume stays normalized - same columns, same types, same join keys. Bulk pulls land as files, continuous consumption runs through the API, and warehouse-native loads write straight into your own storage. Quality flags ship beside every value either way. A sample goes out first, sized to test in your pipelines the same day.

• EU ICT market sizing. Enterprise counts, turnover and persons employed for J62 and J63 give consultancies a defensible denominator for every European software-and-information-services market model - sized from official structure, not vendor marketing.
• Cross-country benchmarking. Every member state reports the same NACE Rev.2 codes into one harmonized frame, so Germany's 10,847 J63 enterprises compares definition-safe against France's, Poland's or the EU-27 aggregate.
• Territory prioritisation. Rank countries by enterprise counts and persons employed to size addressable markets before a sales or expansion plan commits budget to a geography.
• Quant and macro features. A tidy country x activity x indicator x year panel drops straight into models as a sector-cycle signal; turnover and value-added series track the EU IT-services cycle.
• Citation-grade reporting. Official EU statistics with published structure and stable definitions - the reference when a story, paper or prospectus needs numbers no one can argue provenance on.
• Demand-side pairing. Production-side structure joins onto services-trade flows and vendor directories: what the sector looks like where the work is done, next to where the demand sits. When the question turns UK-specific, the head-to-head with the UK business register settles the choice.

Market researchers and consultants get the standard sizing layer for European computer programming and information services - the numbers everyone else's decks cite; see market researchers using data processing outsourced services data. Data scientists and ML engineers get a harmonized panel with verified field documentation and no reshaping tax; see data scientists use cases. Developers and builders find a stable, well-documented structure that maps cleanly onto typed deliveries; see developers builders use cases. Journalists, academics and students cite official European statistics with a persistent dataset identifier; see journalists academics use cases. Investors and quants read turnover and value-added tracks as an IT-services cycle signal. Sales and growth teams turn country rankings into territory plans - with the honest caveat that aggregates name sectors, never accounts.

Because the useful test is a join, not a glance. A sample comes back cut to the countries, activities, indicators and years you nominate, with quality flags intact so you can see immediately how many cells survive your reliability threshold. Test whether your market model, territory plan or research panel keys cleanly onto country-plus-NACE-plus-period before anything reaches production.

Neighbors worth knowing: NOMIS - UK Business Register & Employment Survey by SIC trades the EU-27 frame for UK depth down to local authority; Companies House Public Data API (UK) flips to the entity level this dataset deliberately lacks; BEA International Transactions & Trade in Services (iTable) adds US demand-side flows; World Bank ICT Service Exports and UNCTADstat trade depth for near-global breadth. Start small: get a sample of this dataset scoped to the slice you will actually use, or browse the data processing outsourced services data hub for companions.