NACE codes: the EU industry classification, and what Rev. 2.1 changed

Register a company in an EU member state, pull a Eurostat table, screen an investment against the EU Taxonomy or file a merger in Brussels, and you will be asked for a NACE code. NACE - from the French Nomenclature statistique des activités économiques dans la Communauté européenne - is the EU’s classification of economic activities: the industry counterpart to ISCO-08 on the occupation axis, and the European sibling of the UN’s ISIC.

There is a catch that most NACE code lists online have not caught up with. Since 1 January 2025, European statistics run on NACE Rev. 2.1, adopted by Commission Delegated Regulation (EU) 2023/137. The headline change: the old Information and communication section was split in two, and every section letter after it shifted by one. Section K no longer means financial services - it means telecommunications and computing. A page still describing Rev. 2, and most of the first page of Google does, is now describing a classification the EU statistical system has moved off. This page covers how a NACE code is built, exactly what Rev. 2.1 changed, the full 22-section list with the old letters mapped, and how to use the codes against live job posting data.

How a NACE code is built

NACE has four levels. A section is a letter, A through V, and is not part of the numeric code - the digits imply it. A division is two digits, a group is three (written 62.0), and a class is four (written 62.01). So in Rev. 2, code 62.01 reads: section J, division 62 (computer programming and consultancy), group 62.0, class 62.01 (computer programming activities). Registries write the dot; plenty of datasets strip it to 6201. Both conventions are common enough that any join should normalise first.

Member states extend NACE downward with national versions - WZ in Germany, NAF in France, ATECO in Italy, and the UK’s SIC 2007 is built on NACE Rev. 2 even post-Brexit. They add national detail below the class, which means they stay comparable with each other and with NACE at four digits. For analysis, the two-digit division is the grain that travels: fine enough to separate banks from insurers, coarse enough that classification stays reliable.

Rev. 2 vs Rev. 2.1: what actually changed

NACE Rev. 2 was established by Regulation (EC) No 1893/2006 and carried European statistics from 2008 through 2024. Rev. 2.1 applies to statistics from reference year 2025, with a phase-in across statistical domains and national business registers that runs to 2031. The structural totals:

The big structural move is the split of old section J, Information and communication. Its six divisions now form two sections: a new J, Publishing, broadcasting and content production and distribution (divisions 58-60), and a new K, Telecommunication, computer programming, consulting, computing infrastructure and other information service activities (divisions 61-63). Every section from the old K onward shifts one letter down, and the last section - activities of extraterritorial organisations - becomes V.

Below the sections, the revision reads like a changelog for the 2020s economy. Division 45, motor vehicle trade, was abolished: wholesale moved to 46, retail to 47, and repair to division 95, so your car mechanic now sits in Other service activities. Trade itself is limited to physical goods - selling digital goods, downloads and streaming belongs to the new section J. Property development left Construction for Real estate. And there are new classes for electricity from renewables (35.12), electricity storage (35.16), additive manufacturing machinery (28.97) and solid fuels from vegetable biomass (16.26).

The 22 sections of NACE Rev. 2.1

The list below is the current classification, with the letter each section carried in Rev. 2 - the column to check before trusting any dataset keyed by section letter.

NACE vs ISIC, SIC and NAICS

NACE Rev. 2 is the EU’s implementation of ISIC Rev.4: identical at the section and division levels, with extra European detail below - 615 classes against ISIC’s 419. Rev. 2.1 keeps the same arrangement against ISIC Rev.5, which is where the shared changes like the J split and the abolition of division 45 come from. The UK’s SIC 2007 is built on NACE Rev. 2, so it stays ISIC-compatible at two digits. NAICS, the North American system, is a genuinely different structure connected only by concordance tables. The practical advice from the ISIC post holds unchanged: standardise multi-country data on the two-digit division and crosswalk outward, because the division layer is where NACE, ISIC and the national versions all agree.

Querying live postings by NACE division

Official NACE-keyed statistics are published on survey cadence. Job postings give the same industry axis a live reading. Every classified posting in the JobsPipe index carries an ISIC Rev.4 division alongside its ISCO-08 occupation code - how the classification pipeline works is its own post - and because NACE Rev. 2 divisions are the same two-digit codes, a NACE-keyed analysis plugs straight in. Computing and telecom hiring across three EU markets is one filter each:

curl -X POST https://api.jobspipe.dev/v1/jobs/search \
  -H "Authorization: Bearer $JOBSPIPE_API_KEY" \
  -d '{
    "isic_division_or": ["61", "62", "63"],
    "job_country_code_or": ["de", "fr", "nl"]
  }'

The two-axis version is the one survey statistics cannot give you weekly: which industries are hiring software developers, rather than which industries are hiring.

curl -X POST https://api.jobspipe.dev/v1/jobs/search \
  -H "Authorization: Bearer $JOBSPIPE_API_KEY" \
  -d '{
    "occupation_code_or": ["2512"],
    "isic_division_or": ["64", "65", "66"],
    "job_country_code_or": ["de"]
  }'

Practical traps

Section letters moved in 2025. Anything keyed by section letter across the Rev. 2 to Rev. 2.1 boundary lies silently: K flips from finance to telecoms and computing, R from arts to health care. Key analysis by division digits, which mostly survived the revision, and treat the letter as presentation.

Divisions are strings. Codes 01 through 09 carry leading zeros, and an integer cast turns crop production into an invalid code. Same trap as ISIC and ISCO, same fix: text everywhere.

Normalise the dot. Company registers publish 62.01, many datasets store 6201. Strip or add consistently before joining, or the join quietly drops rows.

Mind the transition window. Until national registers finish migrating, the same company can carry a Rev. 2 code in one source and a Rev. 2.1 code in another. Eurostat publishes correspondence tables; a time series that mixes revisions without converting has a break in 2025 whether it is labelled or not.

Where to go next

The authoritative Rev. 2.1 structure and explanatory notes are free from Eurostat. For the wider family tree, start with ISIC codes; for the occupation axis that crosses it, ISCO-08. And if you want the industry codes on live postings rather than last year’s survey, the free tier includes 1,000 jobs a month - enough to rank EU hiring by division and cross it with a few occupations every day.