Eurostat PRODCOM - EU production statistics for construction machinery

Every machine that digs, lifts, crushes or drills in Europe has a paper trail, and this is it. PRODCOM - PRODuction COMmunautaire - is the EU's survey of manufactured-goods production, organized on a Prodcom List of roughly 4,000 products whose 8-digit codes descend from 6-digit CPA headings inside 4-digit NACE divisions, linked to NACE Rev. 2 and CPA 2008 since 2008 (CPA 2.1 from 2016). The national statistical institutes of the Member States - joined by Albania, Bosnia and Herzegovina, Montenegro, North Macedonia, Serbia, Norway and Iceland - survey enterprises and transmit results to Eurostat, which computes EU aggregates at EU-15, EU-28 or EU27_2020 level depending on the year.

This record follows the slice equipment makers care about: NACE division 28.92, manufacture of machinery for mining, quarrying and construction, with related earthmoving and heavy-machinery products in adjacent headings. Production is measured two ways at once - by value in thousand euro and by whatever volume unit fits the product: kilograms for steel-intensive categories, square metres where area matters, item counts for complete machines. At scale, about 4,000 products times roughly 30 reporting countries across decades of reference years, the full family runs to hundreds of thousands of observations.

Get a sample of this dataset - name the product headings and countries, and rows come back shaped exactly as delivered.

One observation per row: a product heading, a reporter, a reference year, a unit and a value with its status flag. The shape of the delivered extract:

freq     : A                  # annual reference period
nace_r2  : 28.92              # machinery for mining, quarrying and construction
prodcom  : 28.92.xx.xx        # 8-digit product heading inside the division
geo      : DE                 # Germany, national total
time     : 2023               # reference year
unit     : thousand EUR       # value measure
OBS_VALUE: <value>            # sold production for the cut
OBS_FLAG : <flag>             # status marker, e.g. estimate or confidential

freq     : A
nace_r2  : 28.92
prodcom  : 28.92.xx.xx
geo      : EU27_2020          # Eurostat aggregate beside the countries
time     : 2023
unit     : thousand EUR
OBS_VALUE: <value>
OBS_FLAG : <flag>

freq     : A
nace_r2  : 28.92
prodcom  : 28.92.yy.yy
geo      : FR                 # France
time     : 2023
unit     : items              # physical volume, product-specific
OBS_VALUE: <value>
OBS_FLAG : <flag>

The <value> placeholders are deliberate. Cells move with the product heading, reporter and year you ask for, so printing a number outside your cut would be fiction dressed as data - and the catalog itself holds no pre-captured values for this record. Read the shape instead: Germany's national value cut, the EU27_2020 aggregate beside it in the same dimension, and a French volume cut in item counts rather than currency. Request a sample and those three shapes return populated for whichever headings you name.

Eight documented dimensions define every observation, and seven carry their own example in the catalog record. Each row is one product-country-year cell, so pivoting to a country-by-year matrix for one machine category or stacking every 28.92 heading into a panel takes the same query shape. Dimension labels were reconstructed from PRODCOM's explanatory metadata rather than observed in a live pull, so they are confirmed against real rows at first delivery.

Geography - EU Member States plus Albania, Bosnia and Herzegovina, Montenegro, North Macedonia, Serbia, Norway and Iceland, with Eurostat-computed aggregates layered into the same geography dimension. One pull gets both a Europe-wide view and the per-country breakdown underneath it.

Temporal - annual reference years reaching back to 1995, roughly three decades per series. Aggregation geography shifts with the union itself - EU-15 for earlier years, EU-28 in between, EU27_2020 from 2020 - so long-run comparisons need break-aware stitching rather than one continuous code.

Granularity - one observation per Prodcom 8-digit product, per reporting country, per reference year. There is no company-level, plant-level or shipment-level detail anywhere in the survey; it measures industry output, not firms.

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

The underlying observations are annual, so most teams load history once and keep a scheduled top-up on whatever cadence their models expect. Every delivery ships the complete field dictionary above, sample rows for validation and the coverage profile mapped to your scope.

• European equipment-market sizing - production values by 8-digit product and member state put euro figures behind market-entry theses without stitching two dozen national releases; context lives on the construction machinery heavy transportation equipment data hub.
• Demand forecasting and model features - a dimension-coded annual panel back to 1995 trains construction-cycle models on multiple boom-bust rounds; patterns for this work sit on the data scientists use cases page.
• Product-line competitive tracking - watching which 8-digit headings grow in which countries reveals where excavator-versus-crane-versus-drilling-rig capacity is actually shifting; workflows on the competitive intel product teams use cases page.
• Supply-chain and sourcing analysis - country-level output per product family locates where specific machine categories concentrate, feeding supplier-selection and tariff-exposure work.
• Citation-grade reporting - an official EU statistical survey with published methodology behind every figure, built to survive review.

Developers and data-product builders get the highest-leverage read: a dimension-coded panel whose product-country-year keys never drift, clean enough to power a product feature; see developers builders use cases. Data scientists and ML engineers get three decades of annual observations per product heading, ready-made training columns for cycle models. Competitive-intelligence and product teams get the official category baseline their internal shipment numbers should be judged against. Market researchers and consultants get euro-denominated market sizing from the source of record. Investors and quant researchers get factory-gate output for screening machinery-exposed positions. Journalists, academics and students get citable official statistics with flags doing the caveating.

Within the industry's catalog this record owns the pan-European production job, and the neighbors own different jobs. US Census Annual Survey of Manufactures (NAICS 333120) answers the equivalent question for the United States through employment, payroll and value-of-shipments cubes. Statistics Canada manufacturers' sales reports monthly Canadian NAICS 33312 sales back to January 1992 - faster, but one country and one industry aggregate rather than product-level splits. FHWA Highway Statistics counts registered trucks, not machines built. Inside Eurostat itself, the road freight transport statistics panel measures tonne-kilometres moved rather than machinery manufactured. PRODCOM's differentiator is 8-digit product granularity across roughly 30 European reporters since 1995 - nobody else in the slice covers that combination. The ranked slate sits on the best construction machinery & heavy transportation equipment datasets page.

Four things worth having in hand.

First, the field definitions above were reconstructed from PRODCOM's explanatory metadata rather than observed in a live response, so they are confirmed against real rows at first delivery - the dictionary you sample is the dictionary you buy.

Second, pin the exact table before quoting numbers. During August 2026 research the DS-066341 identifier did not resolve on programmatic paths even though its catalog entry stayed live, and the explanatory metadata documents the family as DS-059358 for annual sold production and DS-059359 for annual total production. We confirm the working table and return populated rows in the sample, so nothing gets quoted off a dead identifier.

Third, aggregates sit beside countries: EU15, EU28 and EU27_2020 members share the geography dimension with sovereign states, and the applicable aggregate changes with the reference year. Filter deliberately before computing averages.

Fourth, values are thousand euro and product-specific volumes, not euros and tonnes uniformly - and confidential or estimated cells arrive flagged rather than removed. Whether suppression bites at the 28.92 headings specifically went unverified on the research pass, which is precisely what a scoped sample settles.