Eurostat Database — Bulk Download (Eurobase)

The bulk download facility is the whole-database door into Eurobase, the European Commission's statistical warehouse — 8,150 dataflows spanning everything from retail-trade turnover to international trade in goods. Where a data browser answers one filtered question at a time, the bulk route ships complete datasets as files: TSV and SDMX for the full database, plus a dedicated Comext domain that publishes international trade in goods as monthly and annual CSV files running from January 1988 up to the latest reference month, each file carrying a complete set of data for the period in one go. A table-of-contents inventory in plain-text and XML versions maps the catalogue, code lists sit in the metadata section, and reference metadata ships in SDMX and HTML alongside the numbers. For apparel, accessories & luxury goods work, that means the entire EU statistical base behind a market study — production, trade, retail turnover, clothing prices — arrives in one pass instead of dimension-by-dimension queries. Datadory turns it into a managed feed rather than a filing exercise.

Two observations from the August 2026 verification pull, shaped the way they land from a bulk extract — full dimension key, flags attached.

dataset       | freq | geo       | time_period | obs_value | obs_status | conf_status
sts_trtu_m    | M    | EU27_2020 | 2026-06     | 105.5     | p          | N
prc_hicp_manr | M    | EU27_2020 | 2025-12     | 0.3       | p          | N

Read the first row as EU-27 retail trade turnover for June 2026: 105.5 is the calendar-adjusted volume-of-sales index, base 2021=100. The second is the December 2025 HICP annual rate of change for ECOICOP subclass CP031 Clothing — euro-area clothing inflation at 0.3 percent. Every row carries its SDMX flags beside the value: obs_status draws on the code list b, d, e, f, i, m, n, p, u and conf_status on C, N, P, the SDMX-aligned sets adopted on 27 January 2025. Missing cells publish as the special value ':' rather than blanks, so a gap never masquerades as a zero.

Six dimensions document every observation in the warehouse, defined in the table below against live responses rather than remembered documentation. The per-dataset classification axes — NACE activities for retail turnover, ECOICOP purposes for HICP, reporter, partner and product keys for Comext — ride on top of those six; they are itemised with examples when you request a sample, so nothing arrives as an unlabeled column.

Geography: EU-27, the euro area, every EU member state and numerous non-EU countries. Geo codes like EU27_2020 resolve aggregates and individual reporters alike, enough to model intra-EU apparel flows next to imports from major sourcing markets.

Temporal: the Comext domain reaches from January 1988 to the latest reference month — nearly four decades of EU goods trade. Other collections generally begin in the 1990s. History arrives whole: one monthly or annual Comext file contains the complete set of data for the reference period.

Granularity: complete multi-dimensional dataset files, one per dataset per period for Comext. The envelope behind that is large — 8,150 dataflows listed in the Eurostat structure as of the August 2026 research pass — and the plain-text and XML table-of-contents inventory lets you see the full catalogue before committing to any of it.

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

Datadory handles the extract, normalisation and load: flag codes resolved to labels, ':' cells typed as null, dimension keys kept stable when the catalogue moves underneath them. You set the cadence per feed — most apparel teams take trade and turnover daily and reserve hourly delivery for the short-term indicators they alert on.

Four jobs it does better than any filtered view.

Run a full-sector study. Market researchers treat it as the complete EU statistical base: when a study refuses to stay inside one dataset — production plus trade plus prices plus demographic context — the bulk route hands over the whole base in coherent formats instead of a pile of hand-stitched exports.

Ingest decades of trade history in one move. Each Comext monthly or annual file is a complete set for the period, so a sourcing-shift analysis reaching back toward January 1988 becomes a load job rather than a pagination project.

Simplify pipelines. Developers rate the bulk facility precisely because one well-formed file replaces hundreds of small requests, and the table-of-contents inventory doubles as a build-time catalogue of what exists.

Cite primary official statistics. Journalists and academics get the European Commission's own figures with reference metadata in SDMX and HTML attached, which is what citation-grade work requires.

Six of Datadory's eight apparel, accessories & luxury goods personas carry this feed.

1. Market Researchers & Consultants — scored relevance 3 in the persona pack for the use case "complete EU statistical base for any sector study." See market researchers use cases.
2. Data Scientists & ML Engineers — relevance 3, "full TSV/SDMX bulk" for feature engineering at scale. See data scientists use cases.
3. Journalists, Academics & Students — relevance 3, primary official statistics for citation-grade work. See journalists academics use cases.
4. Developers & Data Product Builders — relevance 2, "bulk facility simplifies pipelines." See developers builders use cases.
5. Investors & Quant Researchers — broad macro coverage as sector-level context around company-level signals.
6. E-commerce Operators — retail turnover indices among the domains, usable as demand context for assortment and promotion planning.

The registered head-to-head pairs this feed with DeepFashion Dataset (CUHK MMLab): 800,000+ annotated fashion images against official European statistics. They solve different problems — DeepFashion teaches vision models what clothes look like, this feed quantifies the industry those clothes are sold in — and the comparison page works through when each wins: vs DeepFashion Dataset (CUHK MMLab).

Inside the same warehouse, Eurostat Data Browser — Apparel, Textiles & Trade (PRODCOM, COMEXT) serves filtered, keyed, current observations; this page wins when the brief is everything at once — whole datasets, full history, one pass.

Three caveats worth stating plainly.

First, size cuts both ways. Catalogue estimates disagree by scope — roughly 8,150 dataflows cited for the Eurobase structure, about 5,900 disseminated dataflows in the science-and-technology view, some 1,400 tables in the transport description — because each counts a different slice of the same warehouse. Scope the extract before committing to it.

Second, gaps are flagged, not smoothed over. Low-count cells carry confidentiality suppression, and missing values publish as ':' rather than blanks; the flag code lists tell you which situation you are looking at.

Third, identifiers move. Older material still cites retired series codes — the current retail-trade family is STS_TRTU_A/M/Q, not the legacy sts_rtu_a — and the SDMX-aligned flag code lists date from 27 January 2025, so older extracts can carry different flag sets. Datadory maintains the mapping so your feed follows the living codes.