AI-Driven Guide to eCl@ss Classification and BMEcat Export Automation

AI-driven guide to eCl@ss classification and BMEcat export automation

Your largest distributor just sent an email demanding your complete product catalog. They do not want a PDF or a spreadsheet. They require a valid BMEcat XML file, fully mapped to the latest eCl@ss standard, with a 100% spec fill rate for all mandatory features.

If you are a product manager or technical sales lead at a mid-size manufacturer, this request usually triggers a crisis. Your product data lives scattered across ERP systems, legacy Excel spreadsheets, and unstructured PDF spec sheets. You know that manually categorizing 5,000 hydraulic fittings or electrical cables into a highly rigid taxonomy will take hundreds of hours. Outsourcing to an agency will cost thousands of dollars, and buying a massive software suite will not solve the underlying problem: your raw data is missing the required structure.

You do not need a new interface to type data into. You need a way to organize the data you already have, normalize the missing variables, and output it in the exact machine-readable format your distributors require.

This guide breaks down how eCl@ss classification works, why BMEcat exports are non-negotiable for industrial wholesale, and how AI product data enrichment transforms messy internal databases into compliant export files without requiring a dedicated data department.

Deconstructing the eCl@ss hierarchy and BMEcat standards

Before automating your export pipeline, you must understand the rules of the destination. eCl@ss is the only worldwide ISO/IEC-compliant data standard for classifying products and services. Conforming strictly to ISO 13584-32/42, IEC 61360, and DIN 4002, it provides a standardized, machine-readable taxonomy that eliminates ambiguity in procurement and digital data exchange.

Currently used by 4,000–5,000 companies worldwide, the standard is practically mandatory for industrial suppliers operating in the European market and is rapidly expanding across North America and Asia.

The four-level classification model

Every product mapped to eCl@ss is assigned an 8-digit code representing a strict four-level hierarchy:

1. Segment (Level 1): Broad industry category (e.g., 27 - Electric engineering, automation, process control engineering).
2. Main Group (Level 2): High-level product group (e.g., 27-14 - Low-voltage switchgear).
3. Group (Level 3): Specific product family (e.g., 27-14-22 - Circuit breaker).
4. Subgroup / Commodity Class (Level 4): The exact product type (e.g., 27-14-22-01 - Miniature circuit breaker).

The true complexity—and value—of the standard resides at the fourth level. Once a product is classified into a Subgroup, it inherits a mandatory set of properties (features) and predefined values. For a miniature circuit breaker, you cannot simply write "Voltage: 230V." You must provide the exact property identifier for "Rated operational voltage" and use the standardized unit code for volts.

eCl@ss Release 16.0 by the numbers

Keeping up with classification standards is a moving target. The current version, eCl@ss Release 16.0, published in November 2025, covers approximately 40–48 segments. The sheer scale of the taxonomy illustrates why manual product data enrichment is becoming impossible:

• ~50,000 classes in total
• 23,000 unique properties
• 140,000 keywords for search and mapping
• 995 new classes added since the previous version (including 137 new classification classes)
• 1,253 new properties and 985 new values introduced
• 126 new value lists
• Over 155,000 change requests processed

BMEcat as the delivery mechanism

Classification is just the taxonomy; BMEcat is the vehicle. Developed by the German Federal Association for Materials Management, Purchasing and Logistics (BME), BMEcat is a standardized XML format used to exchange catalog data.

Distributors require BMEcat because it natively supports complex classification structures. When you generate a BMEcat export, it packages your SKUs, GTINs, media assets, prices, and eCl@ss properties into a single machine-readable document that can be ingested directly into a distributor's procurement system without human intervention.

Mapping the differences between eCl@ss and ETIM classification

Mittelstand manufacturers frequently receive overlapping requests: one distributor asks for eCl@ss, while another demands ETIM classification. Both are taxonomy standards, but they serve different purposes and operate via different exchange formats.

Understanding the difference determines how you structure your product data.

If you manufacture electrical cables, your products exist in both systems. However, the properties required by each standard will differ in format and unit requirements. Trying to maintain these mappings via a master Excel sheet quickly leads to broken formulas and outdated catalogs, especially as both standards update annually. Centralizing your product data into a format that can map to multiple standards simultaneously is the only viable approach for scaling manufacturers. For a broader overview on consolidating this data, see our Product Data Management for Manufacturers: A Practical Guide.

Quantifying the cost of manual eCl@ss mapping

Let us assume you have a catalog of 5,000 hydraulic fittings. Currently, your spec fill rate is roughly 40%. The descriptions in your ERP look like this: FIT HYD 1/2in M-NPT ST SS316.

To generate a valid BMEcat file compliant with Release 16.0, you must extract the thread type, material, gender, size, and pressure ratings from that text string, convert any imperial measurements to metric if required by the standard, map the product to the correct 8-digit Subgroup, and populate the specific property IDs.

You generally face three traditional paths:

1. The manual agency route You can export your raw data and send it to an agency. A typical manual agency charges $2-3/SKU for basic data normalization and classification. For 5,000 SKUs, you will spend $10,000 to $15,000. You will wait 4 to 6 weeks for the return file. Furthermore, the agency will rely on humans reading spec sheets, meaning error rates typically hover around 3-5%. More importantly, the moment you launch a new product line next quarter, you have to pay the agency again.

2. The internal team route You assign this task to your existing product managers. Assuming an expert can fully classify, enrich, and validate 50 SKUs per day in a spreadsheet, 5,000 SKUs will take 100 working days. This pulls highly paid technical staff away from core product development to do data entry.

3. The massive software route You are told you need an enterprise software implementation. You spend a six-figure sum and six months setting up the system. When it goes live, the system is empty. The software gives you a structured database, but you still have to manually map those 5,000 hydraulic fittings into the system. It solves data governance but fails to solve data enrichment.

The pressure to resolve this bottleneck is accelerating. Regulatory pushes for the Digital Product Passport (DPP)—which relies heavily on standardized taxonomies like eCl@ss—mean compliance is shifting from a competitive advantage to a legal requirement. Industry events like the BME SOLUTION DAYS 2026 (Düsseldorf, May 19–20, 2026) and digitalBAU 2026 (Cologne) are actively dedicating workshops to eCl@ss integration, signaling that distributors expect suppliers to have this solved.

Automating BMEcat exports with AI product data enrichment

The alternative to manual data entry is AI product data enrichment. Instead of paying humans to read technical strings or relying on static rules engines that break when a typo occurs, AI models designed for industrial taxonomies automate the extraction, normalization, and classification process.

This is exactly what FacetFlux is built to execute.

Step 1: Data extraction and attribute mapping

The AI ingests your raw data—whether via API or a flat file upload. It reads unstructured text (FIT HYD 1/2in M-NPT ST SS316) and parses it into discrete technical attributes. It understands that "SS316" means Stainless Steel 316, "1/2in" is a dimension, and "M-NPT" is Male National Pipe Thread.

Step 2: Data normalization

Extraction alone isn't enough — the values still have to match what the destination schema expects. Normalization converts every attribute into the canonical form required by eCl@ss and BMEcat. Imperial measurements convert to metric where the property demands it (1/2 inch → 12.7 mm). Unit strings collapse to canonical variants (PSI, psi, lbf/in² all resolve to the unit code expected by the target property). Value-list properties snap to the standardized eCl@ss values (SS316, stainless 316, 1.4401 all map to Stainless steel 316). Without normalization your raw text might look complete, but the distributor's BMEcat parser will reject it the moment it checks unit codes or picklist values.

Step 3: Taxonomy classification

Instead of a human guessing which of the ~50,000 classes fits best, the AI evaluates the product's enriched attributes and maps it to the precise 8-digit Subgroup. It maps the normalized attributes to the specific property IDs assigned to that class.

Step 4: Generating machine-readable XML

Once the data is structured, generating the BMEcat export is simply a matter of outputting the syntax.

Below is an example of what the AI generates from your raw text to satisfy distributor requirements. Notice how the eCl@ss version, classification class, and property IDs are strictly formatted within the <ARTICLE_FEATURES> block of the BMEcat 2005 standard.

<ARTICLE>
 <SUPPLIER_AID>HYD-5001</SUPPLIER_AID>
 <ARTICLE_DETAILS>
 <DESCRIPTION_SHORT>Male NPT Straight Hydraulic Fitting, Stainless 316, 1/2"</DESCRIPTION_SHORT>
 <EAN>4012345678901</EAN>
 <MANUFACTURER_NAME>Industrial Fittings Corp</MANUFACTURER_NAME>
 </ARTICLE_DETAILS>
 <ARTICLE_FEATURES>
 <REFERENCE_FEATURE_SYSTEM_NAME>eCl@ss</REFERENCE_FEATURE_SYSTEM_NAME>
 <REFERENCE_FEATURE_SYSTEM_VERSION>16.0</REFERENCE_FEATURE_SYSTEM_VERSION>
 <REFERENCE_FEATURE_GROUP_ID>21-06-03-02</REFERENCE_FEATURE_GROUP_ID>
 <FEATURE>
 <FNAME>Material</FNAME>
 <FVALUE>Stainless steel 316</FVALUE>
 </FEATURE>
 <FEATURE>
 <FNAME>Thread size</FNAME>
 <FVALUE>1/2 inch</FVALUE>
 </FEATURE>
 </ARTICLE_FEATURES>
</ARTICLE>

This entire transformation happens in seconds per SKU, not hours.

Structuring exports without a dedicated data team

If your engineering or product management team is tasked with meeting these distributor requirements, you need a workflow that circumvents the need for a dedicated data science department. The goal is to build a repeatable pipeline so that next month, when you add 200 new SKUs, the BMEcat file can be regenerated instantly.

Aggregate what you have

Do not wait until your internal data is perfect. Pull your item masters from the ERP. Export your price lists from Excel. Gather the PDF spec sheets for your most complex product lines. The AI product data enrichment engine does not require perfectly structured inputs; it requires raw material. Ensure every item has a unique SKU and, if applicable, a valid GTIN.

Run automated enrichment and gap analysis

Upload the aggregated file. The AI will classify the products and map the available data to the taxonomy. Crucially, it will flag what is missing. If eCl@ss mandates an operating temperature range for a specific sensor and that data does not exist anywhere in your uploaded files, you will be notified. This focuses human effort solely on sourcing missing technical facts, rather than wasting time formatting units or looking up 8-digit class codes.

Validate the output against standard requirements

Before transmitting data to your distributor, it must be validated against the required schema. A structural error in the XML or a missing mandatory property will cause the distributor's automated procurement system to reject the file. Manufacturers should pass their enriched data through a validation check. While FacetFlux handles this internally during enrichment, ensuring strict adherence to the BMEcat 2005 specification ensures zero friction upon delivery.

Establish a continuous update cycle

Taxonomies are not static. With over 100,000 change requests already submitted ahead of the April 30, 2026 deadline for Release 17.0, the classifications you use today will evolve. Categories will merge, new properties will become mandatory, and deprecated values will be rejected.

By utilizing an API-driven enrichment tool, your catalog remains dynamic. When a new taxonomy version is released, you simply re-run the classification against your centralized, normalized product data to generate an updated BMEcat file. You eliminate the historical practice of starting from scratch every three years.

Common questions about eCl@ss and BMEcat

What is the latest version of eCl@ss? The current active version is Release 16.0, published in November 2025. It contains approximately 50,000 classes and 23,000 properties. It added 995 new classes and supports 31 languages. Development for Release 17.0 is underway, with a deadline for change requests set for April 30, 2026.

Do I need an enterprise software suite to create BMEcat exports? No. While large software platforms offer robust data governance, they do not automatically structure or classify your data. If you are a mid-size manufacturer, you can use AI product data enrichment via API or file upload to structure your existing ERP/Excel data and generate a valid BMEcat file directly, bypassing the need for a six-figure implementation.

How does data normalization improve my spec fill rate? Many product catalogs have the correct data, but in the wrong format. If a standard requires a length value in millimeters (e.g., "12.7") and your database says "1/2 inch", a standard automated export will fail, resulting in a blank field. Normalization translates your native units and formats into the taxonomy's required structure, instantly raising your fill rate without requiring new data entry.

Can I automate mapping to both eCl@ss and ETIM classification? Yes. Because both standards utilize specific property requirements, AI enrichment can parse your raw product data once, extract the core technical attributes, and simultaneously map those normalized values to both taxonomies. You can then export a BMEcat for one distributor and an ETIM xChange file for another.

Does eCl@ss support Digital Product Passport (DPP) compliance? Yes. The standard is heavily involved in the digital twin and DPP initiatives across Europe. By providing a globally recognized, ISO-compliant taxonomy (ISO 13584-32/42), it gives manufacturers the foundational data structure necessary to meet upcoming regulatory requirements for traceability and environmental reporting.

Why is BMEcat preferred over standard CSV files? CSV files are flat and struggle to represent hierarchical data efficiently. BMEcat is an XML format designed specifically for complex catalogs. It natively supports multi-level taxonomies, media asset linking, tiered pricing structures, and multilingual product features within a single, validated package that distributor ERP systems can automatically digest.

Why shouldn't I just use a manual agency for mapping? Time, cost, and scale. An agency charges roughly $2-3/SKU and takes weeks to return a file. They rely on human data entry, which introduces error. Furthermore, manual mapping is a one-time service; when your catalog changes or a new taxonomy version is released, you must pay the agency again. AI product data enrichment offers a scalable, repeatable pipeline at a fraction of the cost.

Upload your product data. See it organized in seconds.