Why commodity classification is difficult

Intrastat declarations generally require an 8-digit commodity code. Product descriptions in ERP systems, however, are often too short, inconsistent, or written for internal purchasing rather than statistical classification. Similar products can also require different codes because of material, function, composition, or technical characteristics.

Manual classification is slow when every item starts from a keyword search. Reusing an old code without checking the underlying product can be just as risky, particularly when nomenclature changes or master data has been copied between products.

Better suggestions need better context

AI-assisted matching becomes more useful when it considers more than the product name. Relevant context can include:

• Material and composition
• Product function and intended use
• Technical attributes, dimensions, or capacity
• Supplier and product-family information
• Previously reviewed classifications for comparable goods

Intrastat.ai uses available product and transaction context to rank likely commodity codes. The result is a reviewable suggestion, not an unexplained final decision.

Put human attention where it matters

High-confidence matches for established products can move through a streamlined review, while ambiguous items are routed to specialists. Reviewers can compare alternatives, inspect supporting attributes, and document why a code was selected.

This human-in-the-loop approach helps teams improve consistency without hiding uncertainty. Confirmed decisions also create a stronger reference set for future declarations.

Maintain classification quality

Classification is an ongoing data-management task. Review code changes at the start of each nomenclature year, recheck products whose specifications changed, and retain evidence for important decisions. Final classification responsibility remains with the reporting business and should follow current official guidance.