Auto-classification: definition and role in PIM

In PIM practice, auto-classification suggests a class or category for each product from its name, description, attributes or images. Good implementations show a confidence score and route low-confidence suggestions to a steward for review.

Illustrative example: a distributor onboards 5,000 new products from a supplier without classification. Auto-classification assigns ETIM classes to most of them with high confidence, and stewards review the remainder manually.

In a selection, buyers should test auto-classification on their own products and target schemes. Points to check include accuracy on a labelled sample, confidence thresholds, handling of new standard releases and the effort to correct and retrain. Classification errors propagate into required features and channel exports, so review steps matter. Confidence thresholds should be set per class group, with every suggestion below the threshold routed to a steward.