Product Information Management and Catalog Enrichment Pipeline
Challenge
Each supplier used different taxonomies and units. A canonical attribute model, per-supplier mapping templates and confidence-based auto-approval kept the steward workload manageable.
Approach
Supplier feeds (CSV, Excel, BMEcat, APIs) land in a staging area where mapping rules and ML-assisted attribute extraction normalise units, categories and specifications. Data stewards work in a review UI with completeness scores per category. Approved products publish to an OpenSearch-powered storefront search with faceting, and to marketplace channels through channel-specific exporters.
Outcome
Illustratively ~50–70% faster supplier catalog onboarding Better faceted search and fewer no-results queries Consistent product data across web, marketplaces and print catalogs Completeness scores show stewards exactly where to focus Fewer customer queries about specifications