Product 360 Survivorship Across ERP, PIM, and Supplier Portal
Meridian Distribution runs Product 360 for 1.2 million SKUs. ERP supplies price and unit of measure, PIM supplies long descriptions and image URLs, and a supplier portal CSV supplies lead time and country of origin. Trust is configured globally: ERP rank 1, PIM rank 2, supplier rank 3. A nightly load from ERP sends null image URLs, and the 'most recent' survivorship rule overwrites PIM images with null. Supplier records from two different manufacturers merge when names are similar, mixing brand attributes. Stewards also see ERP's 40-character description surviving over PIM's 250-character version. Procurement needs lead time, which only the supplier portal provides, but values arrive as text such as '4-6 weeks.' The business wants one golden product record per SKU without losing source-specific detail.
1. How should survivorship and trust be configured to resolve these conflicts?
Read the worked answer
Configure trust and survivorship per attribute, not globally. Price and unit of measure: ERP first, then PIM, then supplier. Description: PIM first, supplier second, ERP last; never let ERP's 40-character text overwrite a longer PIM description just because it arrived later. Image URL: PIM only, with a rule that rejects null and blank values. Lead time: supplier portal is the only source, but add a validation or cleanse rule that converts '4-6 weeks' to a numeric range and rejects unparseable values. Disable 'most recent' for attributes where the latest load is not the most trustworthy; 'most recent' is appropriate only for owned attributes with reliable timestamps. For country of origin, prefer supplier but validate against a reference list. The trade-off is that field-level rules are more work to maintain, but a global source ranking cannot express that different sources own different attributes.
2. What would you change in matching to stop cross-manufacturer merges, and how would you handle existing bad merges?
Read the worked answer
Cross-manufacturer merges usually mean the match rule is too dependent on name and description. Add exact manufacturer part number and GTIN as high-weight identifiers; if either is present and conflicts, block the merge. Include brand or manufacturer as a required match dimension for fuzzy product matches. Do not use description for matching. For existing bad merges, create a review task rather than deleting records; use the unmerge capability or split the record, preserve cross-reference lineage, and correct the surviving golden record. Analyze the merged records to find the pattern before changing rules, then re-match in test. Add a validation rule that flags records with multiple manufacturer part numbers. The trade-off is more false negatives, but for product data a false merge contaminates pricing, inventory, and supplier reporting, so precision is more important than catching every duplicate.