| # Cross-System Product Matching Procedure | |
| ## Steps | |
| 1. Extract product records from both source systems | |
| 2. Normalize product names (remove system-specific prefixes, standardize units) | |
| 3. Compare key attributes: category, sub_category, grade, concentration | |
| 4. Apply fuzzy matching on normalized names (threshold: 0.7) | |
| 5. For matches above threshold, verify with attribute comparison | |
| 6. For ambiguous cases (0.5-0.7), escalate to domain expert | |
| 7. Document decision and update mapping table | |
| ## Key Discriminators | |
| - Product grade (Food vs Technical vs USP) | |
| - Concentration/DE values | |
| - Source vs processed form (crude vs refined) | |
| - Regional naming conventions (German vs English vs abbreviated) | |
| ## Escalation Criteria | |
| - String similarity < 0.5 but attributes match | |
| - String similarity > 0.7 but attributes conflict | |
| - Legacy system codes with no clear mapping | |