This proof-of-concept explores two complementary approaches to product recommendations: Day 1 (Semantic) finds products with similar text descriptions, while Day 2 (Graph-based) discovers connections through shared occasions, seasons, and themes. With only 18% overlap, these methods complement rather than replace each other. The visualizations below demonstrate how graph-based recommendations enable cross-category discovery that traditional similarity algorithms cannot detect.
POC Completed: January 22, 2026 | 20,000 products analyzed | 2 AI approaches tested | 18% overlap = Complementary value ✓