A product knowledge graph encodes entities (products, brands, categories, materials) as nodes and their relationships (made_by, belongs_to, made_of, fits_with) as typed edges. It complements unstructured text data with a structured layer the search engine can traverse.
Typical wins in ecommerce: facet hierarchies (Apparel → Tops → T-Shirts → Graphic Tees) come from the graph; “people who bought X also viewed Y” edges drive related-product widgets; brand entities disambiguate queries like “apple” (the fruit vs. the company); seasonality edges power time-aware boosts.
You don’t need Neo4j to start. A well-designed product schema in your existing database, with clean foreign keys between brand, category, and product tables, is already a graph. Tools become useful when relationship traversal exceeds 2–3 hops or the graph grows past 10M nodes.