Cross-Aisle Recommender · Sample Shelf
What else should this shopper reach for?
Pick any product. The engine breaks it down to its flavor compounds and ranks the rest of the shelf by how much molecular ground they share — the quantitative "you'll also like this," reaching across aisles a purchase-history model would never connect.
How it scores
Each product is reduced to its flavor-driving ingredients. Similarity = the average best-match
compound overlap between the two products' ingredients — the share of flavor compounds they hold
in common (Jaccard), straight from the engine. Every shared-compound count and Jaccard shown here is
the same headline figure you can verify pair-by-pair on
/compare. Ingredient-level only; a finished product
would need GC-MS to profile exactly.
Powered by the CompKitchen engine · sample shelf for
illustration · your catalog builds the same way.