How is RecoMelon different from other recommendation engines?

Conventional recommendations often begin with behaviour history

Collaborative-filtering systems commonly learn from what customers viewed or bought together. That can work well for established products with large volumes of behaviour, but it creates limitations for new products, changing catalogues and visual buying decisions.

RecoMelon begins with the product

VisualDNA™ analyses product imagery and relationships such as silhouette, colour, cut, pattern and other visible characteristics. This gives the system useful product-to-product context before a SKU has accumulated purchase history.

Similarity is only the starting point

RecoMelon does not treat 'looks similar' as the entire answer. Dealbreaker™ Filters let shoppers preserve the characteristics creating desire while changing attributes that make the current product unsuitable.

The same intelligence works across more than a carousel

Visual catalogue understanding can power Reveal™, The Index™, AI collection discovery, RescueGrid™, SecondLook™ and other surfaces. Bundles™ uses related product intelligence for a different objective: finding complementary products rather than substitutes.

Merchants retain commercial control

Dynamic merchandising can combine relevance with boosts, suppression, catalogue priorities and market context instead of forcing teams to choose between manual merchandising and algorithmic ranking.

New products are not automatically disadvantaged

ColdStart™ allows new and low-history products to enter recommendation relationships from imagery rather than waiting for sales data to accumulate.