Embodiments of the present invention disclose a method, computer program product, and system for catalog digitization. A method including receiving observed information including physical attributes of a plurality of products, assigning an importance to each physical attribute, organizing the products into product categories based on the physical attributes and their corresponding importance, determining a product profile for of the products based on the product category of each of the products and a likelihood of any two attributes being observed together with respect to each of the products, identifying a match between two products based on a similarity between the physical attributes of the two products and their corresponding importance, and a comparison between the product profiles of the two products, preparing a product description based on a combination of the physical attributes of the two matching products, and creating a digital product catalog comprising the product description.