Product Recommendation Engines
Recommendation engines are best known for using the wisdom of the crowd to determine which products to recommend to a visitor. Item affinity, an approach made popular by Amazon, identifies items to present to a visitor based upon models such as, “visitors who viewed this bought that” or “visitors who bought this also bought that.” These recommendation engine models can be applied to websites to increase engagement and result in higher conversion rates, cart sizes, and order sizes.
Visitor affinity looks at a visitor’s product interest compared with other visitors and recommends products based upon visitors with similar interests. For example, if John and Mary have similar interests in products, then recommendations for John may be products that Mary has shown interest in that John has yet to see. Visitor affinity is useful for repeat visitors and can be combined with or cascade to item affinity models for newer or less recent visitors.
Combine Multivariate Testing with Recommendation Engines
By combining multivariate testing with recommendation engines, the content and placement of recommendations can be optimized for the highest conversion rate. Different variations of the recommendations and placement on the page are tested with live visitors to determine which combination yields the most engagement and ultimately higher revenue per visitor.
Different recommendation models can also be tested, since a different model may work better based upon the pages on which they are displayed. Product recommendations could be displayed throughout the website, such as the home page, category pages, product pages, basket pages, or on final checkout. In some cases product recommendations are filtered by categories to provide more contextual relevance to the visitor. Different approaches work better on different pages, and testing the models ensures the optimal content, placement, and model for recommendations.
See MaxRECOMMEND for more information.