Product recommendations are growing in ubiquity, and for good reason. They help customers find what they’re looking for, they help brands highlight products, and they streamline the online shopping process. But, product recommendations aren’t one-size-fits-all — they depend on the algorithms that drive them. In this blog, I’m sharing an excerpt from our newest ebook, “A Recipe for Perfect Product Recommendations,” to provide a taste of the different algorithms you may use on your own site. If you want to read more, you can download the ebook here.
The product recommendations algorithm for popular items simply delivers the most purchased items within a set time frame. This is a great option for new users, since it doesn’t require previous user behavior and highlights a cross-section of your inventory. It can also be used to back fill Personalized Product Recommendations where there is not yet enough data. Popular items also tend to perform well around holidays, where users may not be shopping for themselves.
People Also Like
People Also Like product recommendations are based on a variety of factors that make up the “signature” of a given product. The signature is made up of the most common actions, including searches or clicks, that lead to purchasing an item. Items with the most similar signatures are treated as the most similar items. So, let’s say you have landed on the product detail page for a particular pair of shoes. A People Also Like algorithm will search for items with a similar signature to those shoes and recommend them.
Complementary product recommendations use frequent pattern matching to compare other shoppers’ carts and display the other common items in the cart. For example, say that you have added that pair of shoes to your cart. A Complementary algorithm looks at all the past instances of users having those same shoes in their cart, and delivers the other items most often found in the cart with them. This has huge value for helping customers find an item they forgot, like common recipe ingredients or accessories. Whereas People Also Like product recommendations are based on item similarities, Complementary product recommendations are based on cart similarities.
Personalized product recommendations are based on individual user profiles and informed by their past and in-session behavior on your site. This user behavior is then compared to the product signature of other items. Items whose signature most closely aligns with that user’s behavior will be recommended. This method predicts other items the user is likely to purchase. When a new user visits a page with personalized product recommendations, you can default to filling that module with popular items or most recent items until you have enough data to offer them personalized product recommendations.
For a deeper dive into product recommendations methods check out our new ebook here.
To learn more about SiteSpect, visit our website.