About Recommendations

There are two methods for delivering recommendations to users. User-centric recommendations take a user’s behavior into account. Product-centric recommendations are based solely on a product.

When you select the user-centric method, it can be helpful to remember the following information:

• This method finds similarities in behavior or characteristics between users and uses that information to make recommendations.
• New users do not receive recommendations during the first day that they visit a website because there is not enough information about that user available to generate recommendations. However, information about the user is tracked right away and is used to create recommendations for future visits.
• It takes one day for a product view to be eligible to be used as a basis for a recommendation. For example, if a user views a product on Wednesday, the recommendations shown to the user do not take that product view into account until Thursday.

When you select the product-centric method, it can be helpful to remember the following information:

• Product recommendations are made based on associations between products, without taking into account user behavior or characteristics. This is helpful when you do not know anything about the user except that they are interested in a product during the current website visit. For example, when a user is looking at a bicycle on your site, you can recommend other products associated with the bicycle, such as bicycle helmets or bicycle baskets.
• Analytics uses product views as a proxy for the ratings that are traditionally used in recommendation systems. The more that products are viewed by similar users, the stronger the association becomes between the products.

Data for recommendation tasks is stored in these tables: IMPRESSION_DELIVERED, IMPRESSION_SPOT_VIEWABLE, SPOT_CLICKED.