Recommendation engine algorithms examine a series of data points regarding the subject matter, including details about the content on the site or service as well as general user habits, to identify similarities between products. The engines then track a user's behavior and suggest items similar to those that she frequently views.
Many websites and services, such as retailer sites or online streaming movie services, use a recommendation engine to suggest new content to users and increase both activity and time spent on the site. These engines begin by cataloging various aspects of the content within the service, known as metadata, to find trends and similarities between the items and make logical connections. For example, a streaming movie service's recommendation engine may make connections between movies made in the same year, featuring the same actors, covering similar topics or that feature the same type of characters.
The engine also looks at user behavior to create its connections, which often results in more complex associations. In this sense, the engine may see that many users tend to watch shows about vampires after watching shows about zombies, and thus determines that these types of programming have a correlation. As an individual user interacts with the service, the engine tracks her usage patterns and uses its knowledge of the connections between products to suggest other items that she may find relevant.