Our Content Recommendations endpoint is designed to be used without user data. The only required input to the API is a desired “interest” that you want to find recommendations for. For example, a recommendation request based on ‘LaMelo Ball’ might return:
Which are all relevant to LaMelo Ball through things like position, style of play, rookie expectations, and other athlete attributes. With this list of recommendations, you can then determine which interests a user isn’t already subscribed to and push content involving these interests to further fuel the user’s discovery and engagement.
Our User Recommendations endpoint takes the above a step further and ingests your interest catalog along with your users’ preference and consumption data to crowdsource how a single user stacks up to the crowd, and ultimately what hyper-personalized recommendations they’re most likely to interact with. For example, you may provide a payload of NBA interests you serve on your platform:
along with a payload of your users and their defined interest or consumption data:
From here, our algorithms will do the math and learning to determine which content and interests you should push to each user based on the provided information. The output will look similar to the above input, except each list of returned interests will be recommendations specific to each individual user id provided:
With this user-level granularity, you can simply suggest a user follow these interests, or go so far as to passively curate and surface content around these hyper-personalized interests to drive more topic discovery.