This documents all the events/data coming in from clients to be used the recommendation system, analytics, etc.
Client sends events to the backend. The events are actions taken by the user that contain certain information about the user's action that is useful for us to understand the user's usage of the the feed, and help optimize it. This information is essentially datapoints. These datapoints are all listed below, organized by similarity into tables.
We consume data from the client in a raw JSON format which is highly flexible and extensible. We do this instead of consuming perfectly formatted data with an enforced schema because it allows to add/remove/extend/reformat what datapoints we want to consume on the fly with very little overhead.
The goal is to get as much data as possible from the user about the interaction with the feed and the app in general, and then process it as needed for downstream use.
These tables describe this data:
That raw data is organized, parsed, joined and transformed into a more efficient usable format by an ETL process for the Recommendation Engine/other services to consume.
Overview