| I am aggregating events on a user-level and on a session-level so as to gain a higher level understanding of user behavior. Like:
- "What is the median time that users spend on a given page?" - "How many events do users generally fire in their first 3 sessions?" -"Who visited Page A, then Page B, THEN Page C?" -"How many users who used this feature also used this feature within the same day?" This is difficult to do with purely event-level metrics. It also takes a really long time and isn't suited to exploration. We're already using a data warehouse (AWS Redshift) to analyze event-level metrics. But more complex queries are much easier, and much faster to do, when the data is stored in ways that are more suited to such calculations. So, that's what this article is hopefully helping people to do! For a related topic, "Entity-Centric Indexing" is the ElasticSearch equivalent of what I'm working on! |