This document discusses building a user data store and analytics platform for an ad system. It outlines storing user profile data and activity data in a graph database to allow querying relationships between users, sites visited, ads served, devices used, and locations. The challenges of querying huge graphs with "super nodes" with millions of connections are discussed. The document proposes using a directed multi-property graph model with bitmap indexing to optimize expensive queries by avoiding costly graph traversals. It provides examples of how the bitmap indexes would allow set operations to efficiently retrieve user subsets and analyze user behavior patterns over time.