Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for clustering user interface event data for analysis and retrieval are disclosed. In one aspect, a system includes a data store and computer(s) that interact with the data store and execute instructions that cause the computer(s) to receive, for a user interface event, event data specifying a structure of a user interface presented during the user session. The event is assigned to a respective cluster based on a comparison of the structure of the user interface specified by the event data to a user interface structure that represents the respective cluster. For each cluster, a user interface attribute indicative of a user interface state of user interfaces specified by the event data in the cluster is determined. User interface state groups are generated based on the user interface attribute for each cluster.