This document discusses using MongoDB for flexible event logging and analysis of funnels, retention, and viral spread. It provides examples of aggregating web and video system event data, processing it using Python map-reduce jobs, and counting events in real-time and over time. The results can be queried to analyze user behavior and the effectiveness of changes. While initial performance of complex map-reduce jobs was poor, MongoDB 1.5.0 improvements are expected to make this approach fast enough for both batch and interactive use.