User traces are essential for analysis of human behavior and development of opportunistic networking protocols and applications. As user traces are collected with high granularity to apply them in diverse scenarios, they have a high complexity resulting from the large number of user states. We present MobReduce: a methodology for reducing the number of states in user traces. We apply MobReduce to individually to GPS locations and WiFi sightings of the Nokia Mobile Data Challenge data set and show how to trade off state complexity vs. granularity.