Elevators are the primary method for occupants to move between floors inside modern buildings. Therefore, there have been efforts in minimizing the service delay for a elevator call. Recently, the industry has started to leverage human behavioral patterns in optimizing the elevator dispatching algorithm. However, we argue that it is difficult to judge their gains in the real world, mainly due to the lack of real-world data sets and analysis based on human behavior. We take the first step in studying the human behavioral patterns in the elevator usage. Our analysis is based on real-world traces collected from 12 elevators in a 18-story office building.