Abstract
This paper investigates the phenomena and effects of heavy tails of execution time in business processes. We approached the heavy tails as a particular variation of business process execution, and analysed them from empirical data of its execution time. The features and possible causes of the heavy tails were investigated. A workflow simulation was conducted to simulate the heavy tails, in order to evaluate quantitatively their effects to business process performance. The simulation outcomes demonstrate that (1) even a heavy-tailed activity in a business process can cause large variation of end-to-end execution times of the process and (2) there are significant prediction gaps resulted by different assumptions of execution times between an exponential distribution and a heavy-tailed distribution.