We propose a methodology applied to complex systems to analyze operational risk events in banks, with the objective of determining an understanding of the key characteristics and their relationships in initiating operational risk losses. We applied our methodology to operational risk losses in Australian banks over the period 2010-14. The analysis identified that there are a small number of characteristics that are common to many operational risk events, and these "level 1" characteristics are stable across time, which implies operational risk losses could be controlled by managing these characteristics. The methodology adds value to the existing analysis by identifying the main characteristics of operational risk events in a rigorous manner.