Towards discovering erratic behavior in robotic process automation with statistical process control

Abstract Companies that frequently use robotic process automation often encounter difficulties in maintaining their RPA portfolio. To address these problems and reduce time spent investigating erratic behavior of RPA bots, developers can benefit from exploring methods from process sciences and applying them to RPA. After a selection process, we examine how variability and deviations impact robotic process automation. Indicators of statistical dispersion are chosen to assess variability and analyze RPA bot behavior. We evaluate the performance of RPA bots on 12 processes, using statistical dispersion as a measure. The results provide evidence that variability is an undesirable form of erratic behavior in RPA, as it strongly correlates with the success rate of the bots. Importantly, the results also show that outliers do not affect the success rate of RPA bots. This research suggests that variable analysis can help describe the behavior of RPA bots and assist developers in addressing erratic behavior. Additionally, by detecting variability, we can more effectively handle exceptions in RPA.