Measuring Complexity of Bot Models in Robotic Process Automation

Robotic process automation (RPA), a technology to automate structured tasks on computers in a light-weight fashion, thrives on graphical models, enabling business users and citizen developers to create automation workflows without programming knowledge. However, due to the created flows often not being optimized and the fine-grained nature of the RPA instructions, these models quickly become complex and extensive, complicating their maintenance and comprehensibility. In this paper, we draw inspiration from the related fields of software programming and business process modeling to introduce complexity metrics for RPA bot models. These allow to objectively measure and compare the complexity of such workflows, and can thus, for example, provide an indication of where to start refactoring models in the bot repository.