4 Process selection for RPA projects

Abstract

The identification and selection of suitable processes are an essential success factor for robotic process automation (RPA) projects. Numerous studies show that 80 % of RPA projects fail due to wrong decisions in this phase. Insufficient decision quality often results from inaccurate qualitative analysis methods based on interviews, observations, and estimates. Several research frameworks have recently been developed to address this problem. They postulate to use process mining to collect accurate and robust information that can be used for decision support. However, process mining is based on analyzing event logs, which are only available in sufficient quality and information width and depth in process-oriented information systems. Activities performed in these systems are typically not suitable for RPA, as they can be automated directly via system-specific workflows. To address this shortcoming, we introduce a novel approach that combines desktop activity mining and process mining techniques in the present chapter. We use desktop activity mining to record and analyze all user interactions during the process execution, such as clicks and keystrokes. To comply with privacy regulations, we limit recording to a short period of time. Furthermore, we extract process execution data from ERP systems, merge it with our data set, and extrapolate previous findings to obtain a more holistic understanding of a process's suitability for RPA. We conceptualize our approach in a five-step iterative framework and demonstrate its practical implications by applying it to experimental data.