Towards User-Oriented Process Mining: A Collaborative Approach to Minimize Late Payments in Accounts Payable Process

This paper presents an application of process mining techniques to a real-world accounts payable process within the context of ECE, aiming to minimize late payments. (a) Situation faced: The initial stages of a process-based insight project posed challenges for the process mining team as they were confronted with numerous process variants and comprehensive dashboards that hindered the identification of critical issues within the processes. These complications impeded the understanding and implementation of process analysis for business users, necessitating extensive training and leading to localized optimizations that were inconsistent with the overarching end-to-end process improvement goals. (b) Action taken: To address these challenges, the team initially concentrated on the accounts payable process to demonstrate the practical implementation of process mining techniques. They developed a process analytics pipeline that integrated advanced process mining with machine learning models, generating targeted and actionable insights and recommendations. Furthermore, they leveraged an informal network of experts to address the difficulties of discerning complex patterns in interconnected processes and delivering root cause insights. (c) Results achieved: The process mining team devised a process analysis pipeline comprising the automatic generation of object-centric event logs that were subsequently analyzed using event knowledge graphs. Machine learning models were then employed to uncover root causes, yielding actionable insights that were systematically organized and correlated to facilitate the identification of areas for improvement and the customization of recommendations for various stakeholders. (d) Lessons learned: This project underscored the significance of a clear purpose, a user-centered design, the targeting of the appropriate audience, effective dashboards, an actionable design, and proper data modeling in the process mining and dashboard designs. The findings suggest that a well-constructed framework can offer tailored insight for various process stakeholders and that selecting a suitable data modeling approach is essential to achieving substantial results from process mining initiatives.