MINING E-MAIL CONVERSATIONS TO ENRICH EVENT LOGS: AN EXPLORATORY CASE STUDY OF A HIRING PROCESS IN A NORWEGIAN MUNICIPALITY

Authors

  • Aina Goday-Verdaguer
  • Felix Mannhardt
  • Hans Yngvar Torvatn

Abstract

Process improvement is an important challenge for the public sector, which struggles with reduced budgets and raised expectations on service quality. Process mining is a set of techniques that use data from information systems to understand how business processes were actually performed. It has been used widely to help with the process improvement challenge by identifying improvement and automation potential based on facts. We analysed the hiring process at Røros Kommune, a Norwegian municipality which hires about 150 new employees each year, by using interview-based process mapping as well as process mining. Our goal was to identify whether there is potential for automation as well as to explore the use of recorded event data for process analysis, which was not yet done beforehand. We used both structured data extracted from two existing information systems as well as unstructured data from e- mail communication between employees. The hiring process is comprehensive consisting of many steps, involving many responsibilities, and making use of several information systems. Many manual activities give raise to a large potential for mistakes. We found that data from the information systems was not always reliable and, therefore, developed a novel data correlation tool for connecting relevant e-mails to process activities. The results show the complexity of the hiring process as well as lacking support for several parts of the process through the used information systems. Including the e-mail communication showed promising results to improve understanding; however, many challenges remain given the complex and ambiguous relation between e-mails and process activities. Our case is unlikely to be unique, for instance a lot of Norwegian municipalities use the same information systems for similar tasks, and the low data quality can likely be replicated there. Thus, any automation or artificial intelligence project likely requires much work on improving data quality.

Published

2020-11-23 — Updated on 2020-12-01

Versions