Process mining : construction of an event log and process discovery within a return-order process

Abstract

In recent years, organizations have expressed a rapidly growing interest in improving their end-to-end processes by using the powerful tool of Process Mining, taking advantage of data in order to discover their actual business processes. Currently, poor data quality costs around $3 trillion per year and only 3% of firm’s data meets basic quality standards. Consequently, businesses have acknowledged the potential of utilizing unstructured raw data, transforming it into an event log, thereby enabling improvement of their operational processes.

In context of the Supply Chain Management process of return orders in SAP, this thesis emphasizes on developing a step-by-step guide for the construction of an event log, in order to enable Process Mining and subsequently evaluating the Discovered process model.

Through an analysis of a 2019 SAP-data extraction of a company in the car parts business, this study develops a six-step guide towards a complete event log aimed at visualization and analysis of the return-process of sales orders. The analysis describes an approach to identifying and separating process instances, order events and construct timestamps, extract activities, in addition to extracting and enriching event data to form the event log. Process analysis in the form of Process Discovery is made possible by utilizing the steps of the developed guide. Furthermore, the quality of the resulting process model including the representative behavior seen in the event log is evaluated by applying a four-dimensional framework. The dimensions Replay Fitness, Simplicity and Precision is characterized with a plus-symbol (+), whereas the dimension of Generalization is characterized by a minus-symbol (-).

The approach to construct activities in the event log is highlighted as a likely root cause of the process model’s low score on Generalization. Furthermore, the current method of evaluating the quality of process models is considered to be lacking proper scaling capabilities, and further research on the topic is advised. After the evaluation, a supplementary case study utilizes the step-by-step guide on the extracted SAP-data, in order to illustrate the possible business insights that Process Mining may extract from the constructed return-order event log.

In closing, the thesis sums up the step-by-step guide, subsequently concluding that all steps are considered essential and that the resulting process model is of medium-plus (+) quality.

De siste årene har sett en markant økning i etterspørselen etter å utnytte Process Mining til å forbedre eksisterende forretningsprosesser. Process Mining muliggjør at selskaper kan unytte store mengder data til å analysere de facto modeller, ved å konstruere høy-ytelses hendelseslogger. Totalt, er det estimert at lav datakvalitet medfører årlige kostnader på 3 billioner dollar, der 3% av selskapenes data er av holdbar kvalitet. Dette kan ha en årsakssammenheng med selskapers økte fokus på datadrevne kvalitetsforbedringer for å ta i bruk Process Mining. Målet med denne gradsoppgaven er å utvikle en steg-for-steg guide for konstruksjon av en hendelseslogg, i konteksten retur-ordre i ERP-systemet SAP, for dermed å muliggjøre Process Mining. Deretter skal prosess-modellen evalueres for å avgjøre ytelsen.