Process mining uses data from IT systems to identify and analyze the underlying business processes. The production of oat or soy drinks is a multi-layered process. The natural raw materials are processed step by step via large tanks and long pipes into 1L cartons. Berief wants to use its extensive data pool to further optimize its processes and uncover evidence-based potential.

How do you lift the treasure trove of data?

 Food products must meet the highest quality standards. At the same time, raw material costs are increasing, consumers need to save and plant-based foods are in greater demand - attracting international competitors.  

Berief Food GmbH is meeting these requirements and challenges by digitizing its processes. In order to convert this treasure trove of data into evidence-based insights about internal processes by means of process mining, three challenges must be met. First, machine data must be converted into usable process data. Second, process mining algorithms work with a unique process identification number - for a process that ultimately merges a thousand liter batches into 1L cartons this identification capability must be worked out in a structured manner. Third, process mining must be used efficiently and purposefully in the enterprise environment to achieve the greatest possible effect.  

 The overall goal of the transfer pilot is the use of process mining to identify optimization potentials in production processes. In parallel to the challenges, three questions have to be answered for this: 

1. How can the different data pots be merged? 

1. What analysis perspectives result from the batch trees? 

1. How can process mining be integrated into the digital "value chain" in the long term? 

For this purpose, various solution approaches from previous research projects such as BPM-I4.0 are used. These include data maps, key figure systems and, of course, visualizations.  

Light in the dark of the silo

Since the transfer pilot is still running, only partial results can be presented. Early in the project, level changes and machine data could be cleverly combined to represent a process model about the typical behavior of a tank (see figure). These results will be used in the further course of the project to map the entire process, develop a dashboard and answer key business questions.