As a former academic, I’m always delighted when a topic of academic research begins to prove its worth in a business context. One example of this is process mining, which was introduced around the turn of the century and has already been put to good use in a wide range of businesses, including banking, insurance, logistics, and telecoms. In some recent cases it has notably been deployed to improve supply chain performance.
Process mining is a term used for a series of data science techniques involving the analysis of huge data logs describing what people, teams, organizations and assets are really doing as opposed to what they are thought to be doing. In the telecoms business, for example, process mining is heavily used to improve customer experience in a continuously changing communications landscape.
More involved examples can be seen in the area of supply chains, where processes such as order acceptance, planning, manufacturing, and customers services can be mined. Analysts mine historical production data retrieved from MES and other systems to understand how products are made in practice. This analysis might reveal some surprises, such as the fact that asset productivity is much lower or higher than thought, or that the quantity of scrap generated by a process is significantly higher or lower than predicted. Lead time deviations or alternative production routings might also be identified, for example in the metals industry.
These examples will certainly ring a bell among supply chain planning experts. Productivity and scrap rates, lead times, routings: aren’t we talking about master data here? Yes we are. These are the types of factors that should be defined carefully if you want your digital supply chain planning solution to do a good job and help you optimize your planning. If these data don’t match reality, plans are optimized based on false assumptions, potentially leading to issues such as late delivery, unexpected stockouts, costs that could have been avoided, or inefficient use of assets.
That’s why an increasing number of manufacturing companies are beginning to use process mining as a monitoring tool, allowing them to periodically recalibrate their planning model and the related master data. Even more ambitious is using process mining to derive the planning model from scratch. This data-driven approach to building a digital twin uses machine learning techniques and consequently requires access to massive amounts of relevant production data to achieve reliable results. If these data are available, it can be a great way to kickstart your digital supply chain planning project.
Broes finds his challenge in shaping OMP’s product road map to support the supply chains of the future. He takes a broad approach to supply chain resilience, deploying techniques such as process mining, building models from execution data, and developing optimization and business rules for hyperautomation, as well as stochastics and ML.