Avoiding Supply Chain Risk with Machine Learning


Almost all Machine Learning techniques use past data in order to predict the future. Although this can be effective however, it is not a true representation of the future. A supply chain model or what we call a supply chain Digital Mirror© represents very accurately how the operation would behave in the future. Note that for this to happen we need to have a true mirror of the supply chain, not just a high level S&OP model that is most probably 60% accurate! A Digital Mirror© requires accurate modeling of the operations, equipment, people and execution capability of the supply chain (what Gartner refers to as S&OE model). Now given that we have a digital model of the supply chain, it is possible that with every run of the model we identify issues causing projected lateness or projected increase in cost. Over time we can then use Machine Learning techniques to identify the patterns that are evolving. For example, certain materials are the main causes of lateness during certain periods of the year for 10% of customers for a given group of products. Such underlying trends are always on-going. The fact that we don’t see them does not imply that they are not significant. By identifying such patterns we are in a position to avoid supply chain risks and be proactive in understanding how the supply chain is changing behavior over time. You can read more on this topic and how we are deploying Attribute-Based architecture to identify causes of undesired issues at Innovations on our website.




Leave a comment

Request Demo