Core Aspects of Self-Driving Supply ChainBy CIOAdvisor Apac | Friday, December 14, 2018
The age-old questions of resolving problems in real-time and building a self-driving supply chain are still prevalent. The other concern is utilizing machine learning to ensure the continual betterment of the system.
The demand planning application is suited for machine learning because the forecast accuracy is precise. If a manufacturer is facing a shortage of crucial component, then they have a series of calculated decisions to make. If they have already promised the customers a certain number of units, then they might run the risk of dissatisfying a few of their other customers. The supply chain planning engine must re-work to generate a new supply chain. Few suppliers are using pattern recognition to resolve the problem as it was addressed in the past. If demand plans are made on a daily basis, and every day a new data set is available, then the supply chain system can improve quickly.
Demand planning is a good application for machine learning because it has a continual flow of data to improve the system. Machine learning observes the forecast accuracy from the model and raises a question if anything was altered and if the forecast would improve in any manner by the change. Employing machine learning in supply planning is difficult. Demand management application constantly monitors forecast accuracy, and the accurate data allows learning feedback loop. Demand planners, people that use the outputs of the system, are crucial because they maintain uniformity and accuracy of the data inputs.
Data flows from various systems in supply planning. The biggest problem of data management is scattered data among different incompatible systems, processes, and formats. The other major issue is the people responsible for data accuracy do not use the system outputs to maintain data accuracy. A business intelligence/middleware must collect, analyze, and visualize the data and automatically alert a human if the data is altered. Bringing together middleware, supply chain applications with optimization and predictive analytics, and decision rules for prescriptive analytics can create a feedback loop capable of taking advantage of machine learning. Its implementation will make the system better over time.
An organization cannot solve problems in real-time with optimization. Companies would need to make a set of rules to make tough trade-off decisions. Startups have designed a workflow engine that helps them create these rules. Users get prescriptive suggestions and they can either say yes or no to the suggestion. Every decision becomes historical data that provides a feedback loop that allows the machine to get smarter over time and that the point where the system becomes a self-driving supply chain.