Metalloinvest Introduces Machine Learning in Procurement Process
Russian mining and metallurgy company Metalloinvest has successfully implemented delivery disruption forecasting based on machine learning into
Russian mining and metallurgy company Metalloinvest has successfully implemented delivery disruption forecasting based on machine learning into supply chain workflows. The innovative solution has increased the efficiency of the work of specialists, while reducing the risk of additional costs associated with deviations in delivery times. The forecast accuracy is 84.7%. Due to the accumulation of data, the introduction of new parameters and the retraining of the system, it is planned to bring this figure to 87% by the end of the year. Procurement Director of Metalloinvest Maria Kovalenko said “Buyers have an average of 150,000 bids from four of Metalloinvest's plants in operation at a time. Of this number, all those related to the risk zone were selected and taken under special control. According to them, a point-by-point development of all identified issues is carried out. In the near future, it is planned to launch an automated distribution of notifications to suppliers that fall into the risk zone. This approach makes it possible to increase the provision of demand without increasing the burden on employees."
To train the system, we used the data of more than 200 thousand already executed orders stored in SAP. Based on this "digital footprint", a forecasting model has been built. To assess at different stages of its life cycle, up to 27 parameters are used, including the volatility of exchange rates, supplier country, mode of transport, seasonality, urgency, and so on. Over time, the list of parameters will expand.
OTIF (on time in full - fulfilment of obligations on time and in full) is one of the main indicators of the efficiency of the supply service. It can be maintained at a high level either through the control of all active applications, which requires significant human resources, or through selection from the total volume and elaboration of applications with high risks of failure. Working according to the second scenario saves resources, but requires a high quality forecast. The project was developed and implemented by the team of the Department of Methodological Support and Supply Development of Metalloinvest under the leadership of Roman Orlov with technical support from JSA Group (part of the ICS Holding diversified IT group) and Metallo-Tech. This allows developing and accumulating competencies within the Company and transferring successful experience to related areas.