An Integrated Lean Service and Machine Learning Approach to Improve On-Time Delivery in a SME Restaurant

Abstract

Food service establishments operate in environments where demand variability, ingredient availability, and delivery speed directly impact operational performance. In this context, a diagnostic assessment conducted in a pizza restaurant revealed many important inefficiencies such as inventory inaccuracy because of insufficient supplies; inefficient mise in place due to long time to find supplies; and unproductive times caused by inadequate control in the process and poorly optimized layout. To address these main issues, this study implements an integrated model that uses a Random Forest algorithm to forecast weekly demand with high accuracy and synchronizes purchasing and production using an EOQ-MRP system with Kanban as a visual control mechanism, in addition of standard work and station redesign to reduce preparation time. The proposed model generates highly accurate demand predictions, enabling more efficient inventory planning and improved coordination between purchasing and preparation activities. Hybrid validation shows that process compliance increases from 79% to 97% after implementation. These findings show indeed that combining Machine Learning forecasting with Lean Service tools significantly improves synchronization between all the processes and stabilizes ingredient supply, enhancing operational performance in foodservice environments. These results open the path for future research integrating predictive analytics with Lean Service methodologies to address operational variability in dynamic foodservice settings.