Demand Forecasting Method of Moving Average considering Variable Selection by Multiple Comparison Procedure

To help achieve Goal 12 of the SDGs, “Ensure sustainable consumption and production patterns,” food loss must be reduced. Therefore, a case study based on industry-academia collaboration with a processed food manufacturer was conducted. The manufacturer has a single production site in Koga City, Ibaraki Prefecture, and it ships products nationwide. The company handles milk, dairy products, various beverages, desserts, and other daily delivery products. All of these products are susceptible to obsolescence and have a short shelf life; consequently, excess inventory can easily lead directly to food loss. Aiming to reduce food loss, we analyzed the current situation and factors according to a problem-solving QC (Quality Control) story, and low accuracy in demand forecasting was found to be the main cause of food loss. As a result of studying countermeasures from the perspective of the 4Ms (Man, Machine, Material, Method) of production factors in quality control, we developed and introduced a demand forecasting method suitable for the characteristics of day-of-week-dependent demand as a countermeasure. By examining the formula for demand forecasting, we proposed a method that applies the multiple comparison procedure used in pharmaceutical development to demand forecasting in this case study. The demand forecasting formula is simple, and its practical use is emphasized in this case study. We compared and verified the accuracy of demand forecasting between the proposed method and existing methods using the absolute percentage error as an evaluation index and confirmed the superiority of the proposed method. This proposed method has practical value not only in terms of demand forecasting accuracy but also in terms of the standardization of operations.