Sales Forecasting System for Van-Sales Channel for FMCG Industry

Abstract

In the Fast Moving Consumer Goods (FMCG) sector, the availability of sufficient product inventory on the delivery vehicle is directly related to the accuracy of the sales forecasts. Insufficient accuracy of the estimations leads to loss of income and increases secondary costs such as transportation and labor costs. In the current situation, sales forecasts are based on the sales personnel's delivery route, knowledge, experience, and relationships. Since the knowledge and experience of the personnel are not brought into the institutional memory, this information is lost with the personnel change, and the new person needs to develop their own experiences about the route. Currently, the sales forecasting accuracy rate is calculated as 70%. It has been determined that a daily loss of 15% on a product basis and a total of 5% daily occurs. In the study carried out within the scope of this research, advanced analytical and machine learning methods that can capture the dynamics of the FMCG industry and analyze the extensive data formed effectively are studied to increase the accuracy and consistency of sales forecasts. Within the scope of the research, machine learning models to be used for sales forecasts were developed using artificial neural networks methods. We evaluated the models' performance according to the recall, precision, and accuracy metrics based on the route, point of sale, and product. It was determined that the artificial neural networks performs well for sales forecasting. Using the artificial neural networks in the experimental study, we achieved an average of 5% revenue increase for the three route groups selected as pilots. The sales forecast accuracy rate increased from 78% to 82%.