Development of the Multi-factor Model for Demand Forecasting for a FMCG Company Production

The paper is devoted to the development of a multi-factor model for solving the problem of low forecasting accuracy of demand on products in FMCG-sector. The aim of the study was the development of a demand-forecasting model for a large FMCG company. To achieve it, a list of tasks was solved: • A sector of FMCG-market was analyzed and the key features of demand on it were identified; • Approaches to demand forecasting in FMCG were compared; • The task to develop a predictive model for company XYZ was set; • The current process of forecasting demand on products in XYZ was analyzed; • The requirements for the forecast model (assumptions, limitations, quality metrics) were established; • The data research and preparation for modeling was made; • The choice of forecasting methods was justified; • A multifactor model that meets the stated requirements was developed; • An analysis of business effects was carried out and recommendations for the application and further development of the model were formulated. The result of the research is a developed hybrid multifactor model combining cluster and regression analysis methods, as well as a corrective prediction unit based on a neural network. The model is implemented using tools based on Python and IBM Watson Analytics. The sequential application of the methods proposed by the author with the cluster analysis of the initial variables that preceded them allowed determining the aggregate volume of demand for the company's products in monetary terms with a test sample with an accuracy of 97%. The resulting solution fully satisfies the requirements formulated in the paper. The practical significance of the obtained results lies in the fact that the constructed multifactorial hybrid model was tested in the company under study and, taking into account the described limitations and requirements, can be successfully applied in the business practice of companies in the domestic FMCG market.