Improving forecast accuracy for seasonal products in FMCG industry: integration of SARIMA and regression model

Increasing forecast accuracy of seasonal products is very critical as production, inventory and customer service depends on it. There has been introduction of new models, techniques and use of advance data analytics in forecasting, however, considering the complexity of the several causal variables and demand, it has been very difficult to get the consistent accuracy. This paper proposes integrated SARIMA (for non-seasonal component of demand) and regression (for seasonal component of demand) models for improving the forecasting accuracy. Further, we evaluate the performance of the proposed model with other known methods such as, SARIMA, ANN and SARIMAX. The performance is evaluated on various parameters of forecasting error. It is seen that for the empirical data, the proposed method outranks the other methods on all the performance metrics. Further, this paper brings into managerial insights, which can be replicated to various industries, indicating the wide scope of the proposed approach.