B2B automobile sales forecasting with participatory predictor using ANN-GA

Purpose The market of a product can generally be divided into business-to-customer (B2C), business-to-business (B2B) and business-to-government (B2G) sectors. Although estimating future sales is important for all sectors, B2B forecasting has not received as much attention as B2C forecasting. The authors argue that B2B forecasting predictors differ from those used in B2C forecasting. This study aims to show what so called participatory predictors to accurately forecast B2B sales. Design/methodology/approach In forecasting, the type of independent variables used and the forecasting method used influence the accuracy of the result. Here, historical sales data are used to forecast B2B future sales using moving average (MA) and simple linear regression (SLR), while a complete set of literature-based independent variables and participatory predictors are used to forecast the sales using multiple linear regression (MLR) and artificial neural network-genetic algorithm (ANN-GA) optimized methods. Findings A case study on B2B automobile sales reveals that participatory predictors enhance forecasting results. The accuracy metric, mean absolute percentage errors (MAPE), is 1.35% using 6 literature-based independent variables and 12 participatory predictors compared to 7.26% using only 6 literature-based independent variables. Furthermore, all forecasting accuracies using the ANN-GA method are better than the MLR, MA and SLR methods. Originality/value This study introduces what so-called participatory predictors for forecasting B2B sales. The predictors are extracted from marketing and B2B-related literature. While the values are retrieved through a survey to the purchasing decision unit (PDU) personnel. To get an accurate forecasting result, the predictors are combined with prominent B2C independent variables selected from the available automotive forecasting literature.