Improving automotive battery sales forecast

Abstract

Improvement in sales forecasting allows firms not only to respond quickly to customers' needs but also to reduce inventory costs, ultimately increasing their profits. Sales forecasts have been studied extensively to improve their accuracy in many different fields. However, for automotive batteries, it is very difficult to develop a highly accurate forecast model because many variables need to be considered and their correlations are complex. Additionally, current sales forecasts are derived from historical data and thus do not include any other causal factor analysis. In this study we applied causal factor analysis to determine how the forecast accuracy could be improved. We focused on understanding the relationship between temperature and sales. Using regression modelling, we found that there is a quadratic relationship between temperature and battery sales. We validated the model by comparing the actual and predicted sales for various geographies and times. We concluded that the model is more robust for predicting sales across various times than through various geographies.