Identifying Leading Indicators for Tactical Truck Parts’ Sales Predictions Using LASSO

This paper aimed to identify leading indicators for a case company that supplies truck parts to the European truck aftersales market. We used LASSO to extract relevant information from a collected pool of business, economic, and market indicators. We propose the efficient one-standard error rule, as an alternative to the default one-standard error rule, to reduce the influence of sampling variation on the LASSO tuning parameter value. We found that applying the efficient one-standard error rule over the default one, improved forecasting performance with an average of 0.73%. Next to that, we found that for our case study, applying forecast combination yielded the best forecasting performance, outperforming all other considered models, with an average improvement of 2.38%. Thus, including leading context information did lead to more accurate parts sales predictions for the case company. Also, due to the transparency of LASSO, using LASSO provided business intelligence about relevant predictors and lead effects. Finally, from a pool of 34 indicators, 7 indicators appeared to have clear lead effects for the case company.