Marketing predictions under macroeconomic volatility: empirical evidence for automotive SMEs from a machine learning perspective

Abstract Small and medium-sized enterprises (SMEs) in the automotive industry rely critically on predictions of future demand for new cars to inform downstream marketing activities – particularly in current times of macroeconomic volatility. This empirical study analyzes 21 quantitative prediction methods within the German automotive market, encompassing benchmark/rule-based, econometric/statistical, single-method Machine Learning (ML), and ensemble methods with ML integration. The findings indicate that in predictive performance, methods integrating ML significantly surpass those that do not. Furthermore, multivariate methods – based on predictors selected from 30 economic, demographic, and industry indicators – significantly outperform univariate methods. While a non-linear, single ML method achieves the highest accuracy, the study also provides recommendations tailored to the specific priorities of SMEs. For example, a substantial “Forecast Value Added” is also provided by the method AutoGluon, which automatically ensembles multiple ML algorithms and may yield additional value to SMEs with lower technical barriers. Beyond point prediction, the paper exemplarily demonstrates how SMEs can use scenario-based analyses to estimate demand-side impacts of macroeconomic shifts. In this way, the study’s results offer guidance for automotive SMEs to manage the recent volatile period by benefiting from market-oriented predictions.