Dominik Hammann

2020

At present, controlling departments often rely on simple methods to predict their budgets, which
often leads to large deviations from actual values. The spread of business intelligence and analytics along with
machine learning offers the opportunity to improve forecasting accuracy and simplify the budget
forecasting process. What is missing nowadays, however, is proof of the implementation of such
methods and their applicability in practice. Hence, this thesis examines the role of predictive
analytics and especially machine learning for the prediction of R&D budgets. The research in form
of a case study takes place in the R&D controlling department of a large German utility vehicle
manufacturer. I find that predictive analytics can help to improve the forecast quality and machine
learning algorithms do not outperform parametric algorithms in general. Furthermore, employees have
a positive, yet cautious attitude towards the application of predictive models and pay most
attention to accuracy, integration of feedback, and acceptance by top level management.