In an increasingly fast-paced and volatile world, accurate revenue forecasting is critical to companies in capital-intensive industries. Applying machine learning (ML) provides a more objective alternative to the inevitably biased human judgments commonly used. In this context, this work examines the added benefits of imperfect advance demand information (ADI) through a case study by employing exogenous information and cross-learning to predict different time series simultaneously. Here, training the model jointly on revenue and imperfect ADI data yields the best results. Additionally, clustering the underlying data into homogeneous subsets is most beneficial in terms of cross learning, yet, the margin to the heterogeneously trained models is minimal. In a similar sense, tailoring the model to portfolio products with higher contributions does not yield particular accuracy advantages over a uniformly optimized approach. In summary, this work fosters further research into the previously neglected revenue forecasting field by outlining an end-to-end ML approach.     «

In an increasingly fast-paced and volatile world, accurate revenue forecasting is critical to companies in capital-intensive industries. Applying machine learning (ML) provides a more objective alternative to the inevitably biased human judgments commonly used. In this context, this work examines the added benefits of imperfect advance demand information (ADI) through a case study by employing exogenous information and cross-learning to predict different time series simultaneously. Here, trainin...     »