Cash flow forecasts play a pivotal role in corporate financial management tasks. The forecast are for instance used for liquidity management to ensure solvency or for currency risk management to identify and hedge exposures resulting from foreign business activities. Inaccurate forecast can be an unreliable basis for corporate-wide financials plans and lead to liquidity shortages, uncovered currency risks or increased hedging costs. Hence, the quality of corporate financial management tasks therefore strongly depends on the accuracy of available cash flow forecast data.
In this project we develop novel measures for assessing the quality of cash flow forecasting processes, we develop techniques to identify and automatically correct biased forecasting processes, we explore forecast blending techniques to appropriately "mix" forecasts generated by experts and model-driven forecasts generated by statistical prediction mechanisms, and explore the structure of forecast- and revisioning processes to identify processes with weak efficiency that can and should be improved. The artifacts developed in this project are learned, adjusted, and evaluated using a database of millions of real-world cash flow forecasts with different forecast horizons and corresponding actual items provided by hundreds of subsidiaries from different regions and business divisions over a period of multiple years.