TY - JOUR
T1 - Forecast value added in demand planning
AU - Fildes, Robert
AU - Goodwin, Paul
AU - Baets, Shari De
PY - 2024/12/26
Y1 - 2024/12/26
N2 - ‘Forecast value added’ (FVA) is a term commonly used to measure the improved accuracy achieved by judgmentally modifying a set of forecasts produced by statistical methods or algorithms. Assessing the factors that prompt such adjustments, and when they are likely to improve accuracy, is important in company demand forecasting and planning but has not been studied sufficiently. The published research has taken various individualistic approaches, both in the questions examined and the data analysis and modelling. In this paper we have collected the publicly available data from these studies, six in total, to analyse them using a common framework. Questions include when do demand planners adjust their statistical forecasts, do adjustments improve accuracy and reduce any bias, does the size of the adjustment signal a more substantive and useful piece of information gathered by the demand planner, and are improvements consistent across companies? These questions are important in practice since the costs of error are substantial, while the process of adjustment is expensive and time consuming, but they are also theoretically interesting raising the question of why consistencies across companies arise and the circumstances when one organization is more effective than another. The key question is how organizations can improve on their current forecasting processes to achieve greater ‘forecast value added’.
AB - ‘Forecast value added’ (FVA) is a term commonly used to measure the improved accuracy achieved by judgmentally modifying a set of forecasts produced by statistical methods or algorithms. Assessing the factors that prompt such adjustments, and when they are likely to improve accuracy, is important in company demand forecasting and planning but has not been studied sufficiently. The published research has taken various individualistic approaches, both in the questions examined and the data analysis and modelling. In this paper we have collected the publicly available data from these studies, six in total, to analyse them using a common framework. Questions include when do demand planners adjust their statistical forecasts, do adjustments improve accuracy and reduce any bias, does the size of the adjustment signal a more substantive and useful piece of information gathered by the demand planner, and are improvements consistent across companies? These questions are important in practice since the costs of error are substantial, while the process of adjustment is expensive and time consuming, but they are also theoretically interesting raising the question of why consistencies across companies arise and the circumstances when one organization is more effective than another. The key question is how organizations can improve on their current forecasting processes to achieve greater ‘forecast value added’.
U2 - 10.1016/j.ijforecast.2024.07.006
DO - 10.1016/j.ijforecast.2024.07.006
M3 - Article
SN - 0169-2070
JO - International Journal of Forecasting
JF - International Journal of Forecasting
ER -