Asymmetrical Evaluation of Forecasting Models Through Fresh Food Product Characteristics

Forecasting accuracy in context of fresh meat products with short shelf life is studied. Main findings are that forecasting accuracy measures (i.e. errors) should penalize deviations differently according to product characteristics, mainly dependent on whether the deviation is large or small, negative or positive. This study proposes a decision-based mean hybrid evaluation which penalize deviations according to type of animal, demand type, product life cycle and product criticality, i.e. shelf life, inventory level and future demand.