Forecasting Sproadic Demand In Supply Chain Management

With increasing number of SKUs(stock keeping units) in the supply chain, demand data for many products become more sporadic with few nonzero observations. Almost binary-like pattern of the demand data makes the forecasting difficult. There are several established methods such as Croston’s exponential smoothing or SBA with limited general success. Most of the time forecasting of a sporadic demand requires modification of a standard method to incorporate some domain specific information into the forecasting model. In this research we present a demand forecasting approach for a local candy manufacturer. The company has a portfolio of products with highly sporadic demand. Our approach includes standard methods and an alternative model that is based on the dynamic aggregation of the demand data.