Preliminary analysis of an adaptive smoothing factor for lumpy demand forecasting

Forecasting has become increasingly complex due to high and variable customer demand of a wide array of products. Over the last years the demand has become increasingly erratic due to some structural changes in the market. These structural changes are characterized by numerousness of potential customers in the market, heterogeneity of customers, frequency of requests, variety of customer\´s requests, and correlation between customer\´s requests. These changes generated an extremely irregular demand, which is also referred to as "lumpy". This demand is not only variable but also "nervous", thus entailing great differences between successive demand observations and sporadic demand peaks that follow several periods of zero or low demand. In this paper, the selection of a smoothing factor in a structured and adaptive manner for lumpy demand environments is investigated. A smoothing factor based on the Kalman Filter weighting function is developed and tested