Forecasting of Sporadic Products: Practical and Theoretical Considerations

Abstract

This paper focuses on forecasting of products with sporadic demand. The demand for such products is not continuous but diffused seemingly at random, with a large proportion of zero values in the analyzed time series. The sporadic character of demand patterns actually means that the information available on the demand for previous selling periods is patchy, resulting in lower quality of data available. Under such circumstances demand forecasting is a challenging task. We present the results of a case study, where forecasting practice of a pharmaceutical wholesaler firm –we call it Pharma– is analyzed and developed. We present state-of-the-art knowledge related to demand forecasting of sporadic products and test suggestions related to them. We show that these suggestions can only partly be backed. We extend therefore the suggested product classification scheme and recommend using the concept of demand data aggregation. This will reduce sporadicity and result in higher quality forecasting. Aggregation also helps to specify the recommended forecast period, the length of time recommended to calculate the forecast for. The managerial consequences of these suggestions are also discussed, and future research directions are highlighted.