Intermittent Demand Forecasting

11Classification Methods

11.1 Introduction

Demand forecasting sometimes needs to be performed at a group level for a collection of different items or, indeed, across all the items that are held in stock. This would be the case, for example, when deciding on the mode of transport to distribute products to customers or when investing in a new warehouse, the capacity of which needs to be determined. In such cases, a manager would be interested in the future demand of more than one single item, and the forecasts would typically inform medium‐ and long‐term planning.

As the forecast horizon gets shorter, typically the level of aggregation (the number of items for which an aggregated forecast is required) gets smaller. For inventory management, short‐term ordering decisions are at the lowest level of aggregation, namely the individual stock keeping unit (SKU). This necessitates the selection of an appropriate forecasting method for each item. However, a typical organisation holds hundreds or thousands, or in the case of a military entity, like the US Defense Logistics Agency for example, even millions of distinct inventory items.

To streamline the organisation and management of such large collections of items, businesses often utilise classification schemes. Under such a scheme, the items are separated into various classes and common decision processes apply to all items belonging to a particular class or category. (The words ‘class’ and ‘category’, as well as ‘classification’ and ...