Intermittent Demand Forecasting

6Forecasting Mean Demand

6.1 Introduction

Identifying a statistical distribution that may adequately represent intermittent demand data, as discussed in the previous chapters, is an important first step for successfully managing inventories. Although it may be reasonable to assume, at least in the short term, that the form of the distribution does not change over time, it would be less reasonable to expect that the mean and variance of the distribution also remain unchanged over time. In practice, these parameters of the demand distribution are not known, and forecasting is concerned with their ongoing estimation. Certain distributions, like the Poisson for example, call for the estimation of one parameter only, namely the mean. For other distributions, such as the negative binomial and the normal, the estimation of the mean demand is not sufficient. Although forecasting the mean demand is often emphasised, estimating the variance is also important for most distributions. In this chapter, we focus on the issue of forecasting mean demand, leaving the forecasting of variance to be covered in Chapter 7.

One crucial issue involved in forecasting of mean demand is the determination of which estimates are relevant to the inventory decisions to be taken. Suppose that the decision is whether to continue stocking the item, or to discontinue re‐ordering. In this case, as discussed in Chapter 2, forecasts generally will be made for a whole collection of stock keeping units (SKUs). Whether ...