# 10.6 Summary

A demand forecast is an expression of the probable course of demand along the time axis. A demand must be forecast if the cumulative lead time is longer than the customer tolerance time. Such a situation occurs, for example, in trade in consumer goods, in components for services, or in single parts of investment goods. Forecasts are transformed into demand for resources later and then compared with the organization’s supply capacity. However, every forecast is associated with uncertainty. There­fore, forecasts must be compared to demand continually, e.g., in a rolling manner. A significant deviation in demand may require the selection of a different technique.

We distinguished two basic types of forecasting techniques: historically oriented and future-oriented. Both basic types are further subdivided into mathematical, graphical, or intuitive techniques. The selection of a technique is made according to a series of criteria intended to produce a reasonable alignment of the forecast to the demand, at reasonable expense.

Historically oriented techniques calculate demand based on consumption with the help of mathematical statistics (extrapolation of time series). There are simple techniques for continuous demand, such as moving average or first-order exponential smoothing. For linear trends, we may make use of linear regression or second-order exponential smoothing. In addition, the Trigg and Leach adaptive technique examines and adapts the parameters used in exponential smoothing. All the techniques may be expanded to account for the effect of seasonality. Trend extrapolation, the Delphi method, and scenario forecasts were discussed as future-oriented techniques, although these also contain historically oriented elements.

The more discontinuously consumption occurs, the more difficult it is to fore­cast reliably. The definition of consumption distributions as an overlay of the distribution of consumption events and the distribution of consump­tion quantities per event helps describe discontinuous conditions. A suita­ble length of the statistical period can lead to a smoothing of demands. Where there are few variants and repetitive production, forecast for variant demand of a product family may be calculated using option percentages. This is a stochastic variable with an expected value and standard deviation.

In all cases, larger fluctuations in demand lead to safety demand, which is calculated on the basis of standard deviation. The expected value and standard deviation are related to the statistical period, while indepen­dent demand is related to the planning period. The conversion of expected value is proportional to the ratio of the two time periods, whereas in the standard deviation the conversion is proportional to its square root. The expected value of the demand increased by safety demand is set as independent de­mand per planning period; the latter is then available as stochastic demand for further handling in the context of materials management. When dependent demand is calculated later, using a quasi-deterministic bill of materials explosion, it will contain the corresponding safety demand.

For each independent demand, the item ID, the forecast quantity, and the quantity of the forecast already “consumed” by orders are recorded, as well as the planning date. The total of all independent demands belongs to the production schedule, or, when referring to trade items, the purchase schedule. Independent demand can be recalculated or canceled by rolling planning, either manually or with automated techniques. In general, actual demand successively replaces or reduces independent demand.