Objective methods of forecasting are statistical methods that range in complexity from relatively simple trend extrapolations to the use of sophisticated mathematical models. More and more companies are tending toward the use of advanced methods in which the computer correlates a host of relationships.
Trend analysis via extrapolation: A simple objective method of forecasting is the extrapolation of past sales trends. In this method the assumption is made that sales for the coming time period will be equal to the current level or that sales will change to the same degree that sales changed from the prior period to the current period. Such simple predictive models are more reliable than might at first be thought – especially for very short time periods (a month or a quarter) under stable conditions. This forecasting method assumes that some past pattern in sales can be identified and measured, and that if reflects accurately what will happen in the coming period. Thus, the forecasting task centers on quantifying the trend or tendency in such a way as to project it into the future. For example, using historical data on total births per 1,000 and current population statistics, it is possible to forecast the number of births in the coming year. In undertaking any kind of trend analysis, the researcher must keep in mind that each time series is made up of four factors: long term trend, cyclical variations, seasonal variations, and irregular variations. If the pattern of these factors is at well developed, each of them can be separated from the other. The first three (trend, cyclical and seasonal) can then be projected to determine the sales pattern for the future.
Regression analysis: Regression analysis can be used in sales forecasting to measure the relationship between a company’s sales and other economic series. For example, automobile manufacturers may find their sales are related to personal income – when income go up, their car sales go up and when incomes go down, their sales drop. To use this relationship in forecasting car sales, the manufacturers must determine the degree of relationship. If income rises 10 percent, do car sales rise 10 percent, 30 percent, 2 percent or what? Regression techniques enable the producers to estimate the relationship between changes in income and changes in car sales.
One may wonder how the discovery of a relationship between sales and one or several other factors helps to forecast sales. The problem is merely shifted from forecasting sales to forecasting the other factors. However, this indirect approach has two advantages. First, a number of other factors, such as general economic series and personal income, are forecast by many people. A particular company then can take advantage of the forecasts of a number of experts. On the average, this should enable the company to make a better forecast of the related series than it could of sales. Second, in some cases a lead lag relationship may be found between a series and the company’s sales. Income changes may precede changes in auto sales by three months. When such a relationship exists, the correlation with the related series has a direct advantage. A building supply company, for example, has found a high correlation between the sales of its products and building contracts awarded; however, sales lag five months behind building contracts.
Regression analysis has the advantage of being more objective than the previous methods discussed. If sales are related to a widely used series, forecasters have the advantage of many opinions to aid them in forecasting the other series. Another advantage of the method is that it can be done by an office staff or a consultant, thus leaving the executives and sales organization free to carry on their regular operations. In general, regression forecasts are considered highly accurate for short terms such as two years or less.