extrapolative methods) such as exponential smoothing. Regression’s track record against these is mixed. Regression is likely to outperform univariate methods when
the causal relationships are strong;
the causal variables can be accurately measured; and
future values of the causal variables are under management’s control or readily predictable.
In forecasting, there are several benefits in using lagged variables, both of the variable being forecast and of the causal variables. Regression also works well when one or more of the causal variables is a control variable; that is, its value can be set by the organization, such as the amount in the promotion budget and the timing of its use.