While forecasting short term sales using moving average, the problem is that all the data is weighed equally. In general, latest data should be given more importance compared to old data. However, the weights depend on forecast error.

Hence, we use exponential smoothing model to overcome this limitation. In this, recent forecast is adjusted by considering the errors in preceding forecast. There are two popular ways to compute this error:

MSE is better because it highlights larger errors more effectively than MAD.