The manager is open to suggestions for forecasting models and decides to try both an exponential smoothing model with an alpha of 0.9 and a simple moving average of two periods

He decides to use the actual value for January as the forecast for February just to get the exponential smoothing forecasting party started. For the moving average approach, he uses the actual for January as the forecast for February and the actual for February as the forecast for March before beginning to apply the moving average equation. Generate forecasts for the year using these technique and then calculate forecast errors using MAD to determine which is the superior method in this scenario.

Answer: The forecasts and their errors are indicated in the table:

Month Actual Forecast MA MAD MA Forecast Expon MAD Expon
January 474
February 485 474 11 474 11
March 501 485 16 484 17
April 588 501 87 499 89
May 579 588 9 579 0
June 673 579 94 579 94
July 594 673 79 664 70
August 679 594 85 601 78
September 608 679 71 671 63
October 699 608 91 614 85
November 719 699 20 691 28
December 732 719 13 716 16

The MAD for the Moving Average technique is a whopping 52, while the exponential smoothing has a MAD of 50.1, so exponential smoothing would be the preferred approach from an accuracy standpoint.