BASIC IDEA: Data from a number of consecutive past periods can be combined to provide a reasonable forecast for next period. The greater the number of previous periods used, the more "smoothing" takes place.
PROCEDURE: To forecast next period's sales using a simple moving average, use the following procedure: Compute,
Yt+1 = St + St-1 + St-2 + .....+
St-J
-------------------------------
N
where,
Yt+1 = Forecasted sales for next period
St = Current period sales
St-1 = Actual sales of previous period
St-2 = Actual sales two periods ago
St-J= Actual sales J periods ago
N = Number of time periods included
• Note that this forecast is simply an average of past values
Important considerations:
1. How many periods should be used?
• If overall pattern is one of regular growth or decline with few fluctuations, smaller number is usually better.
• If data have been very stable but interrupted by frequent random variations, larger number is best.
• In short, try various numbers and choose one that tends to give the best results (i.e. least forecast error).
2. If a trend is evident in the raw data, try a double moving average model to capture it:
• This is merely the moving average of a moving average! To calculate (using 3 periods), use the following formulas and plug in the appropriate values.
Yt' = St + St-1 + St-2
-------------------------
3
Yt'' = Y't + Y't-1 + Y't-2
-------------------------
3
then the forecast for next period is calculated as:
Yt+1'' = 3Y't - 2Y''t
COMMENTS:
• Fairly simple method
• Slow adjustment to changes in data pattern
• Assigns equal weight to each past observation
MarketingProfs uses single