If your company uses 13four weekly forecasting instead of monthly forecasting, then you would follow the same processes as described below, but have 13 columns of data for each year, instead of 12 columns of monthly data for each year.
Notice that the composite for August shows a value of 7, even though the last three years show values of 9, 10, and 10. Below you will be able to follow each step of the composite build process.
Year 
Jan 
Feb 
Mar 
Apr 
May 
Jun 
Jul 
Aug 
Sep 
Oct 
Nov 
Dec 
2011 
5 
7 
10 
10 
4 
15 
16 

2012 
11 
16 
10 
14 
15 
10 
15 
10 
12 
9 
9 
13 
2013 
10 
5 
7 
7 
6 
6 
1 
9 
1 
6 
6 
23 
2014 
1 
5 
3 
5 
4 

Composite 
3 
5 
3 
5 
4 
5 
4 
7 
3 
5 
6 
15 
At least two complete years of history must exist, before the system will generate a calculated composite. Otherwise, the composite line will contain the same values as the most recent 12 months.
If a third complete year of history exists (as in our example), then it will be used. Otherwise, the composite is built from two complete years of history.
The first step would be to sum up the values for each of the complete years.
Year 

Year 3 
11 
16 
10 
14 
15 
5 
7 
10 
10 
4 
15 
16 
133 
S3 
Year 2 
10 
5 
7 
7 
6 
10 
15 
10 
12 
9 
9 
13 
113 
S2 
Year 1 
1 
5 
3 
5 
4 
6 
1 
9 
1 
6 
6 
23 
70 
S1 
Next, two tests are performed to see if the third year values can be used in the composite calculation. The two tests are:
1) Is the sum for the current year (S1), more than twice the sum for the third year (S3) ?
S1 > 2 (S3)?
70 > 2 (133)?
No
2) Is the sum for the current year (S1), less than half the sum for the third year (S3)?
S1 < 1/2 (S3)?
70 < 1/2 (133) ?
No
If the answer to either of these two tests is YES, then the third year will not be used in the calculation of the composite year. In our example, the third year will be used.
The next step will be to scale year 2 and year 3 to equal the sum of values for year 1. This will help to take out any business cycles (growth or decline) during the composite building process.
In our example below, only the August values will be calculated, to help you follow this process.
For year 3, the sum for year 1 is divided by the sum for year 3 (3 decimal positions, not rounded), and multiplied for every period in year 3.
S1/S3 * Year 3
70/133 = .526
Year 


Jan 
Feb 
Mar 
Apr 
May 
Jun 
Jul 
Aug 
Sep 
Oct 
Nov 
Dec 


Year 3 
11 
16 
10 
14 
15 
5 
7 
10 
10 
4 
15 
16 
133 
S3 
Year 1 
1 
5 
3 
5 
4 
6 
1 
9 
1 
6 
6 
23 
70 
S1 
Adjusted Values (2 decimal positions, rounded) 5.26
For year 2, the sum for year 1 is divided by the sum for year 2 (3 decimal positions, not rounded), and multiplied for every period in year 2.
S1/S2 * Year 2
70/113 = .619
Year 


Jan 
Feb 
Mar 
Apr 
May 
Jun 
Jul 
Aug 
Sep 
Oct 
Nov 
Dec 


Year 3 
11 
16 
10 
14 
15 
5 
7 
10 
10 
4 
15 
16 
133 
S3 
Year 1 
1 
5 
3 
5 
4 
6 
1 
9 
1 
6 
6 
23 
70 
S1 
Adjusted Values (2 decimal positions, rounded) 6.19
After all values in year 2 and year 3 have been adjusted, the system will weigh the most recent year (Year 1) larger than the previous year (Year 2), which should have more influence than its previous year (Year 3).
If year 3 has passed all of the tests from above, then the factors used in the process are:
Adjusted Values (2 decimal positions, not rounded)


Factor 

Aug 


Year 3 

1.00 
X 
5.26 
= 
5.26 
Year 2 

1.50 
X 
6.19 
= 
9.28 
Year 1 

2.50 
X 
9.00 
= 
22.50 












Total 
37.04 
Divided by the total of factors: 5
(Same as multiplying by 20%)
Composite value for August: 7.40
Rounded for screen display: 7
If year 3 did not pass the reasonability tests from above, then only year 1 and year 2 are used, and the factors would then be:
Year 2 1.00
Year 1 1.50
From here you would divide by the total of factors, which is 2.5 is this case. This would be the same as multiplying by 40%. The screen value is not rounded for this type.