Seasonal Index in SAP APO

[Former Member]

Good Morning,

I am having difficulty replicating the Seasonal Index for the Seasonal Model in APO DP.  Looking at a 36 month history horizon, I have been able to replicate the initialized seasonal indices for months -24 to -13 (with current month, 4/2015, as month zero).  This is documented to be the order history for the same month divided by the average demand for months -36 to -25.

However, the next 12 months' seasonal indices (months -12 to -1) are not working according to the documentation (shown below):

When working with a very low gamma (such as 0.01), my values using the formula are coming very close to the values that SAP is returning in the Exceptional Statistical Forecast Planning Book.  However, when I change the gamma to something higher (such as 0.30) in both my formula and the Exceptional Statistical Forecast Planning Book, the values are increasingly different.

Anybody have any ideas on what I may be missing?

I appreciate the help!

Best Regards,

Steven DeLisle

[Former Member]

Steven,

As I understand the calculation: For 36 months of history and a 12 month season (L),

1. the initialized (starting) Seasonal index values are the Actual History of each period/Mean of the Sum of months -36 to -25. So you end up with each month having a seasonal index value (for periods -36 to -25).

2. For the next two years (-24 to -1), you will reference the previous year's(Season) seasonal index value (L=12) and then using the gamma value you weight the period's history/Basic value and remove the last season's seasonal index value (The gamma is a 'weighting value'). Think if a  Gamma is '1', then the last year's seasonal index has no impact on the next year. If the gamma is .01, then last year's seasonal index value is 99% of this year's value. Almost

I say almost because the formula for the Basic value G(t) is impacted by last season's index value, but that is weighted by the alpha or constant parameter.

Your example above I think you skewed the values so that for -24 to -13 you used the mean from the 3rd year (-36 to -25).

I hope this helps,

Steve

[Former Member]

Steven,

Are you using model 30 or 40 ?

Couple of pointers

The value of Seasonal index depends on existence of trend (beta) and also the assigned value of alpha. What is called Additive seasonality (without trend) and multiplicative seasonality (seasons with trend) resp.

So you also need to check how G(t) - Base ? is calculated.

Attached are a couple of images. I compiled these long long ago from a pdf. document written by a good teacher.

Seasonality in Level Demand

Seasonality in Trended Demand (Winters model)