Seasonal linear regression

[Former Member]

Hi there, APO Guru's

We have created a forecast profile that uses strategy 35.

However when we run the forecast with this model for say the last 36months of history. The system proposes an almost straight line forecast. If we run it with 18 months history then the system proposes an acceptable forecast. I can now clearly see the sales history pattern in my forecast.

We do have enough data and seasonality in our history for the 36 months to work.

Can anyone maybe shed some light on this.

We also recently upgraded to SCM 5.0

Thanks for your assistance

Robert

[Former Member · Accepted solution]

Before the system applies seasonal linear regression, it carries out a seasonal test. The system uses this test to check if the historical data contains any seasonal patterns. For this, the system determines the autocorrelation coefficient for all periods.

If the value determined is at least 0.3, the system applies seasonal linear regression. If the value is less than 0.3, the system does not recognise a seasonal pattern and applies linear regression.

So I think the test against your 36 months of data returns a value of less than 0.3 and hence it uses linear only, and the value for your 18 months seems to be greater than 0.3.

(P.S. You can change this limit in the /SAPAPO/SCM_FCSTPARA BAdI – Method PARAMETER_SET)

[Former Member · Accepted solution]

Thanks Ian,

This solved our problem. I do however want to know why never had this problem in 4.0. The month before we upgraded to 5.0 everything was fine and now suddenly we have these rules coming into play.

Thank & regards

[Former Member · Accepted solution]

Hi Robert

I believe that this test was introduced in SCM 5.0

Prior to that the calculation working slightly differently.

Pre-5.0 - Linear Regression is performed on the actual values first, then the seasonal indices (calculated during the forecast run) are applied to the result of the linear regression to produce the forecast results. So it will still show some seasonal effect if elements of the forecast show some seasonality.

Regards

Ian

[Former Member · Accepted solution]

Thanks Ian,

The problem I have is that the history does have seasonality to it but because the autocorrelation value is below the limit defined it returns a linear forecast.

Even if I change the limit then I will still have a few of my selections that will fall below the limit.

I do not fully agree with this new logic. Do you think I should log a call with SAP or do you know of another possible alternative?

Thanks

Robert

[Former Member · Accepted solution]

Not sure if it will get you anywhere though it may be worth raising a query on OSS asking why the functionality has changed and if a workaround exists. I cannot think of one off the top of my head unfortunately.

[Former Member · Accepted solution]

Thanks,

I've done so.

regards