Tolerance limit in DP

[Former Member]

Hi Experts,

Does any one help How Upper Tolerance Limit and Lower Tolerance Limit calculates for Corrected History KF?

The sap formula is

Upper Tolerance Limit = Ex-post Fcst + sigma*MAD

Lower Tolerence Limit = Ex-post Fcst + sigma* MAD

In the above formula i know Expost Fcst value and sigma value But do not know how MAD calculates?

Can any one help how MAD calculates?

Advance Thanks,

[rajkj]

Hi Sumanth,

MAD is the ratio of sum of absolute values of differences between actual and forecast values and number of periods considered.

i.e. MAD =  SUM( e(i) ) / n

where e(i) = | Actual value(i) - Forecast value(i) |

and n = number of periods.

http://en.wikipedia.org/wiki/Mean_absolute_error

Thanks,
Rajesh

[Former Member]

Hi Thanks for your information Rajesh,The system is not showing right result for MAD.

I have history 3 years in past (156 weeks in past) and Forecast 2 years in Future (104 Weeks in future)

so you want me take total of Past History and Total of Future Forecast

say for example:

The total of history in the past is = 98457
The total of Forecast in Future is = 151302 98457-151302 / 260

This 260 =  Past 156 weeks + future 104 weeks?

Is that above formula correct?

i am not getting the desired result in the system the way you mentioned. please help me in the below screen.

or

Give me some guide in the below screen which is for week week12 and week13.

Week 12 and 13 have corrected forecast. How the system is determining Upper and lover limit for week12 and week13.

the sigma factor for this is 0.60

[Former Member]

Sumanth,

The system is not showing right result for MAD.

Quite a statement.  A better statement might be that the system does not meet your expectations; the system is most likely calculating exactly what it is programmed to calculate. 

Read SAP's documentation of what logic THEY use in their various forecast algorithms (including their formula for MAD).

http://help.sap.com/saphelp_scm700_ehp02/helpdata/EN/ac/216b8c337b11d398290000e8a49608/frameset.htm

If your system is not consistent with the SAP documentation, then search OSS notes; or raise a message with SAP.  There is noone here who can modify your system.

I have personally never found it profitable to try and analyze or duplicate SAP's forecasting algorithms.  Much effort can be spent, but in the end, all you have is still just a forecast, which will inevitably be WRONG when compared to the actual.

Best Regards,

DB49

[Former Member]

Sumanth,

Pay attention to your selection on which forecast was run. MAD or any other measure of forecast error is meaninggful only at the selection level at which forecast is run.  Rest is disaggregation and aggregation.

If you try to measure it at some other level in your excel sheets, it will never match. So make a distinction between WHAT you are actually forecasting and WHAT you are reporting the error for.

Hope this general gyan helps

Thanks

BORAT