on 01-25-2013 10:23 AM
Hi Experts,
i am using weekly forecast profile
Strategy is 11, constant model.
just wanted to know how the system calculating Ex-post and Basic value for this strategy 11.
In the forecast profile i have mentioned alpha is 0.10, no beta, no gama, no singma, periods is 52,
in the outlier correction i mentioned Ex-post forecast.
The values which i marked i could not understand how the system calucting.
Please help me in this.
Thanks,
Hi Sumanth,
The Ex-post forecast is nothing but the forecast for the past periods for which you have actual values.
A forecast that is run in past periods for which actual demand history is also available. The system calculates the forecast accuracy measurements by comparing the differences between the actual values and the ex-post values.
It is not possible to influence the length of the initialization and ex-post horizons.
The system determines the ex-post horizon based on your history horizon (forecast profile). Since you used outline correction settings as ex-post method, the system calculates the ex-post forecast twice.
(Ref: http://help.sap.com/saphelp_scm50/helpdata/en/c5/eea7f4741c4e39aff1d6e4fce4b024/frameset.htm )
Basic Value:
In first order exponential smoothing (i.e. strategy 11), the basic value is determined as
F(t) = alpha*A(t-1) + [1 - aplha] * F(t - 1)
where
F(t) = Forecast value for period t
A(t) = Actual value for period t
Forecast value
Forecast = Basic value + Trend value + Seasonality value
(Note: for strategy 11, there won't be any trend and seasonality)
Ex-post forecast is derived using the above formulas for past periods.
Thanks,
Rajesh
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Ranjesh,
Thanks for your information.
The values in the Basic value are different without mentioned Ex-post in the Master forecast profile in the outlier tab.
In this case the values are correct as your formula.
But if i mentioned Expost in the outlier in Master forecast profile then then values which are populating in Basic value is different and also Ex-post is different. Just want to know how this ex-post value calculating?
Define threshold limits (ex-post + sigma * MAD) - where i have to maintain this threshold limit?
How the system does know the value lies threshold?
Sumanth,
Outlier correction based on ex-post method can't be controlled by means of configuration or any other setting. The formula given by SAP is dependent on the tracking signal. Typically, it's expressed as a ratio of sum of forecast errors and MAD. Generally, the forecast is considered of good quality if | tracking signal | < 4. Then, statistically, sigma can be defined as 1.25*MAD (approx). Using statistical formulas, SAP calculates the threshold limits as Ex-post forecast +(or -) sigma * MAD.
The corrected value will be the average value of upper and lower limits. Please check the following screenshot for an example.
Then, the following example with and without outlier correction should explain the way ex-post and basic values are calculated in the DP. Please check and let me know if you have any queries.
Outlier Correction: Ex-Post Method and Alpha = 0.1 | ||||||||||||||
Period | W 48.2012 | W 49.2012 | W 50.2012 | W 51.2012 | W 52.2012 | W 01.2013 | W 02.2013 | W 03.2013 | W 04.2013 | W 05.2013 | W 06.2013 | W 07.2013 | W 08.2013 | W 09.2013 |
Forecast | 1049 | 1049 | 1049 | 1049 | 1049 | 1049 | ||||||||
Orig Hist | 1703 | 958 | 920 | 1031 | 1703 | 958 | 920 | 1031 | ||||||
Cor Hist | 1157 | 1321 | 920 | 1031 | 1096 | 1072 | 1038 | 1031 | ||||||
Ex-Post | 1019 | 1033 | 1062 | 1047 | 1046 | 1051 | 1053 | 1051 | ||||||
Cor Fcst | 1049 | 1049 | 1049 | 1049 | 1049 | 1049 | ||||||||
Seasonal | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
Trend | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Basic | 1032.85 | 1061.62 | 1047.46 | 1045.81 | 1050.86 | 1052.98 | 1051.46 | 1049.42 | 1049.42 | 1049.42 | 1049.42 | 1049.42 | 1049.42 | 1049.42 |
Basic(t) = Basic(t-1) * 0.9 + Cor. Hist(t) * 0.1 | Basic W49=1032.85*0.9 + 1321*0.1 | |||||||||||||
Basic | 1061.665 | 1047.458 | 1045.814 | 1050.829 | 1052.974 | 1051.482 | 1049.414 | |||||||
Ex-Post | 1062 | 1047 | 1046 | 1051 | 1053 | 1051 | Ex-post(t) = Basic (t-1) | |||||||
Outlier Correction: None and Alpha = 0.1 | ||||||||||||||
Period | W 48.2012 | W 49.2012 | W 50.2012 | W 51.2012 | W 52.2012 | W 01.2013 | W 02.2013 | W 03.2013 | W 04.2013 | W 05.2013 | W 06.2013 | W 07.2013 | W 08.2013 | W 09.2013 |
Forecast | 1149 | 1149 | 1149 | 1149 | 1149 | 1149 | ||||||||
Orig Hist | 1703 | 958 | 920 | 1031 | 1703 | 958 | 920 | 1031 | ||||||
Cor Hist | 1703 | 958 | 920 | 1031 | 1703 | 958 | 920 | 1031 | ||||||
Ex-Post | 1178 | 1231 | 1203 | 1175 | 1161 | 1215 | 1189 | 1162 | ||||||
Cor Fcst | 1149 | 1149 | 1149 | 1149 | 1149 | 1149 | ||||||||
Seasonal | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
Trend | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Basic | 1230.58 | 1203.32 | 1174.99 | 1160.59 | 1214.83 | 1189.15 | 1162.23 | 1149.11 | 1149.11 | 1149.11 | 1149.11 | 1149.11 | 1149.11 | 1149.11 |
Basic(t) = Basic(t-1) * 0.9 + Cor. Hist(t) * 0.1 | Basic W49 = 1230.58*0.9 + 958*0.1 | |||||||||||||
Basic | 1203.322 | 1174.988 | 1160.591 | 1214.831 | 1189.147 | 1162.235 | 1149.107 | |||||||
Ex-post | 1203 | 1175 | 1161 | 1215 | 1189 | 1162 | Ex-post(t) = Basic (t-1) |
The minute difference between the numbers might be due to the rounding logic used internally and please note ex-post and forecast values are determined in the same way. In case of first-order exponential smoothing, your basic value determines the forecast.
Thanks,
Rajesh
Hi Rajesh,
Thanks for your valuable information. Now i am clear.
You mean to say the Ex-post value for the current period is belongs to Previous period Basic value.
I am i right. I calculated the values that what you are saying that is the current period Ex-post value is belongs to previous period Basic value. There might be little bit difference the values in ex-post due to the some rounding.
Now i am clear in Ex-post calculation for Strategy 11 First order exponential smoothing.
Could you please explain in the details in the below screen?
For week 44--> History is 920, Corrected History is 1014, How corrected History is 1014 in week44.
the formula for week44 corrected history is 1090+938/2=1014.
From where the values 1090+938 is taking?
in the below screen Outlier is X. is that mean no outlier happens for that period. right
Thanks a lot lot for your information. I am clear now that what you said. just need few things here.
Hi Sumanth,
Please check the following answers to your queries.
Q1. Corrected history and tolerance limits.
- The system generates the tolerance limits based on ex-post forecast, MAD, and supplied sigma factor (Upper limit = ex-post + sigma*MAD and Lower limit = ex-post - sigma * MAD).
- The corrected history value is the average value of upper and lower limits.
Q2. Outlier correction periods are marked with X
- If there is a outlier correction, the period's outlier column is marked with X and the corrected history value is displayed in red color.
- If there is no X, no outlier correction takes place as the original history falls within the tolerance limits.
Thanks,
Rajesh
User | Count |
---|---|
7 | |
4 | |
3 | |
2 | |
2 | |
1 | |
1 | |
1 | |
1 | |
1 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.