Statistical forecast

Dear Experts:
I created master profile and a univariate profile as per the building block instructions. In the planning view, when I click on the Stat button, i am taken to a different screen with different planning view. I noticed the same history is copies into the forecast keyfigure row. I tried using different models (with different alfa values for exp models and moving average etc). I get the same figures in the forecast key figure row. Also the green arrow on the top which is supposed to take me back to my planning view is not active. What am i doing wrong?
Thanks
Steve

Hi Steve,
Did you compare the forecast errors? are they same with all the different values? You can check forecast errors by clicking on the icon with two downward arrows. Which forecasting model are you using? May be the values you are entering for alpha are not making much difference. Another check would be to use different forecast models and compare errors. Here is a document with formulas for [statistical models|https://www.sdn.sap.com/irj/sdn/go/portal/prtroot/docs/library/uuid/7a4025f8-0a01-0010-0fb2-b7ab22597675]
Try using one of the models and with the help of formula for that model, see if your values make sense.
Hope this help.

Similar Messages

S&OP on HANA: Error in statistical forecasting - escapenull/complete missing data = 1

Hello,
we created a planning area with storage time profile level days. For the statistical forecast, the input and output key figures are defined per month. We defined a statistical forecast profile with Single Exponential Smoothing as method as well as two values for alpha attached to this profile.
When executing the forecast, it stops with the following error message:
Using parameters: forecast model id = 10010, Input key figure = AGGRORDERENTRYQTY, Output key figure = STATFCSTQTYSINGEXPSM, Starting historical period = 11048, Ending historical period = 11411, Time horizon start = 10682, Time horizon end = 12142, Current period = 11412, Period offset = 0, Period attribute = PERIODID3, Past period offset = 0, Future periods to forecast = 365, Filter id = 0, Scenario = BASELINE, ESCAPENULL/Complete missing data = 1
We changed the time parameters of the forecasting profile and the planning area and tested different combinations, but the error repeats.
I would appreciate any idea how to solve this issue.
Cheers,
Jonathan

Scot,
K is not needed...Ref to Stat Forecasting - Moving Average
Thanks,
Krishna

Selecting the Planning Level for Statistical Forecast in S&OP on HANA

Hello together,
we currently face an issue regarding the statistical forecast in S&OP on HANA:
The customer wants to implement an S&OP process containing the two steps statistical forecast (automatically calculated) and sales forecast (manually determined). The planning level of the statistical forecast should be higher than the planning level of the sales forecast, i.e. the statistical forecast is calculated on a higher planning level and then disaggregated to be an input for the manually determined sales forecast.
Since it doesn't seem to be possible to disaggregate any key figure below its base planning level we need to have the lower sales planning level as base planning level in all key figures, including the input and output key figures of the statistical forecast.
When running a statistical forecast with a defined forecasting profile, the planning level used for the forecast seems to be the base planning level of the input/output key figures defined in the forecasting profile.
However, we need to define a higher planning level for the statistical forecast in order to have a better data base for recognizing trends and sesonality. In APO DP it is possible to define the planning level for the forecast when running/scheduling the forecasting job. Is this also possible in S&OP on HANA? Can we define the planning level for the forecast in the forecasting profile or when running the forecast from the Excel AddIn? I tried to use the field "Filter" in the forecasting profile as it listed all the planning levels (as well as a lot of other things, we don't know where they come from), but the forecast was still carried out on the base planning level.
Do you have any other idea where to define the forecasting planning level that is different from the base planning level of the input/output key figures?
Thank you in advance,
Jonathan

Hi Bert,
thank you for your fast answer. I checked the forum again and found this post:
Question about Statistical Forecast
However, I don't really get how the disaggregation is meant to work; the suggestion seems to be to calculate a forecast on an aggregated level, i.e. with input and output key figure with an aggregated base planning level, and then disaggregate it "somehow".
Do you have any idea how this disaggregation might work and what other steps I have to consider?
It seems as if I'd have to define a whole bunch of new key figures and have more than one output key figure but no input key figure for the statistical forecast, which doesn't make sense in my eyes.
Actually I was looking for a solution where I can define my forecasting profile with both input and output key figure on the lower planning level and then insert the level I want the forecast to be done at either in the profile as well, or even more flexible when running or scheduling a forecast. Defining more key figures any time I need another forecasting level does not appear very comfortable. Is it meant to be like that or is there another way to reach the goal mentioned above?
Cheers,
Jonathan
Edit: The mentioned thread contains some valuable information on how to disaggregate key figures. However, the really tricky decision is how to define the disaggregation ratio.

Statistical forecasting....Input and Output at different levels? & MAD and MAPE calculations

Team,
Couple of questions on Statistical Forecasting...
1) Is it possible to have input and output KF at different planning levels?...
2) Is there anyway to see MAD and MAPE values for each Forecast run?
Any help is appreciated....Thanks for your help...
Thanks,
Krishna

Hi Krishna,
#1 - The Inputs and Outputs should be at the same base planning level. Output should be a Stored KF. The input can be a calculated KF.
To achieve input and output at different plan levels, you can do something like this:
I/P KF1@PL1 (stored) Output: KF3@PL2
Create a Calculated KF: KF2@PL2 with calculation like SUM(KF1@PL1) and use this as input for StatFcst
#2 - The result of Fcst run MAD cannot be seen from Excel Client yet, we will address this in near release. If you are on-premise, you can see the results in table: sap.sop.sopfnd.catalogue::SOPFCS_FORECAST_RUNTIME_INFO_TAB
Hope this is useful.
Thanks,
Raghav

The statistical forecast not for the current month

Hi All,
The statistical forecast is done in the monthly buckets.I would like to run the statistical forecast job at the beginning of the month but the statistical forecast value for the current month must be not changed. I've tried to fix the key figure for the current month but the new forecast is not saved at all. Do you know how to achieve it?
Best regards

Hi Guz,
You should be able to do that using the 'offset' functionality in the Forecast profile screen...
Thanks,
Subash

Statistical forecast varies in back ground and foreground !

Hello friends,
When i generate the Statistical forecast at *product and plant* level in foreground it gives some values, then generate in background at *product and plant level* again it gives different values.Could you pl through me some lights on me.
Regards,
siva

Hello Siva，
Maybe you have different aggregation level for your foreground forecast and background job.
Please first check your 'aggregation level' setting in /sapapo/mc8e.
Or maybe you're using different forecast profiles for your foreground and backgournd.
Please check the settings in /sapapo/mc8t of the forecast activity.
Do you have "always use job profile" setting on?
You can also check the background run's result in SM37's spool file, to see on what level did it execute, and what forecast profile it used for each planning object.
You can check note 372939 to compare the results. In this note, you will have great tips on how to compare both results. Besides that, note 374681 will give you detailed information on how to have the selections when using mass processing.
Also, check the note 546079, which contains more informations on how to have Interactive and batch forecast to run equally.
Best Regards,
Ada
Edited by: Ada Lv on Jul 29, 2011 3:22 PM

Selection exclusions not considered in statistical forecasting

HI,
Our planning object structure has total 6 characteristics and some of them are product, location, division and region. Product has product class as a navigational attribute.
We have created 2 selection ids in planning book.
id1: display "division" and entered some divisions and excluded some regions and product class
id2: display "division" and entered some divisions and excluded some regions.
When we ran statistical forecasting, the statistical models ( moving average model used but we tried other models too) the system is still forecasting for the excluded selections in planning books.
Here are my questions:
1. In case of id1 since there is a navigational attribute, can we use navigational attributes in statistical forecasting.? I remember not but not totally sure...any supporting documentation or references are helpful.
2. In case of id2, I can not imagine why the excluded values are forecasted..even the POS characteristic value exclusions are ignored. I think this is a bug...did any one see this kind of behavior? any suggestions or OSS notes are appreciated.
Thanks.

Hi,
Your aggregation level explains your results.
What is happening is that based on your selection criteria, divisions are getting selected.
Since your aggregation level is division, hence for whatever divisions have got selected as per your selection criteria, forecasting would be done based on history aggregated at division level. Now it doesn't matter whether you maintained exclusions at any other level or not.
Result would be forecast for the selected divisions at an aggregated level. Now, this forecast would get disaggregated to ALL the CVCs which contain these selected divisions, and the disaggregation would happen as per the calculation type that you maintained for the stat forecast.
If you don't want the forecast to be disaggregated to levels below division, then do the forecast at a lower level of aggregation, meaning select all the characteristics in "aggregation" corresponding to the levels which you want to exclude.
Hope this explains the situation to you.
One way you can say that the "display" characteristic (topmost row) that you specify in selection when you load data in interactive planning corresponds to the aggregation level that you maintain in background.
Thanks - Pawan

Statistical Forecast cannot be deleted

Hello everybody,
How do I go about replicating the issue in the system's test environment.
But the main issue I am having is that statistical forecast cannot be deleted. Normally this should be done via transaction /SAPAPO/MSDP_FCST1 - Lifecycle Planning .
While selecting the phase out profile - the material number disappears. Not sure if this is a known issue or not.
As a workaround I deleted the statistical forecast via SDP94 - univariate forecast execute manual phaseout - the statistical forecast disappeared. But over the weekend run the statistical forecast was populated again.
Thanks.
Edited by: apo_dp_consultant on Jan 8, 2012 4:38 PM

Hi,
You should use /sapapo/md74 to delete the forecast.
Select all the 3 - "display log", "delete orders" and "all planning segments".
Also, take out the setting "test mode".
What exactly is your requirement and your issue?
For some materials, do you want that forecasting should not take place? And is it that you don't to forecast these materials due to certain reasons, or because these materials are phased out, you want that no forecast should take place?
Answer to above questions would help in guiding you in the right direction.
Thanks - Pawan

Statistical forecast question

Regarding Stat Forecasts......Where does the calculation of the statistical forecast actually run? And the historical data that APO uses as the basis for the stat forecast, where does that lie? From what I understand the historical data is actually in the planning area, and therefore livecache. But the macro to then create a stat forecast, does that just run in the background on the APO server?

Hello David -
All the calculations for the forecast run depends on the forecast model that you chose. based on your choice calculations will be performed in background and the stat forecast keyfigure will be populated.
about the historical data, it comes from the infocube as you are uploading the historical data into the infocube. your planning results are stored in livecache for easy access.
normally or best practise is that : current planning results are stored in live cache and the previous forecasts are moved to infocubes for storage purpose as well as to generate reports depending on the requirements of the business.
Hope this helps.
Regards,
Suresh Garg

Statistical Forecast Models Using Forecast Assignment Table

Hello all. I need some help.
I am trying to setup a background process to execute Statistical Forecasts in background using the models stored in the table /SAPAPO/DP_FCST2
Can anybody offer advice with regards the best approach.
Is there a specific Forecast Strategy or User Exit that should be used?
Thanks

Hello Christopher
Control Parameter in the activity needs to be setup correctly.
Additionally please refer this old thread
Forecasting interactively and background in DP
Just want to add some additional information on this topic -
In Planning Book, Goto Settings => Forecast Profile => Assignments: There are three options available -
1) Save assignment of selection for forecast profile - This allows users to save changes to original forecast profile
2) Save assignment with unique forecast profile - (this allows for the planner to fine tune a model but then save it to a unique GUID instead of over-writing the original profile itself. Typically, DP Planner are expected to use this option.
3) Do not save assignment and forecast profiles - This also allows users to over-write the original forecast profile settings without saving the assignment
Additionally if this /SAPAPO/FCST_GUIDS parameter is set to “X” then it will result into system saving any forecast profile changes as a unique forecast profile(GUID based).
Also when generating the forecasting (using option#2 above )interactively, it is important to use same planning levels(selection id) as defined in aggregation levels in background job.
Hope this will help.
Thank you
Satish Waghmare

Statistical forecast - automatic model

Dear Experts,
I have specific question regarding statistical forecast in DP
i am using automatic model selection -2
outlier correction - expost method
mape upper limit set to -20 % in diagnosis group
forecast profile - uni variant
now i have
Invoice History -
Jan-13
13-Feb
13-Mar
13-Apr
13-May
13-Jun
14-Jul
13-Aug
13-Sep
13-Oct
15-Nov
13-Dec
105
204
244
455
22
45
25
67
76
68
89
100
now i run statistical forecast
what will be the result
Jan-14
14-Feb
14-Mar
14-Apr
14-May
14-Jun
14-Jul
14-Aug
14-Sep
14-Oct
15-Nov
14-Dec
105
204
244
455
22
45
25
67
76
68
89
100
my question is does jan 13 invoice history is taken for jan 14 forecast ?? like wise for other month too
please advise how the system use time bucket and to calculate the forecast
i also have question regarding ex-post method and median method and none
whats the difference between these methods
Regards
Raju

Hi Raju,
Take a look of the SAP documentation. Here you have the detail of what the Auto Model Sel 2 is doing:
Automatic Model Selection Procedure 2 - Univariate Forecasting - SAP Library
Anyway, maybe because I don't understand exactly what the system is doing or maybe because I have found inconsistencies, I would not use the automatic model selection models in APO DP.
Take a look of these articles talking about some inconsistencies:
Getting Best Fit to Work in SAP? - Demand Planning
Best Fit Forecasting in SAP DP - SAP Planning
As you can see in the first article, Auto Model Sel 2 uses to copy the data.
If you are trying to identify the model, take a look of the Sales Pattern, then run different methods, play with the parameters (alpha, beta, gamma) and determined what should be the model and the set of parameters base on the MAPE.
For ex-post method (constant), you can read this article:
APO DP - Forecast Model Parameters: First-Order Exponential Smoothing
Looking at the SAP help documentation you can see the difference between Ex-post and Median methods:
Ex-Post Forecast - Univariate Forecasting - SAP Library
Median Method - Univariate Forecasting - SAP Library
Kind Regards,
Mariano

How do I delete Statistical Forecasts for components that have no history

Hi
How do I delete Statistical Forecasts for components that have no historical data, but the Stat forecast model is still forecasting values for components.
Other than creating a macro to clear the key figure columns interactively in the planning book, where in the model or background job do I activate or set that components without 12 months historical data are not forecasted .i.e. zero values in forecast for these components?
Thanx
Keegan

Hi Keegan,
To my knowledge there is no standard way you can tell the system not to forecast if the last 12 periods are zeros.
We have the same issue and we use 3 methods to get around this:
1. A macro as you said
2. Create a phase out profile in life cycle profile with parameters, start date today to some 12/31/9999 and set the "Before start date, apply constant factor" 100% and "After start date, apply constant factor" 0%
and check the indicator " Life cycle planning " in master forecast profile.
This puts zeros for all future values but does not check if the history has zeros in it.
3. Delete unwanted CVCs so that we will not see any issues with disaggregation and junk data.

Statistical Forecasting - Directing Forecast Results to Spool

hi,
We are executing planning jobs ( /n/sapapo/mc8g) for Statistical Forecasting. The planning jobs have been set up through /n/sapapo/mc8e with the required parameters and forecast profiles attached.
Post successful completion of the planning jobs, we get the results in the following manner :
1. Forecast Results are getting stored in the specified key figure in the interactive planning book
2. Job execution messages and information on parameters such as MAPE, MAD etc are directed to the spool.
Is there anyway in which we can also direct the forecast results to the spool ?
regards,
Anirudha

hi Luiz,
Basically I wanted the forecasting run results to be directed to the spool to be viewed at a later point of time.
What we are currently doing is trying out the various models for statistical forecasting ( using Univariate profiles ). SInce we are having to execute a large number of runs, it is pretty time consuming as we have to execute it for a specific brand, note down the forecasted numbers ( populated in the interactive planning book for the forecast horizon ) and then move on to the next model / brand.
To simplify and speed up the process, we wanted to try out the following :
1) Create the required number of univariate profiles with various models and parameters
2) Create Planning Activities ( with the corresponding profile attached ) and Planning Jobs
3) Execute the planning jobs through mc8g
4) View the job output in the job spool at one go
The issue we are facing is that the forecast results are getting written only to the interactive planning book. And not directed to the spool. I have listed the sample spool output of the planning run job below. If you observe the spool, only run related info and MAPE, MSE etc are listed.
How do we write the forecast results ( for the forecast horizon ) to the spool also ?
Sample Spool
10.03.2010              Run planning job in the background
Charact. Combination(1 Frm 1):
DELM ADV
Following forecast profile selected: Mil_Forecasting
Required Time for Forecast in Seconds: 0.007
Univariate forecast successfully executed
Seasonal test called
Number of periods: 12 , Autocorrelation: 0.359 , Limit: 0.30
Season test is positive
MAPE: 49.78 MSE: 787878.39 RMSE: 887.63 MPE: 28.64-
MAD: 485.02 Error total: 1895.38
Data saved successfully
regards,
Anirudha

Statistical Forecast not saved when generated with background planning job

Hello All,
The system I am working on is SCM 5.0 Support pack 15.
When I execute the planning job in the background to generate a statistical forecast, the job completes sucessfully, I have a spool that shows the forecast profile was executed (Auto Mod 2) with the appropriate model messages and the last two messages are 'Data saved successfully' and 'No data has been changed; data must not be saved', but when I go into interactive planning to view the forecast, there is no forecast in my key figure.
My planning area has the forecast settings maintained for Statistical Forecast, Corrected History, and Ex-Post Forecast key figures. My forecast profile is looking at the History key figure and Statistical Forecast key figure. My time series has been initialized for 36 months of History and 18 months of forecast. My planning book is set to have the Statistical Forecast Key figure as output only. My activity and planning job are set to execute in the appropriate planning book (in order to see the forecast alerts I have selected) with the specific selection profiles (no navigational attributes), and the aggregation level set (no navigational attributes).
The only way for me to see a forecast is to change the Statistical key figure setting in the planning book to output/input, execute the forecast job in the background, and then go into interactive planning. Am I missing something? Any help you could provide would be greatly appreciated.
Regards,
Roxane H.

Hello Senthilrajan,
I have executed the forecast in the foreground using the same planning book as the background job and same forecast profile (Auto Model 2)with a selection from the batch run (pulled a selection from the spool list) and it generated a forecast with the same messages as my background job except for the last two lines (Data saved successfully and No data has been changed, data must not be saved) that I was able to copy back into interactive planning. I then used this selection in the background job to execute the forecast in the background and did not see a forecast in interactive planning and received the same messages as before -
Data saved successfully
No data has been changed, data must not be saved
The reason we run the forecast in this book is to allow our Demand Planners to view the forecast alerts by book and/or selection directly in interactive planning instead of having to get their from the Alert Monitor. It has worked in the past and is currently working in the same environment just with a different planning book, selections, forecast profiles, and data. It generates a forecast with no 'No data has been changed, data must not be saved' message and the forecast is written in the key figure and is viewable in interactive planning with out manipulating the key figure to be input/output (it is set to output only).
Any further assistance you can provide would be greatly appreciated.

2 statistical forecast lines in one planning area

I would like one planning area to have two statistical forecast key figures based on two different history key figures.
You can assign only one key figure as a statistical forecast in a planning area, so it seems DP doesn't have a capability to create and store two separate stat forecasts.
Please let me know if you had an experience to configure the system in the way.
Thanks!
Regards,
Yuji.

If you create a forecast profile (demand planning --> environment --> Maintain Forecast Profiles (/SAPAPO/MC96B )) , then you can use multiple forecast key figures in the same planning book and data view. Each profile will have a specified forecast key figure and historical key figure.
I hope this helps.
Regards,
James

No Statistical Forecast showing on the Graph

Hi All,
We have an issue in that when you enter into the GRAPH to look at the forecasts, there is no line shown for Statistical Forecast though are settings are selected to show the line. However, when you enter design view, it is shows the statistical forecast. Then when you return back to the live view, the statistical forecast is showing now on the graph. Anyone else had this experience and can something be done to put it right.
Mark

Hi Garry,
Since this forum is for powershell related issue, to solve the Excel VBA issue, I recommand you can post in Excel For Developers Forum for more effective support:
http://social.msdn.microsoft.com/Forums/office/en-US/home?forum=exceldev
Thanks for your understanding.
If you have any feedback on our support, please click here.
Best Regards,
Anna
TechNet Community Support

Statistical forecast

Similar Messages

Maybe you are looking for