Interactive planning questions

[Former Member]

Hai,

I have the following questions in DP.

1. How often is a demad plan generated? Is it done interactively everytime or done in background?

2. forecasting workshop- is it part of blueprint summary

3. Is a forecasting model setup in the univariate forecasting and the user uses it? or a general decided model is entered and the user decides which one to go forward with while running interactive forecast planning?

4. How is the best forecast model decided in interactive forecast planningI mean based on MAD or MAPE or MSE etc.? Is it based on Mean Average Deviation(MAD)?

Thank you.

[somnath_manna · Accepted solution]

1. How often is a demad plan generated? Is it done interactively everytime or done in background?

Depends on what you mean by demand plan generation. Normally demand planning is a monthly exercise in an organisation. There are different activities as part of demand planning which are done interactively (like review of demand plan and/or management by exceptions) and in background (statistical forecasting, macro computations, alert generations).

2. forecasting workshop- is it part of blueprint summary

Not sure where this question is coming from. During Blueprint Phase you do an exercise to understand the business need and processes followed for existing statistical / timeseries forecasting methods. However detailed analysis and decision on the methods and parameters to be used is done later - mostly with realisation. Note this is just a starting point - the organisation needs to continuously track, measure and manage statistical forecasting methods after go-live to get the best out of the system.

3. Is a forecasting model setup in the univariate forecasting and the user uses it? or a general decided model is entered and the user decides which one to go forward with while running interactive forecast planning?

Normally you run statsitical forecasting in background and then the user reviews it interactively. He can rerun forecast using some other method or parameter values interactively and then save the results if he likes them.

He may setup such that next time the product (actually product family/category) when runs in background follows the new parameters and/or method.

4. How is the best forecast model decided in interactive forecast planningI mean based on MAD or MAPE or MSE etc.? Is it based on Mean Average Deviation(MAD)?

MAPE is supposed to be the better measure for forecast errors compared to MAD and MSE.

Thanks,

Somnath

[Former Member]

am not sure if there is one answer to many of these questions and would depend on the business requirements a lot. I can give a bit of my opinion though

1. how often would depend on the demand planning business process. ideally before a demand plan review meeting ...so that it has a current data. also depends what part of the demand plan has interactions with other parts of the business- what parts of the demand are you planning? forecasting? promotions? do planners/managers get to edit data interactively, what are the inputs, and how much time is needed after runnign the demand plan to finalize the plan or to present it. plot your whole process and see what frequency fits in.

assuming you have a lot of data and forecasting and copying and running macros to do it would be in the background with an option of running selectively online

b) forecasting workshop - are you talking about the SAP presentation on the same topic? depends on whether your customer wants it and if its part of your scope? some might have good forecasting practices and knowledge. you need to probably explain jsut the functionality. again the stage is dependent on your project plan and delivery methodology. it could also be for training or proof of concept or any other stage...

c) if you know what model fits a product assign it. else use a auto model selection and see if it has selected the right model. tweak the model paramets and see how it goes.

d) All the forecast accuracy results measure different things. MAPE is suppsoed to be better. you should try this out and see what is the best fit based on forecast vs actuals

also you can give your errors different weights and run a comparison in forecasting and see which model you want to choose