Running LC Planning piecewise

[srinivas_krishnamoorthy]

Hi

We have a situation where we need to forecast of A to be based on history of A until Dec end beyond which it needs to be based on AB. We have thought of couple of ways of doing this but none of them looks elegant on paper. (Eg. Using two forecast profiles - one with LC and another without LC, or using multiple Key Figures where one Key Figure could be some type of indicator to switch to history of AB ) PLease let me know your views.

Although we found settings for date effective Phase-in/out profile we could not find any levers around LifeCycle planning.

Thanks

Srinivas

[Former Member]

If you dont have too many of these products to add up then you can use a new Keyfigure which has a macro calculation that copies A till dec and then does the adding up of history of the two products

Use this new KF as the basis for forcasting.

you might have to run forecasting for these added up products in a separate run so as not to affect the other products

the macro might affect performance but if you hide the KF from interactive planning and run it on the background it should not matter

[srinivas_krishnamoorthy]

Harish

Just wanted to make sure I understand. The Dec I was talking was the one in the future. The forecast for Product A till Dec 2007 should be based on Product A beyond which it needs to be based on AB history. Does the Macro look at some indicator kind of thing to switch to Forecast based on AB? Where will planner maintain these A+B relationships to get copied to a new key figure? Am sorry if this looks to be a dumb question, but I need to sort this out in my mind.

Thanks

Srinivas

[Former Member]

This is what i was thinking of

these are just conceptual blocks. you can replace the steps with other steps if they look better

lets say you have these KF

History

Compounded history

Forecast

Compounded Forecast

You first Calculate Compounded history using a macro that will drill down history. pick up values of the infoobj A and B and add it up

Next you calculate Regular Forecast based on History for A

Then you calculate Compounded forecast based on Compounded history of A+B (using a different forecast profile) - just make sure it goes to char value A

Use a macro to copy from Compounded Forecast into Regular forecast starting Jan 2008. Macros are time dependent. you can design the copy to happen from Jan 2008. that should not be a problem

Is this a bit more clear...

But these might turn out to be complicated macros particularly if you are going to define calculation of individual char values within a KF

[Former Member]

talking about using another version, i suppose you can do the compounded forecasting in another version and use TSCOPY to copy back just the values after dec 2007 into the version you are using

[srinivas_krishnamoorthy]

Thanks for the answer.

Just an additional question. Is there a way to parameterize the macro? I dont foresee that planners will be manipulating the macros for achieving the desired result. It is just not their nature to be technical. They can however be taught to do very basic stuff such as checking a check box or setting some date somewhere etc. Also, hardcoding Infoobject values in macros may not be long term solution.

The business scenarios can easily get complicated. Like it can be a situation of

Use History of A until Dec 31st

Use History of A+B until Mar 31st from Dec31st

use History of B until 30th June from 31st Mar

This happens in very regularly in our business scenario where sometimes we have groups of products upgrading to new ones in short lifecycles.

Using multiple Key Figures might acheive the results but I was wondering if we could have some more elegant solution. I was pondering on the other solution proposed in the mail thread using Time Series Weighting factors.

Thanks for the reply anyway

Srinivas

[Former Member]

macros can take user input from popups (check the example macros in 9AEXAMPLES)

they can also take highlighted rowes and columns

I would tend to incline more towards running this in two versions and copy back to overwite for the periods after dec 07 (in your example).

you can control the forecast profile in the second version and you can control the copy back by char and by dates

[Former Member]

Hi Srinivas,

I think that you have a good idea, you can use two forecast profiles one with LC and the other without LC, but you can not assign two profiles to the same selection, so how are you doing this switch between profiles?

I'm not sure, but may be you can use only one forecast profile with LC, in this LC you can use a weighting profile that mitigates the first historical values for product B. May be this could work.

If this can help you or if you find a better solution, let me know please. i'm interested in this topic.

Regards.

Marí

[srinivas_krishnamoorthy]

Hi

I thought while defining the activity I could define two of them with different Forecast Profiles. So while running the Forecasting calculation in background, we could leverage the activities with an appropriate planning job.I understand that online forecasting is possible only using one of the profiles at a time based on the selection assignment to the profile. My hope was to more get around the background job problem. We could seed the Manual Forecast with the stat forecast generated through the above method.I was planning to develop a 2nd Forecast Profile to generate forecast from 31st Dec using LC planning onwards with an appropriate Forecast Horizon. I am not sure if I can work with such "piece-wise" Forecast Profile approach. let me know if anyone has experiences of using similar approach

Thanks

Srinivas

[Former Member]

Ok, may be you can run a forecast with LC from today to X number of periods, and after this forecast, you can run a second one without LC and only running from today to 31st dec. I think that these forecasts must be ran in this order because the second can overwrite the first periods.

I wouldn't run the forecast with LC at a second place because in order to this forecast not overwrite the first periods, you would have to use an offset so the forecast would show the behavior of present period in January 2008.

May be this can help you.

Let me know if this works or if you use another approach.

Regards.

Marín.

[srinivas_krishnamoorthy]

I am not sure if we can specify X number of periods for LC planning as it will become a situation of rolling buckets. So when we come to close to Dec 31st, the dates would have slipped by the elapsed time. I like your approach though and may be we should have fixed Forecast Horizons set for the profiles. So as an example LC planning with history of A+B can run all the way to the end of time zone. The second run could just be based on history of A with fixed dates in the Forecast Horizon upto Dec 31st. This second run hopefully should overpost the prior run for the overlap period. This solution can possibly avoid multiple Key Figures though we may have to have multiple runs with specific dates in the Forecast Horizon. I think that looks to be decent solution to work upon. Will let you know how it works or any alternate/tweaked process I come up with.

Thanks

Srinivas