Applying seasonality in Statforecast when there is not enough history

[Former Member]

Hi,

We have a requirement where we don't have enough history for some products. These are new category of frozen products that have no similar products in the company to use a like profile.

Our demand planners are looking at competitor or category in market and using their seasonality and trends to manually add the seasonality through manual corrections but this is getting very cumbersome.

Ideally if there is a functionality in APO-DP where in you can define that you may have a demand seasonality in this period of the year, then another demand pattern in another season of the year etc. and if we can attach this to stat forecast model, then the system would just generate forecast following the seasonality for that product.

Not to be confused with the case where there is enough history that the system could figure out seasonality and apply that in stat forecast. I am talking about a case where there is no like product with enough history and the product itself has not enough history for generating any seasonality pattern...

Has any one came across this kind of situation? If so any insights are greatly appreciated.

There is a functionality named seasonal planning in APO (new I think) but this is only to display or aggregate your data based on seasons...not to apply seasonality.

Thanks,

KP

[former_member187488 · Accepted solution]

Hello KP, I don't know there's a functionality in DP to define seasonality as you're expecting ... I think maybe you can use copy fucntionality in /sapapo/rlgcopy or macros to populate historical data. Realignment copy can be used to copy data from one CVC to other CVCs, while macro can be used to copy data from a certain period to other periods ...

[Former Member]

KP,

Automatic forecast models can come in handy here. There is no foolproof way but here are the forecast strategies you can try as appropriate

52		When you THINK history has Seasonality and no Trend
53		When you THINK history has Seasonality Or a Trend pattern
54		When you THINK there is Trend and when you KNOW that there is seasonal pattern
55		When you THINK there is Seasonality and when you KNOW that there is Trend in Historical data
56		No knowledge of patterns of history

System will do the rest.

However you might like to play with periods per season in forecast profile and see the results,

When you don't have much history to decipher seasonality, system will propose either a constant model or a trend model.

[satish_waghmare3]

Hello KP

I agree with Ada's comment -  I don't know there's a functionality in DP to define seasonality as you're expecting and option she mentioned are very handy as well.

But if I have such scenario to deal with I would recommend to use  Simple Forecasting Method to begin the forecasting process like -

For all such SKUs I will suggest to go with Naive Model (The forecast is equal to the actual value observed during the last period) Although it is good for level pattern, but then it is good starting point for forecasting process to begin and then build market intelligence on it.

Or simple method like  -
Simple Mean(The average of all available data - good for level patterns) or Moving Average(The average value over a set time period, responsive to trend but still lags behind actuals ) or Weighted Moving Average (which will allow emphasizing one period over others; like more weight on recent data, more responsive to trends )

And the Fundamental objective of any forecasting process is simply "DO NO worse than the naive model." Otherwise it is of no use.

Beside a process like this can be helpful as well  -  To enhance Statistical forecasts based on Planner/Forecaster inputs based on market intelligence.

Thank you

Satish Wagmare