IBP Forecast Error assesment

[OliverA41]

Hi,

I have a general question about the topic "Forecast Error". How can I say, the Forecast Error e.g. Mape is good and on the other side, when is it a bad value.

When is the point achieved, that you check the input parameter and other thinks ?

How do you deal with the values ?

[piyush_parekh · Accepted solution]

Hi Oliver,

Usually, higher the error value, lower the forecast accuracy. (not always though!!) If you've observed Best Fit forecast models, by design SAP IBP also compares errors from different forecast algorithms and considers the algorithm with lowest error as output of best fit model.

Choosing one forecast error measure over the other depends on whether demand has trend, seasonality, intermittency, pattern etc. For e.g. MASE is most suited for intermittent demand. It also depends on what information are you trying to extract from error and at what level.

Consider following scenario -

Refer to MPE values at week level, one can understand it's either overforecasting or underforecasting. But on an average, error is 0!! So it makes sense to refer MAPE values at aggregated level. However, referring MAPE values at week level doesn't indicate whether planner overforecasted or underforecasted.

You can further study how different error measures are calculated in SAP IBP here.

https://help.sap.com/viewer/feae3cea3cc549aaa9d9de7d363a83e6/1905/en-US/32cc1c552770a20ce10000000a44...

Based on the context and looking at the mathematical expression, you can decide which error measure is best suited for a particular business scenario.

Regards,

Piyush

[OliverA41]

Hi Piyush,

thanks for your response.

You wrote a sentence, that is exactly that, what I search

"For e.g. MASE is most suited for intermittent demand."

When is which forecast error usefull

- MPE

- MAPE

- MSE

- RMSE

- MAD

- ET

- Mase
> intermittent demand

- WMAPE

Can you expand this list ?

Thanks in advance

Oliver

[OliverA41]

furhtermore, you have also 2 results of two different Forecast Errors. e.g. MAPE and MASE.

What is the best choise.