Statistical Forecast's illogical result (zero values)

[Former Member]

Dear community,

Would anybody have the pleasure to explain me why my forecast gives zero values when the adjusted actuals are not zero except for the last?

I use the double exponential smoothing method of which the parameters that I have set are demonstrated below. Normally the parameters that I have set shouldn't lead to a zero flat forecast .

* The sales forecast key figure defaults to the statistical forecast key figure unless manually overwritten

* Adjusted actuals quantity key figure defaults to actuals quantity unless manually overwritten. Besides, its values are copied to the stored key figure adjusted actuals quantity stored that is used as an input for the calculation of the statistical forecast

Thank you for your time

Best regards

Vasilis

[anna_linden]

Hi Vasilis,

the key figure that I see in your Excel screenshots which is zero is Sales Forecast Quantity. In you Forecast Model I see that your results should be saved in the Statistical Forecast Quantity KF. Is the latter one also zero then after you refresh your planning sheet? You state that they equal each other.

The model itself looks OK to me. Unless you have enough data in your system to cover the historical periods you should get some values. Did you try with another algorithm? Something like simple average?

Best Regards,

Anna

[Former Member]

Hi Anna,

Vasilis and me work on the same issue. Sales Forecast Quantity defaults to Statistical Forecast Quantity if no value in entered manually. Therefore, the statistical forecast KF also shows 0.

Kind regards,

Katrin

[anna_linden]

OK. Here are 2 more checks:

1) I assume you have enough sales history to cover your horizon?

2) Did you try with another algorithm, e.g. simple average? Does that provide any values?

3) is it the first time these key figures are filled in that system?

[Former Member]

Hi Anna,

thanks for the input

1: We have sales history for the past two years - starting from December 2015, that is why we defined the Historical Period Offset

2: Simple exponential smoothing for example works. Only Double & Triple Exponential smoothing result in 0 values

3: No, we already performed a lot of statistical forecasts

Can we provide you with any more data?

Thanks already for your time,

Katrin

[anna_linden]

Hi,

it would be intersting to see if the system is calculating something at all. Can you enable a forecast error measure (e.g. MASE) to be calculated and check in the log of the forecast run the results? Are they also zero?

Also, if you are in a 6.x system, can you run the Automated Exponential Smoothing with "Optimize Within Selected Algorithm" and then select the Double Exp Smoothing one?

In any case, it would make sense to open a support ticket here. It is rather unusual that one algorithm works and the other does not when having the exact same configuration.

BR Anna

[Former Member]

Hi Anna,

we just tested it and we have a MAPE (40% for Product 1 and 140% for Product 2). These measures also seem to be way too low. I guess we will have to open an OSS then, we are using IBP 5.0.

Thank you for your help,

Katrin