Understanding BEAM

Michael Feith · 2023-04-13T13:31:22+00:00

There was an error rendering this rich post.

Hi everyone,

I'm currently testing the Forecasting with BEAM, but I can't get it running.

In my first try, I had a cube with month, customer, product, scenario and version and created a second cube for beam with the same dimensions (Absatz & Beam - Absatz). Those two cubes are not in the screenshots, I just wanted to explain my approach.

It didn't work, so i simplified the cube to only have product and month and created another cube for beam with those dimensions (Absatz clean & BEAM - Absatz clean).

the two cubes two Absatz and Absatz clean have fictional sales data from january 21 to december 22. It has seasonal data with lows in march and max volumes in august, so a fairly standard model I would say.

However when I run the forecasting, it shows nothing and also doesnt process the time series.

Does anyone know why it doesn't work/what I did wrong? I had a look at previous conversations like this one, but it didn't solve the problem for me.

Kind regards and thanks for the help,

Michael

Find more posts tagged with

Accepted answers

Tommaso Riva

Hi Michael,

thanks for posting on Board communty.

Does it make any difference if you set the target cube dense, i.e. also set the product as dense?

If not, would it be possible to extend the time range of the observed data by including other observed months, if any?

To check whether the problem only occurs with "competition" method, would it be possible to check whether "Naive seasonal" method draws any figure?

Please advise.

Best Regards

Tommaso

All comments

Willians Santana

Hi, @Michael Feith . Thank you for your message and all the details provided! I do appreciate that.

I hope you are having a wonderful day!

Please, review the following BEAM training set with an example available for partners and customers and let me know if it can help you:
https://edu.board.com/board-enterprise-analytics-modeling-beam

We are all ears for you and will be available at your convenience.

Note: if you also want to send us the data load files and cube definitions, we can reproduce the same issue from our side, then let you know about the results.

Kind Regards,

You are very welcome!

Willians

Tommaso Riva

Hi Michael,

thanks for posting on Board communty.

Does it make any difference if you set the target cube dense, i.e. also set the product as dense?

If not, would it be possible to extend the time range of the observed data by including other observed months, if any?

To check whether the problem only occurs with "competition" method, would it be possible to check whether "Naive seasonal" method draws any figure?

Please advise.

Best Regards

Tommaso

Michael Feith

Hi, thank you for your answers!

Setting the target cube dense made the difference for the Absatz/Sales Volume cube with all 5 dimensions, thanks a lot @Tommaso Riva !

The Absatz clean with only two dimensions actually worked as sparse, I just had the wrong dimensions. Classic error of the person behind the computer, not the computer itself for this one :)

By any chance, is there an explanation behind when you have to use sparse/dense in Forecasting? Or is it just one of the things one knows through experience?

Best regards

Michael

Tommaso Riva

Hi Michael,

I'm glad it works now.

Before "Board 2021 Spring Release", the target cube in the BEAM scenario had to have a dense structure.

Starting with "Board 2021 Spring Release", with the adoption of the new engine based on the tuple definition concept, sparse entities are also supported in the target definition.

In general, we suggest to set the sparsity for the first test and, if this fails, to switch to the dense structure.

Best Regards,

Tommaso