How can we achieve a uniform distribution of the contacts in the random splitting segments?

[alex_gaevoy]

Hi, Experts!

We have request from business to split segment randomly with certain conditions in SAP Marketing Cloud 2008.

It found out segments that we received after standard random splitting not equal by quantity contact which:

Use certain communication channel (e.g. one group have more contact who prefer sms than another)

By spends on goods (e.g. contacts from one group have average month order sum more than this parameter in another group).

So, how can we achieve a uniform distribution of the contacts in the random splitting segments by the certain parameters?

How you have solved such issue?

Thanks for your advice and help!

[former_member226]

Hello,

As the name suggests "Random", hence there is no uniform splitting of the data when you really use random filtering. In this case, system will randomly select a subset of data irrespective of data distribution and create a new segment. There is no way to influence this behavior of standard.

In case you are really looking to uniformly distribute the data then, you should try to, at first, create & distribute various segments using random split like communication channel or spend on good and then finally use the "merge" option to cumulate everything into one.

merge.png

Thanks

Saurabh

[alex_gaevoy]

Saurabh, thanks for your advice!

This is a excellent approach for binary parameters, e.g. gender, using certain communication channel etc. I added a pic. with example.

But how can we work with parameters with a value from a range, for example, avg sum from orders?

[former_member226]

Hello Oleksandr,

Hmmm got it. I think then there is no direct way to deal with such a case where you want to distribute even the sum of order or any other measure(as there could be n number of possibilities or set). Maybe you need to check with your customer and confirmrm if they can pre-define a certain "band/range" for each measure based field and then you can just follow the same approach as you did for other attributes. Unfortunately, I can't think of any other way to get around right now. 😞