The audience was not equally divided by the random split.

[MuhammedSuhail]

I have a scenario, where I tried to segregate people into 3 different static lists based on below criteria:

Add to list:

Choice 1:

If 'preferred Language Segment' is 'Bahasa'-
Then add to 'Test List 1'

Choice 2:
If 'Random Sample' is '50'
Then add to 'Split list 1'

Default choice :
Add to 'Split list 2'

Total audience count is 51488.
Total people with Bahasa Preferred language segment from above audience is 27039.

We expected rest people (24,449) would be equally split (12,224 &12,225) between remaining two lists (Split list 1 & 2), but as you can see below,  the vast majority of people were added to 'Split list 1'.
Split list 1 - 21922
Split list 2 - 2527

Can you please someone help me understand why choice 2 and default conditions weren't split equally?

If we use Random sample as choice 2, will it consider only remaining people from choice 1 or entire audience who qualified?

Test List 1 - 27039
Split list 1 - 21922
Split list 2 - 2527

[Michael_Florin]

I rebuilt your setup internally, and for me it works exactly as expected. I was wondering if this was due to having the Random Sample flow step combined with something else - which I usually don't do - but it works fine.

So I would assume you have outside effects changing your results. I'd recommend to try again with a fresh scenario.

----

EDIT: Wait a second. I tested again and this time with a segmentation filter in the first choice. And in fact I got incorrect results as well. So now, I'd recommend to not use the "member of segment" filter in Flow Step 1, but instead the filter that defines that segment. Most likely that is "Language = Bahasa". That will work.

I don't have an explanation as to why that segment in a filter wouldn't work. Probably a case for Marketo Support, if no one else has a better understanding.

[SanfordWhiteman]

@Michael_Florin afaik there’s no situation where mixing Random Sample and other choices work as OP expects. Filtering on a field value and filtering on a segment — which is ultimately a field, after all — should have the same result.

That is, as noted in the other thread, while multiple Random Sample choices do cooperate (so 25 followed by 25 will be 2 different 25% cohorts) other choices are not taken into account. Random Sample considers everyone who enters the flow step when calculating the %.

[Miray · Accepted solution]

Hi @MuhammedSuhail 

I noticed a very similar question was asked a few days ago, and you might find it helpful to check out this " Uneven Distribution in Marketo Random Sample Flow  ". One important point to note is that you should not mix the Random Sample method with other options. The link also includes a great explanation on how Random Sampling works.

[SanfordWhiteman]

@MuhammedSuhail please read the thread linked by Miray. And also make sure to search the community before posting, as this was asked literally within the past week!

[MuhammedSuhail]

Thank you all!. I read that discussion @SanfordWhiteman , I wonder how Marketo calculates default audience in this scenario.

Though Random sample takes 50% of all audience qualified for this flow step and remove anyone who were part of choice 1, how will Marketo split leftover audience among choice 2 and default?

[SanfordWhiteman]

The default cohort is everyone who’s left over after the other choices.

Doesn’t matter what the other choices were, it’s just everyone who didn’t match.