Uneven Distribution in Marketo Random Sample Flow

Raul88SD
Level 1

Uneven Distribution in Marketo Random Sample Flow

Hello,

I'm seeing some unexpected results in my Marketo Smart Campaign flow. Here's how it's set up:

  • Step 1: Send Email
    • Choice 1: IF condition is "X" then Send Email 1
    • Choice 2: IF Random Sample is "50" then Send Email 2
    • Default Choice: Send Email 3

After running the campaign, the email send counts are:

  • Email 1: 900 records
  • Email 2: 700 records
  • Email 3: 250 records

I expected Email 2 and Email 3 to have roughly the same number of sends because I used a 50% random sample split. Can anyone explain why I'm seeing this uneven distribution between Email 2 and Email 3?

 

2 REPLIES 2
Agrawal_Kirti
Level 2

Re: Uneven Distribution in Marketo Random Sample Flow

Hi @Raul88SD ,

Can you please confirm your target list count? Was it more than 1850?

If yes, can you tell me the count and how many were blocked when you sent the email?

 

If above is the condition, then I believe it randomly picked 50% records and then chose to send email to whoever is eligible and left the rest of the records. I have also faced this situation, but in my case, mostly records were blocked (email invalid), and all records received email 2 (default).

Regards,
KIRTI
SanfordWhiteman
Level 10 - Community Moderator

Re: Uneven Distribution in Marketo Random Sample Flow

I think you’re misunderstanding how Random Sample works. Bottom line: you shouldn’t combine Random Sample with other choice types in the same flow step.

 

Random Sample isn’t taking 50% of the people who didn’t match Choice 1. It’s taking 50% of the people who run through the flow step.*

 

So you have overlapping cohorts. The if condition of Choice 1 can include people who otherwise would be part of the 50% group. But of course if they match Choice 1, they cannot match Choice 2.

 

If the design is to...

 

1. send to a cohort matching a certain condition

2. split the remainder 50/50

 

... then remove the first cohort from the flow and use a different flow step. Even better is a separate campaign, as this kind of setup with a single campaign can be very hard to debug.

 

 

* actually more complex than that, but close enough for this discussion