Forum Discussion

Evgeniia's avatar
Evgeniia
Active Member
4 months ago

Control group based on custom attribute

Hey everyone! I am wondering if anyone had a similar situation: standard tool for isolating control group (global or Canvas- specific) does not fit my business needs, because I want to measure incremental change of product conversions per B2B subscriber, and each B2B subscriber can have more than 1 employee (aka user ID in Braze). Therefore, there is a risk of having representatives of the same B2B entity in both control and treatment groups. How would you deal with this issue? I have an idea of assigning custom attribute (control_group = true/false) when importing each new user (so that all users within one B2B entity fall into either control or treatment group), but it makes managing campaigns and building reports more complicated.

I'll be grateful for your ideas and sharing your experience. Thank you! :) 

  • bob's avatar
    bob
    Specialist

    This could, depending on how many subscribers you have and how many employees in each, easily unbalance a control group.

    That said, if you already had a subscriber identifier as a custom attribute on each record, you could define a control group based on that.

    If necessary, you could hash that value to a single numerical value (like the random bucket number) and work with that. Even just a 0-99 number would let you easily carve out any reasonable control.

    If you didn't already have a value, that hash could be done easily enough in liquid - perhaps triggered at point of user creation (or whenever the subscriber values are set).

  • My approach is to have a array custom attribute where I insert the typology of audience control is it. So for instance could be control zero: all product, product1, product2

    When segmenting I have a segment call control group that I use to across ALL my canvas. In this way is easier to update at scale if need to suppress certain attributes.