Flag-Dependent Experiment Allocation

  • 28 March 2023
  • 1 reply

Userlevel 2

I have a question regarding experiment allocations.

We have two flags that we want to be dependent on each other. For example "For users who are shown this, give them another experiment".

For example using two flags, "show_tabs" and "show_new_shiny_tab". I would like to configure an experiment that is "For users who see the tabs, perform an A/B Test to users to test a new shiny tab". 

So to achieve behavior, we select on the "show_new_shiny_tab" experiment the following: 

"Users where show_tabs = true/false"


Will this work as expected? From what we see, there is room for doubt, but overall it looks like it is working. Is there any concerns with this approach? 


Best answer by Jeremie Gluckman 31 March 2023, 07:54

View original

1 reply

Userlevel 6
Badge +9

Hi @liukaichi I asked around within our team, and in general, the setup will work. 
The only thing that was called out is to be mindful of the timing of the assignment of the flags. For rule-based targeting based on a user property in Analytics, our cache doesn’t get invalidated when a user’s properties are changed in Analytics, so in the worst-case scenario, it’s possible that there is a max delay of 60 minutes when targeting existing user properties. If this time lag matters for your setup and you need additional clarification, please let me know and I'm happy to pass this to our specialists or Product team to clarify!