Using different audiences with mutually exclusive experiments
Hi, I'm wondering if its necessary to use the same audience targeting for two mutually exclusive experiments. I see the helpful tip (i.e., "Use the same url targeting and audiences in experiments that are in the same group.") in the documentation about mutually exclusive experiments, but I'm wondering if that's required, and if not, what is the practical effect of having two mutually exclusive experiments with different audiences?
For example, say I have mutually exclusive experiments A and B (both modifying a part of my site's registration flow). I want A to be seen by 10% of mobile traffic only, whereas I want B to be seen by 10% of overall (i.e., mobile + tablet + desktop) traffic. What is the best way to accomplish this? My idea was to create mutually exclusive experiments, and restrict A to a "mobile-only" audience and keep B for the general audience - would this accomplish my goal?
Here's essentially the most important information from that documentation:
This code allows you to bucket users into different experiments that would otherwise all run simultaneously. Implementing mutually exclusive code does REDUCE the amount of users exposed to each experiment tagged with a mutually exclusive tag.
You can set up epxeriments with different targeting conditions, but this setup is typically used to ensure that a visitor who would normally fit the same criteria for both experiments (fits the audience condition, is on the targeted page, etc) only sees one. If both experiments were targeting different pages or different types of visitors, there would be no needto make them mutually exclusive as there would be no overlap.
Technical Support Engineer
I think you can do this without complicating it with Mutually exclusive experiments.
- Create Experiment A > Use Audience to target MOBILE PHONE (You can do it from Devices ) > set % traffic allocation to 'new' variant A at 10% and Original at 90%.
- Create Experiment B > Use Audience 'Everyone' > set % traffic allocation to your 'new' variant at 10% Original at 90%
Hope that helps
Thanks for the response, Robert! One question though - wouldn't this approach mean that a user could be allocated to the test variant for both experiments (i.e., be allocated to both A and B, since A's audience target is a subset of B's) and that each time this user would enter the registration flow, there could be no guarantee of which experiment she would see (since both experiments modify the same page)? I was just worried that without using mutually-exclusive experiments, I would have some potential overlap between the two experiments' test variants that would muddy up my results.