Hi, is there a good way to conduct dimensional analysis when users can take on multiple values of the dimension? For example, the dimension is device (mobile or desktop) and some users have activity on both devices. We'd like to look at metrics by device for each user's behavior on only that device. (cc @ambitious-apartment-58735)

That's not easy to do right now. You would basically have to make separate metrics for mobile/desktop. Being able to apply dimensions to metrics is an interesting idea though. I'll make an issue for that on GitHub to track it.

Thanks for the info! To make sure I understand, a Dimension (whether defined within the Experiment Assignment Query or not) can only take on one value for every value of the Identifier?

Not exactly. If a single user has 2 rows in the experiment assignment table, one mobile and one desktop, then they will be included twice in the dimension breakdown, once for each dimension value. So the total number of users if you add up the different dimensions will be greater than the total number of users on the overall results (no dimension selected). However, the dimension doesn't flow through to metrics currently, which I think is what you wanted. So both times the user is included in the report, it will see all of their conversions. If they converted on mobile, it will also count as a conversion in the desktop dimension.

Yup, that’s the issue. Thanks for clarifying.

Got it, thanks! Is there a link to the Git issue you can share?