This message was deleted.

Spurious thoughts;

because typically use multiple devices

GA4 Data Stream to bridge platforms merging over

user_id

&

pseudo_user_id

and our period of conversion is long enough

Ramp up conversion window

and we test usually features before login

match by

pseudo_user_id

?

There aren’t great ways to ensure consistent variations across devices without knowing who they are first. A few thoughts/ideas: 1. Bucket your users by IP, if they are using different devices on the same network. 2. GrowthBook will automatically remove double exposed users if you set the experiment to the user_id for your case (and make sure the join works between the anon id and user id on the data sources page) 3. Set up a simple caching/lookup API to assign users- but I’m not sure what other information is available here that would make this an improvement.