Is there a way to randomly, but consistently, divide users into segments, say by id, which can then be used to ensure that different experiments are applied to different groups of users? As a trivial example, to divide users into two groups. Group 1 would then be used for an A/A experiment, and Group 2 would have a new feature rolled out at 50% and be used for an A/B experiment. Looking through the docs I haven’t seen anything that does exactly this?

I believe you're looking for Namespaces, though I don't see specifics on what field it's using to assign users

That seems likely, but I can’t see enough info in the docs to find and use namespaces? I can create namespaces but I don’t see how you activate or use them? I can’t seem to access the controls shown in the doc you linked to.

You can active the namespaces within the experiment over ride rule in feature flags

Thanks…I’m a bit confused, how does an A/B experiment override rule relate to an experiment? And would this put half of all users into the ‘blue’ namespace? And half of those users would see the new feature?

it takes your 100% of traffic and allows you to allocate that audience as you see fit to experiments to ensure user cohorts don’t overlap

Ah, so you only need one namespace to set up two experiments? One would use the 0 0.5 range, the other the 0.5 - 1 range?

you can do that, yes

Interesting, what would be the use case for overlapping?

you can use that for holdouts - to keep a group of users unexposed to any experiments