This message was deleted.

First, we apply namespaces, then we apply reduced traffic coverage on the remaining traffic. So there are 2 ways to accomplish what you want. First is to only use 20% of the namespace (instead of 50%) and don't reduce the traffic percent at all. Second is to do 50% of the namespace and 40% traffic percent.

It's a bit different though isn't it? Namespaces exclude only traffic from another feature and the traffic restriction applies to total traffic? Or am I misunderstanding this?

Yes, they are slightly different, but the end result of both approaches is that you will get 20% of your overall traffic included in the experiment and none of that 20% will be part of the other experiment in the namespace.

Thanks 🙂