Having users that want to return is objectively good. We've just taken this to the wrong conclusions. I strongly suspect that this has to do with attribution a

Having users that want to return is objectively good. We've just taken this to the wrong conclusions. I strongly suspect that this has to do with attribution and shallow "data driven" choices - there is a heavy bias towards things that are easier to measure. Especially in startups with limited resources. Measuring retained users regardless of anonymous/known state is very hard and lossy. So the nearest easy to measure analog that is often picked is logged-in users and we start the fucked up cycle you mention. Similar things happen around all metrics that have to do with retention. We know that a better experience through higher performance and better UX is likely to create happier users that return... but we cannot prove it (easily) so we focus on optimizing other things.