I agree that you can get momentary insight into changes, but that doesn't help you understand if you are indeed alienating potential users, for two combined reasons.
First, you can't measure the preferences of those who never were your users, or those who already dropped out.
Second, software is not static, and whenever you apply changes to it - including those not related to UI - its utility to different demographics changes. And you can't know how much of the new demographic would be your users if your software had a diffrent UI.
So while you can measure the impact of UI changes on your userbase via behaviour analysis, you can't find out whether you're excluding potential users this way.
First, you can't measure the preferences of those who never were your users, or those who already dropped out.
Second, software is not static, and whenever you apply changes to it - including those not related to UI - its utility to different demographics changes. And you can't know how much of the new demographic would be your users if your software had a diffrent UI.
So while you can measure the impact of UI changes on your userbase via behaviour analysis, you can't find out whether you're excluding potential users this way.