The more general version of your argument makes more sense, but I still think you are taking for granted that you can know your users that well, and how much "y

The more general version of your argument makes more sense, but I still think you are taking for granted that you can know your users that well, and how much "your user is different than my user" really applies. You don't seem to account for the magnitude of how positively/negatively the user is responding. If 70% of your users rate it a +2 on a scale of -5 to +5, and 30% rate it -4, that would seem to meet your scenario, but is that really a good goal, to be barely tolerable to a "mere" 30%? Maybe I'm being fallacious here since my own argument is that you won't have data that good, but I mean it as an analogy for why I disagree with the idea of knowingly pushing a design that won't work for some users. Around here at least, it's weird to bake the assumption of a non-large product into your design philosophy. You don't want to have a large, diverse userbase? Is your design actually defining who your userbase is, rather than the other way around? If you want that, we can't stop you, but it seems strange to us.