I once worked on a large application whose features were driven largely by customers and it wasn't pretty. You easily end up with a bloat and 'death by preferences' [1]. So I agree, your product vision should drive your features.

What I didn't realize at the time (and I think many don't) is that when you have customer requests for features that don't align with your vision you should consider making your product extensible. Keep a small core product and provide plug-ins for additional features. Customers pick the plugins they need or write their own.

[1] http://insideintercom.io/product-strategy-means-saying-no/