There's a huge difference between understanding user needs, successfully shipping features, and having those features actually loved by users after launch.

Over the past 2-3 years, I've developed numerous new features for our internal customer profiling system. Yet more than half of them - like adding RFM models or algorithmic audience expansion - saw minimal adoption after being built. 

These seemingly sophisticated features, requested by headquarters operations teams, were barely used in practice. The requesters had naively assumed other teams would love them.