Feedback arrives shaped by who bothered to give it. Loud users are not representative users, and the silent majority is silent in both directions.
Weight feedback by evidence of the behaviour behind it. A complaint from someone who uses the feature daily is worth more than a suggestion from someone who never has.