Features with a low relative importance across your total userbase but a high score amongst a segment of customers that can be clearly identified. Perceived importance of a feature or relative importance of a customer need/goal can quite easily be measured with a forced-choice survey method like conjoint or maxdiff analysis. I generally find these methods to be a bit more applicable than a willingness-to-pay survey where you have to measure the value of each feature separately. As @Luc van der Tuin points out, that is with a caveat of dependent vs independent functionality, although more functionality is independent than product teams tend to initially presume.