In my experience and that of many authors, users at bad at predicting the value of potential/future features. In my view this reduces the validity of applying the Kano model for prospective research.