I have to applaud the Jawbone ICON people for setting up a feedback mechanism for their online MyTalk service. It's fairly cool! It lets you change the voice (from "The Hero" to "The Catch!") and install one of three minor apps on the headset. 
At the conclusion of the survey I get an opportunity to win--wait for it--a free Jawbone ICON. That is, I'm taking a survey for a service I use with my Jawbone ICON and I'm entered into a drawing to win... a Jawbone ICON. Hmmm. Something's wrong here.
But that isn't really my point.
After asking me a few fairly interesting questions about the application, the last question is:
Someone must've written a book on surveys that says to always end with an open-ended question. But "What other products should we develop?" is exactly the wrong question. I doubt they get any valuable insights from this. A better question is "What other problems do you encounter?" but even that I wouldn't have been able to answer. Maybe "What else would you like us to know?" is a good ending. That is, leave space for comments of any type: satisfaction with the device, the service, the survey.
You don't discover problems in a survey; you validate and quantify what you've learned from personal experiences and interviews.
Before a survey, someone in product management needs to spend some time observing problems.
