Honestly, I don't know. But from what I can tell, the ones solving problems people have, even if they were not aware before, are the ones working. The other ideas usually don't. What are problems worth solving?
I'd define a problem worth solving as something that has 3 attributes: It is persistent, in a sense that it reoccurs too many times, that it has an impact on some activity or workflow. That is something worth resolving. Another point is the "factor of annoyance". Remember the Microsoft's Paperclip? High factor of annoyance. One dialog asking for a bit of love in the form of a ok-button? Not that annoying. But keep in mind, that's all a very subjective rating. So if you find Dialog Boxes more annoying than speaking Paperclips, just change the order accordingly.
The third factor is the most important one: the Time spent working around. The attributes mentioned above may be pointless if it takes no time to work around the issue. If something takes a serious amount of time to be worked around, solved, corrected or understood, it's time to do something. Recently, GMail and my Mac synced. The result was a Address Book disaster. Time to correct it: 3 hours. That's a problem.
Anything that's built to return money should solve a problem or improve something. Anything else? Naaa..That's my point. What is yours?