Users demand delightful experiences (see: Pinterest, Square, Path), but at what point does polish get in the way of finding product market fit?
The first versions of Facebook and Twitter weren't pretty and users flows probably weren't great, yet their products were inherently compelling. There was an underlying addictiveness to the experiences they offered their users.
So when should a startup ship their product? If the fully-realized app has four main components, should the startup only build one or two of those components with the finest attention to detail. Or should they bust out all four in a semi-embarrassing way?