"Build it and they will come" seems to be true in relation to Second Life. The problem is that 90% of people who register don't stay. They leave within the first three months. It seems obvious to me that the one primary reason for the astronomic departure rate is that most people don't find something worth doing. Right now, finding something interesting enough to make it beyond the initial learning curve is left up to chance. And the odds seem to be about 9-1 against.
So I offer a relatively simple solution. Treat Second Life nubes like conference attendees. When they sign-up, have them register for a specific track. Then provide a series of classes, self-guided courses and resources to lead them from neophyte to journeyman.
The chart below shows just some top-of-mind category idea for the tracks, along with one level of detail for a few of them to illustrate the concept.
So if someone is interested in building, have them start with building curriculum right after they are registered. Start with a small project they can build in less than an hour. And provide a place for them to display their work. Show them how to do a screen capture or video capture to share their effort with friends outside Second Life. And then offer additional education and projects so that they don't need to figure out what to do until they've built up the skills to start working independently.
This same concept holds true for more recreational tracks such as sightseeing and art viewing. Virtual art appreciation classes and self guided tours would be 101 level courses, followed by short tours of specific art styles or artists, lectures, exhibition trips, etc.
Outreach to physical world artists and musicians is another promising area. Why not advertise to them and offer education to help extend their work to virtual worlds?
Anyway, that's my lightbulb idea for this morning. Of course, there is a great deal of thought that would need to go into the structure of classes, curriculum, etc. But I think it would be a relatively modest investment with huge potential returns. Just bumping retention up 10% would have an enormous cumulative impact.