Boxed Ice, creators of Server Density, wrote up a nicely detailed post on how to run a successful beta program for a web application.
Knowing when to release a product is difficult. The maxim “release early, release often” certainly applies but that has to be balanced against making sure you have a minimum viable product and features that work well, with a minimum number of bugs. You also need to get feedback as early as possible to either completely change what you are going to do or introduce new feature you would never have thought of on your own.
As my company’s product, Investify, is now in a private beta, these tips gave me a lot to think about in structuring the remainder of the test period.
I guess the most important features are the first thing developers take care of. I'd recommend taking care of bugs though. If I work with beta that's so full of bugs it makes me want to kill the developer, I'm not likely to return for the full version. It's better to have couple features less but almost bug-free beta than the other way around.