I generally do a preview release for techie friends, but I've never had a formal beta for BCC or AR. If the software is stable enough to provide value, then I release it and start charging money. You get much better feedback when someone is invested in your software and wants to use it to get their work done, versus someone who is not invested in your software and for whom using it is uncompensated work.
(My conscience is totally clear about doing this: I give refunds for any reason as a matter of policy, and if people didn't sell buggy software then there would be no software on the market. In return for higher than average suffering compared to later users, they get me at their beck and call for the first few weeks. In the very unlikely event that AR failed to phone the customers' customers, I am more than capable of calling Lucinda Smith to tell her that her appointment with Mindy's Hair Salon is on Tuesday at 3 PM.)
BCC 1.0 was terrible software. Somebody bought it anyhow. (He got a refund.) 1.02 was heavily influenced by his comments, and sucked a little less.
AR 1.0 is not terrible. It is bare-bones, and it certainly isn't right for all customers, but it does what it says on the tin. It will probably improve fairly rapidly with customer feedback.
I get users for software four weeks old the same way I get users for software four years old: organic search. (There is a reason I got AR's website up and some links pointed at it six months before I actually had software to sell: I now get a couple hundred searches a month "for free", which gives me a small trickle of brave souls to try it out on.)
