Andrey makes some very good points. With Rallypoint what we are trying to do (and is working out so far) is not launching fully public straight from the get go, but slowly letting people in.
Even though we have quite a bit of experience launching / running apps for clients, we have never really done it on our own. We didn’t know how our architecture would scale, how many of what service we’d need, how many bugs were lying dormant, etc. On the other hand, we do know what standards we want to hit, how we want our users to experience our app, what our levels of support are, etc.
So we decided that once our app was more useful than broken, we started letting people in (which we’ve just started doing a couple of weeks ago). And actually before we let people in, we used the app ourselves quite intensively.
Now with a few active users we’ve got more of an idea how our app will perform and if it has any crazy issues. In the worst case scenario, if anything really hits the fan, any big problems only affect a couple of people, who already know and accepted that with early access come a few risks.
Once we’re happy with how it runs and with ourselves reaching our standards, we’re letting more people in, with hopefully some informed confidence that we won’t completely crash and burn instantly. And we’l keep going like that. One of our milestones will be starting charging for it and another instant access for the public.
This approach is surely making it easier for me to relax