I have a critical mass of paying users and I
have to disappoint a small portion of them.
I hope I'm not misrepresenting your comment because on the spectrum of problems afflicting early stage startups, them's the good kind.On a more serious note, we didn't have too much of a problem with it and we had to say "no" quite a few times. We'd just offer to refund their money and explain the situation the best we could.
I guess it's just another factor to consider. They both seem to be viable models. And I'm not experienced with either.
Free users were the opposite. They would be angry when things didn't work out quite right and complain about the beta price (roughly $5, though we moved it around some). Free also has more issues on the customer service side simply because you have to do more of it.
I can certainly think of scenarios where a free beta makes sense, but if I'm starting another B2B company I would charge from day one all over again.
I think for any product that results in a sale, ie not a social net or the like (not disparaging!), this is the right approach.
What happens when your learnings pivot the app in a direction that [some] paid users aren't interested in? They have bought access to a service which might not align to their needs anymore and they likely wouldn't have access to the version which they paid for. Sure, a dollar is a small amount, but I feel as though the effect is more pronounced than with free services. Heck, free services get flack a lot of the time for changing strategy or shutting down.
I guess if your feedback and learnings are from the vocal majority, you might underestimate the backlash from the silent real majority.
Furthermore, what happens when you have 10 paid signups (let's say $10 total) for a service that needs $100 worth of infrastructure? Do you end the service and refund the money? Do you pocket the money and move on? Do you fork up the remaining infrastructure cost from your pocket and continue? I thought the value of kickstarter and the MVP model was the validation to support the product, whereas this model seems to have a grey area in that regard.