I have long been an naysayer about the "free" model of technology, preferring to PAY, rather than use a free service. Beyond the ability to bitch for not getting what you pay for, I've often found the "free" (i.e. we're burning up our investors money) model as one that is not a direct sibling of "sampling" something I consider a proven marketing concept and one that does indeed accomplish the three basic tenants of promotion:
1. Stimulate Trial
2. Drive repeat purchase
3. Promote brand switching
You see, as Om points out in his Cricket League Fail post about the IPL, they went with a model that was both free and underpowered.
This is not a surprise, nor in the purest sense of the concept of broadcasting, ad supported would not be purely free. But we're in an era where supply (server drain, bandwidth capacity, access devices and clients) are at levels never before seen, and despite all the modeling that can be done to predict if startup x has enough of everything in really still an art form, not a science. If it was a science, would we see Twitter outages? Would Amazon blink from time to time or would Google Mail ever go down?
The "free" model is limitless, and in turn the free users tend to only hang around as long as someone else is paying the freight. It also leads to faulty projections of users. What used to be 4-6 percent of conversion in media free to paid sites, is now in the one percent range. With telecom services the range is even less. Offer someone free minutes and they'll sign up. Ask them to pay, and they'll go elsewhere. Same with video.
So, if you have a free service that wants to go paid, my advise is simply this. Start with a paid model, offer samples. Time limits and tracking. Once the sample period has expired, survey and monitor the users behavior and attitude towards your service. The real stat to look at isn't users, it's conversion. The ones that convert are your customers. Everyone else is just a freeloader.