Hmm, maybe I am simplifying this but can't you do:

Fraction that convert to paying * How much they per on average per month * Average months before cancellation * Some sort of Discounting for value of future money (unless you plan to increase price) = value per member

For avg months before cancellation count all those that are yet to cancel as 2x there lifetime (seems reasonable). Would give you a fairly reasonable guess. Then you just have to make sure that the value of a paid member is the same.

discounting: http://en.wikipedia.org/wiki/Discount

Take the average length of time that a user stays with your service.

Find out the average number of purchases they make.

Multiply to get average lifetime value.

Spend less than this number on your customer acquisition, ie Adsense.

Example: The average subscriber stays with you 1 year. It's a monthly service, so the average customer makes 12 purchases. You have a low/no cost service with a gross profit of $2/mo.

12 x $2 Profit per Unit = $24 LTV of the average customer

If you have some sort of viral component where users are getting other users, you'll want to account for that as well.

The one suggestion I would add is to segment your users in the model: flighty, regular, loyal (or something similar) Focus your analysis on clearly modeling usage by segment in the first few months to get a good feel for who is dropping, when, and why (good exercise). There is nothing wrong with modeling a 'loyal' segment that simply loves your service and never or rarely cancels, so long as you also model 'regular' and 'flighty' users.

The problem with these methods is that they assume you can figure out average months before cancellation.

If 10% of your members are still active (have never cancelled) since you launched, how can you estimate this? They may stay on for another week or another 5 years.

Clearly, this isn't something that can have a hard answer, but it would be nice to reduce the guesswork.