Personally I think the data is of limited use even for benchmarking but, eh, your funeral. Users of my downloadable app convert to the paid version at about 1.8%~2.2% most months. There is a bit of guesswork involved there as counting downloads is not quite as reliable as "select count(*) from accounts;" .
Overall, Freemium conversion rates are way down right now. I might suspect that users are feeling less wealthy, and not opting to pay for anything that's not absolutely necessary.
Comparing virtual goods sales to premium web app conversions is a bit of a stretch, but I happen to like Nabeel, and I think that link is a good article.
I can absolutely see your point, but I tend to see them as two branches of one tree- You're inviting people into your service for free, and then trying to sell them a higher quality version.
The difference is that one transaction is a one-off sale, and the other is a subscription.
It's more then that though. People subscribe to web applications for completely different reasons than people whom purchase virtual goods. The demographics, spending patterns, motivations to spend, lifetime value, upsell opportunities and a million other things of the customers differ. Comparing money spent on entertainment and money spent on tools isn't a good idea.
You can't really compare yourself to those guys. Sure they might have ~2% conversion rates...but thats because their users trust that they pay for quality, You on the other hand will be an unknown, so they'll be more hesitant to give up the credit card info. To be on the safe side, put your conversion rate at .5%.
My opinion: it depends on a lot of factors. From my experience without much effort you can get 1% users convert. But like I said there a lot of variables that can mess it up or down.
Personally I think the data is of limited use even for benchmarking but, eh, your funeral. Users of my downloadable app convert to the paid version at about 1.8%~2.2% most months. There is a bit of guesswork involved there as counting downloads is not quite as reliable as "select count(*) from accounts;" .