The most talked about EverNote numbers on freemium conversion shows, 3% paying customers and 97% free users (freeloaders and hopefully some Do-Gooders).
When a user signs up, if we use priors as an indication of posterior, the probability this user will upgrade to the paid version is 3% (will not upgrade is 97%).
Let us assume the Lifetime Value of a User is $250. Since all the infrastructure costs are sunk and the marginal cost to serve an additional user is $0, there is no cost in serving the free user.
So the expected Lifetime Value of this user is $7.5 (3% * $250). Since this a positive value, it would appear that there simply is no downside to sign up another user for free.
However, what about users who would have upgraded had it not been for the free version? This could be either due to the $0 reference price or the value they get from free is good enough (See: value step function). This is the opportunity cost.Let us assume the same conversion numbers and assume that an additional 3% would have upgraded had it not been for the free version. The value you would have gained is a forgone opportunity and hence coded as red. This brings the expected Lifetime Value to just $0.225. Model this for different forgone opportunities, does Free look attractive any more?
You could argue that not all of the 3% who pay now would not have upgraded had it not been for free.
You could argue that it isn’t another 3% that was forgone.
You are correct on both aspects. However, if you do not know how many will pay for your product or if your prospects do not value the product enough to pay for the value they get, isn’t that a bigger problem?
In a talk at Haas Alumni Luncheon Mr. Chris Anderson, the author of The Long Tail and Free, talked about his new book Free. First few minutes into the talk he talked about his previous work, The Long Tail, and how the marginal cost of shelf space. He said how physical goods (what he calls atoms) have a marginal cost to produce, store and sell and how information goods (what he calls bits) have declining marginal cost that approaches zero.
I will set aside my previous arguments on why costs do not matter with pricing and focus on just what is marginal cost. According to Mr. Anderson, the definition of marginal cost is simply the cost to produce, store and sell one widget. That is the right definition if there are no opportunity costs. In his example he talks about how there is a marginal shelf cost to selling physical goods through Wal Mart. Why is there a charge? Because Wal Mart is looking at opportunity cost of storing your wares vs. someone else wares. If you look closely all shelf costs are sunk for Wal Mart, but still it charges a marketer a fee because of this opportunity cost.
Marginal cost is not about the widget you sell, it is the “relevant cost of serving one additional customer”. It is the higher of either the cost to produce/store/sell the widget and the opportunity cost of serving that additional customer. The definition is easy to see for physical goods. For digital goods, for example music, the marginal cost is not zero (as Mr.Anderson says regarding digital music just because there are no CDs) but the revenue lost by selling this music vs. another or the future revenue lost by setting a very low reference price in the minds of customers.
Once you consider the opportunity costs you can see the deficiency in the definition and argument based on marginal cost being zero.