I think more than economics it's statistics that the credit card companies struggle with. Many people assume correlated data has causation and make massive blunders. Here is an example.

CC company X determines that they make about $650 per cardmember per year based on their historical data. They want to make $15,000,000 more this year so they decide to get 100,000 more customers at an 'acquisition' cost of $500 each. since they will have 100,000 * (650-100) or $15,000,000 more in profit.

The problem is that the correlation between customers and profic does not dictate that one will cause the other. So when a bunch of us take those deals and then bolt they are screwed. Not all customers are the same value and averaging what you have doesn't mean your new customers will act the same. Especially when you used outside forces to acquire them.