Comment by awillen

> Taken in isolation, the optimal amount is clearly zero.

The post makes it clear that the discussion is not about theory or taking anything in isolation - it's about fraud in the real world. In that context, the way it's stated is correct - if you have zero fraud in the real world, that means that you designed the tradeoffs wrong and that the cost of your fraud prevention (in terms of actual dollars as well as inconvenience to customers, etc.) is greater than the overall cost would be if you allowed a small amount of fraud to occur (looking at the total cost of that fraud as well as the cost of preventing additional fraud).

I suppose the problem is that whether or not the title of the post is true or not depends on the context in which it's taken, and the title itself doesn't have any context. Since the post does offer context, though, I think it's reasonable to take the title in that context.