Indeed, the relevant number is not (students who complete) / (students who sign up). A better number to look at is (students who complete) / (cost of class) - i.e., an actual benefit/cost ratio. (It's not the best number, since it ignores the value gained by students who partially complete.)

If I pay $2,000 and have a 4% chance of completion, that sucks. If I pay $0 and have a 4% chance of completion, that's pure win.