Wouldn't a better formula be the expected value of the outcome: utility delta * #people * probability of success?