But indeed, I would be willing to ???reduce risk at the expense of reducing the quality of the model??? ??? especially in marketing, and especially if quality implies trying to nail down every last variable using a faulty dataset in the pursuit of being ???accurate???.