IMO this reeks of data-driven myopia. Slow bleeds to goodwill are incredibly difficult to properly capture into a model. Anything that isn't measured - or poorly measured - is liable to be traded off for what is measured.