It's a less error-prone approach than speculating on how they might improve in the future and using that to judge their present actions.