Inclined to agree that data are essentially constructed, but enough of an empiricist (and realist) to want to believe that the the fields of our theories are constrained at an intersection at some kind of boundary where they correspond with the world we're trying to describe and intervene upon.