For a mechanism designer with an objective such as welfare we propose a method for robustly ranking mechanisms. The method is based on eliminating weakly dominated strategies only, and thus does not require any assumptions about agents' beliefs about each other except full support. We illustrate...