What are the lessons for those concerned with robust evaluations?