Criteria of identity should mirror the identity relation in being reflexive, symmetrical, and transitive. However, this logical requirement is only rarely met by the criteria that we are most inclined to propose as candidates. The present paper addresses the question how such obvious candidates are best approximated by means of relations that have all of the aforementioned features, i.e., which are equivalence relations. This question divides into two more basic questions. First, what is to be considered a 'best' approximation. And second, how can these best approximations be found? In answering these questions, we both rely on and constructively criticize ground-breaking work done by Timothy Williamson. Guiding ideas of our approach are that we allow approximations by means of overlapping equivalence-relations, and that closeness of approximation is measured in terms of the number of mistakes made by the approximation when compared to the obvious candidate criterion.