This paper presents a criterion for dividing observable firms into industries. The previous literature has failed to recognize how important is the lack of a principle for forming industries. Not only does the lack of an industry criterion lead to bad antitrust decisions, but it brings into question the logic behind applying standard economic analyses to observable groups of sellers. Economic theory traditionally has constructed models of markets composed of sellers of a homogeneous commodity. In the application of theories to observable sellers whose production processes and products are rarely precisely identical, the easy and convenient assumption that all sellers are identical is violated. One of the lacks of the theoretical literature is that it has given no guidance on how to put existing producers together to form the markets and industries to which economic modellers make reference. Instead, applied economists have tended to follow the pragmatic advice that violations of assumptions do not matter so long the firms behave approximately the assumptions of the model were satisfied[8]. Logic requires, however, that this as if methodology be applied only to firms which have been selected independently of the model being used. For example, to find the bounds of a competitive industry by finding the group of firms whose behavior is predicted by the competitive model and then to justify the use of the model on the grounds that the model predicts well is tautological. The logic of analyzing industry behavior in contradiction to the assumptions of the model is defensible only there exists a procedure for forming an industry that is independent of the criterion used to select the model to analyze it. This paper provides such a procedure. The industry bounding procedure outlined in this paper may be one of many, and others may be found in the future which have more desirable properties than the one presented here. The main contribution of this paper is to demonstrate that there is at least one definition which can be used to divide existing firms into industries, and that the industry so defined is unique for each firm. The lack of a criterion for defining industries is especially noticeable in antitrust analyses where the question posed is whether a particular firm is a monopoly, not whether it behaves it were one. The current procedure in such cases is ad hoc in the true sense of the term. Every situation is taken unique and while the economic witness will sometimes apologize about measurement problems, a list of industry members is produced for use in court in what Stigler has called an impudent exercise in economic gerrymandering [20,8].