The complete and efficient solution to the enumeration of candidate compounds and mixtures that meet specified consumer attributes is often a difficult mathematical programming problem. Most approaches to this problem involve the solution of a mixed integer non-linear program (MINLP) which may achieve only local optima solutions. In this paper a proof-of-concept study is presented to show that empirical models can be used in a reverse problem formulation to ensure a complete set of candidate compounds and mixtures are found subject to the predictive power of the model. The method utilizes a transformation of consumer attributes to properties described by the group contribution method and solves the reverse problem formulation using the property clustering technique. A case study in refrigerant design is used to highlight the method. (C) 2008 Elsevier Ltd. All rights reserved.