This paper deals with the product mix problem using the concept of Theory Of Constraints (TOC). Theory of constraints is one of the most efficient approaches which have been applied to solve the product mix problem heuristically. Although there are numerous heuristic and meta-heuristics to solve this problem, finding the optimal solution in a reasonable time is still a challenging issue. In this paper, a novel procedure inspired by multi-agent decision making concepts, is developed to generate better initial solutions upon which the existing TOC-based product mix algorithms can reach solutions with better quality. The superiority of the proposed procedure is validated by two existing algorithms through a well-known problem instance in the body of literature.