Product platform concepts are often deployed to achieve product variety and hence effective product customization. One of the popular methods to achieve product variety is to scale one or more design variables called the scaling variable(s). This necessitates efficient methods for identifying the values for scaling variables. This paper presents a graph-based optimization method called Platform Ant Colony Optimization (PACO) for identifying the values of the scaling variable(s) for platform formation. In PACO, the overall decision is a function of the cumulative decisions of simple computing agents called the 'ants.' The method employs an autocatalytic mechanism using a probabilistic search to improve the solution iteratively. We use a universal electric motor example cited in the literature to test the efficiency of the proposed method. Simulation results on the example problem indicate that the PACO method produces promising results.