In the globalization process, postponement is still an effective strategy to reduce supply chain risks. Existing researches on postponement are often based on a fixed product family architecture which is given at the outset, with limited attention to the inherent coupling between product family design and postponement manufacturing process decision. Thus, a hierarchical optimization approach is formulated. By constructing a hierarchical interactive evaluation mechanism, a nonlinear bi-level programming model is established. The upper level designs the architecture of the product family and decides the type(s) of postponed product module(s) to maximize the customer utility per cost. The lower level decides the optimal manufacturing mode for each non-postponed and postponed product module, and the optimal assembly mode for each product variant to minimize the engineering cost. A nested genetic algorithm is designed to solve the optimization model. A case study of smart refrigerator product family postponement manufacturing is introduced to illustrate the feasibilities of the proposed model and algorithm. Finally, a nested GAPSO algorithm is proposed to improve the nested genetic algorithm. The calculation processes and results of the two algorithms are compared and analyzed.