Mass customization, which aims at satisfying individual customer needs with near mass production efficiency, has become a major trend in industry. Adopting the mass customization paradigm, customer preferences have a significant impact on the product design process. Thus, it is important for companies to make proper decisions in translating the voice of customers to product specifications. To facilitate this process, a learning-based hybrid method named KBANN-DT is proposed, which combines knowledge-based artificial neural network (KBANN) and CART decision tree (DT). In this method, the KBANN algorithm is applied to modeling the relationship between customer needs and product specifications. With prior domain theory, KBANN can provide a high generalization performance even if the data set is small. Based on the trained KBANN network, the CART DT algorithm is employed to extract rules from it. To illustrate the effectiveness of the proposed method, a case study in an elevator company is reported. The results show that the proposed method can be a promising tool for product definition.