In recent years, product customization is getting considerable amount of interest from both academia and industry. It is imperative for many companies to survive in the increasingly diversified, fragmented, and competitive global marketplace. In this new paradigm, customers are no longer passive recipients of products or services that are designed and produced for a nominal customer. Instead, each customer has his individual identity and provides key inputs to design, producing, and delivering of the product or service based on his individual preferences. Nonetheless, product customization faces many challenges. It requires product development team to capture customer preferences and transform them into tangible product specifications effectively and efficiently to meet the continuing q...[ Read more ]

In recent years, product customization is getting considerable amount of interest from both academia and industry. It is imperative for many companies to survive in the increasingly diversified, fragmented, and competitive global marketplace. In this new paradigm, customers are no longer passive recipients of products or services that are designed and produced for a nominal customer. Instead, each customer has his individual identity and provides key inputs to design, producing, and delivering of the product or service based on his individual preferences. Nonetheless, product customization faces many challenges. It requires product development team to capture customer preferences and transform them into tangible product specifications effectively and efficiently to meet the continuing quest of reducing time to market. In addition, the choices of products offered in this paradigm also increases dramatically. Customers can easily be overwhelmed by the vast variety of products and the wide assortment of options. It is necessary to assist customers to become informed about choices and make the right decisions without the pressure of information overload.

This thesis aims at developing a theoretical foundation and methodology for product customization and in the meantime overcoming the above-mentioned challenges. A critical corner stone of this research is to represent customer preferences in an appropriate form and incorporate them into product design process. Because customer preferences are often related to other factors such as relevant product attributes and external parameters, conventional methods in deterministic form fall short of providing adequate support to represent and manipulate the uncertain nature of customer preferences. Thus in this thesis Bayesian network is deployed to represent customer preferences information. It is a probabilistic approach to capture diverse customer preferences quantitatively. In addition, customer preferences knowledge can be incrementally accumulated during the specification definition process. A specification definition environment is created which can acquire the most information about a customer’s preferences during each communication round by using information gain as the item selection criterion. As a result the most uncertainty and redundancy are eliminated. Since it is a sequential decision making process, a customized query sequence which is adaptive to the customer’s partial specification will be developed.

A recommendation module is also presented after the specification definition part to allow for configuration converging to customers’ target products quickly. Probabilistic relevance model is developed to calculate the probability that each product meets an active customer’s needs based on the incomplete, ambiguous, and even inconsistent specifications. Probability ranking principle is exploited to sort potentially satisfactory products based on the probability of relevance. Analytical results show that the proposed approach outperforms other methods in terms of specification definition efficiency and recommendation accuracy. Experiments and numerical example are conducted to test the viability of the methods.

This research makes interdisciplinary contribution by integrating techniques in artificial intelligence, information theory and information retrieval to solve issues in product customization. It provides a theoretical foundation for the integration of marketing, design and manufacturing to improve overall design chain responsiveness and efficiency from both an academic and industrial perspective. The developed approach explores the issue of incorporating customer preferences and matching customer needs in the product design process. In a longer perspective, it is expected that the result can provide a new angle to advance the product design methodology in general.