بهینه سازی انتخاب مشتری برای محصول با قابلیت تنظیم در کاربرد تجارت الکترونیک B2C

Many companies provide configurable products on Internet to satisfy customers’ diversified requirements. Most of business-to-consumer (B2C) e-commerce software systems use tree- or wizard-like approaches to guide customers in configuring a customized product on Internet web pages. However, customers may feel confused while they are selecting components of a product from option lists, since they are usually not familiar with the technical details of these components. A few e-commerce sites use recommendation systems to provide suggested products for customers, but they have to maintain user profiles and have limitations such as new user problem and complexity. Therefore, they may not be suitable for small and medium-sized enterprises. This research proposes a new approach to help customers configure their expected products. By using this approach, once a customer inputs the levels of importance of requirements, total budget of the expected product, the software system can figure out a customized product which maximally meets the customer's expectations, and can also provide the suboptimal solutions for further selections. A mathematical model to formulate this optimization problem is established. A case study is used to demonstrate the feasibility and effectiveness of this approach.

In the past decade, e-commerce has emerged as an increasingly important information technology to businesses. Many enterprises have been using e-commerce as an effective and necessary trading tool in their daily business processes [1]. Particularly, as an important form of e-commerce, business-to-consumer (B2C) e-commerce is popular to more and more customers due to its convenience, quickness and price advantage. Today, many products sold on the Internet are designed as configurable products to satisfy the diversified requirements of customers [2], such as computers, cars or software packages, etc. Most of the B2C e-commerce applications use tree- or wizard-like approaches to guide a customer in configuring a customized product on-line. By using these applications, a customer can simply select each component of a configurable product from a list of available options step by step, and eventually get the customized product. For example, Dell™ uses a wizard-like B2C application to sell configurable personal computers on website (www.dell.com). One problem with either tree- or wizard-like approaches for product selection is that a customer may not know the technical parameters of a product and their meanings to this product [3]. For example, to order a computer from the Internet, a customer may be confused by technical terms such as ’CPU L2 cache size’, ’dual channel DDR2′ or ’display refresh rate’, etc. Hence the customer may feel hard to select a CPU or determine the size of the RAM that matches the selected CPU. Moreover, a customer may not know how to select well-balanced components within a predetermined budget and other expectations. For instance, a customer is going to buy a computer on-line for 2-D drafting purpose within a budget of US$ 1500. The computer has to be very reliable for professional use and has to be available in 4 weeks. In this situation the customer may find that, without a profound understanding of computer system, it is difficult to select appropriate components for the computer to avoid overall performance bottlenecks with the given conditions. To solve this problem, a promising technology called recommendation system is used by some e-commerce sites to provide the suggested products for customers [4]. The most popular recommendation methods in these systems are content-based filtering (CBF) and collaborative filtering (CF) [5]. Content-based filtering involves recommending items similar to those the customer has liked in the past; collaborative filtering, on the other hand, involves recommending items that customers, whose tastes are similar to the user seeking recommendation, have liked [6]. However, besides limitations of each paradigm indicated by Adomavicius and Tuzhilin [7], a common drawback for both of them is the new user problem [8] and [9], i.e., a new customer is unable to get accurate recommendations because there is no historical preference records available in the system. The recommendation system could not understand the customer's preferences before the customer has to rate a sufficient number of items. In addition, recommendation systems have to record customers’ rating profiles and apply various advanced techniques [7], such as clustering, artificial neural networks, neighborhood search, case-based reasoning (CBR), data mining, semantic analysis, etc. Therefore, the complexity of the recommendation systems should not be underestimated [10]. For many e-commerce enterprises, especially small and medium-sized enterprises, the complexity may keep them away from applying these recommendation systems. Regarding this problem, we propose an alternative approach to help customers to find the rational product profiles from a product family. The customer requirement (CR)–technical attribute (TA) mapping matrix of HoQ (House of Quality) is used to translate customer requirements to technical attributes, and conjoint analysis is applied to construct the component performance matrix. Eventually an optimization model is established to maximize the overall performance of the selected product profile with the total budget constraints. By applying the software system based on this approach, a customer is only required to input the levels of importance of his/her requirements for the product and total budget of the product. Then the system can automatically generate a set of product profiles, which are optimal or suboptimal solutions of this problem, for the customer as recommendations. The rest of this paper is organized as follows. Section 2 reviews the relevant literature on product family, product configuration and recommendation systems. Section 3 specifies the relationship matrices used in the mathematical model, then describes the optimization problem and establishes the mathematical model. Section 4 illustrates a case study to empirically verify the feasibility and effectiveness of the developed model. The characteristics of this approach, limitations, and some future research potentials are discussed in Section 5. Finally, conclusions are drawn in Section 6.

A new method is presented in this paper to automatically and optimally configure a customized product from a product family in B2C e-commerce environment. Based on the discussion and presentation made in this paper, the following points can be concluded: (1) The proposed method is a convenient way to configure a product for those customers who are not familiar with technical details. It can provide the optimal and suboptimal solutions considering customer's requirements and their preferences as well as the total budget of the product. (2) Conjoint analysis can be used to estimate the contribution index in a component performance matrix, which provides an effective way to explore the relationship between the performance of a component and the part-worth contributions to technical attributes.