روش استخراج ترکیبی برای بهینه سازی سیاست های بازده در خرده فروشی الکترونیکی

The returns policy has long been considered as a critical yet controversial issue in the development of supply chain and marketing strategies. Up-stream manufacturers or distributors may offer returns policies to the down-stream retailers or customers to increase order and sales quantities. There are trade-offs between returns policies and customer satisfaction, product sales, and operating costs. The goal of this paper is to use a hybrid mining approach for analyzing return patterns from both the customer and product perspectives, classifying customers and products into levels, and then for adopting proper returns policies and marketing strategies to these customer classes for sustaining better profits. A multi-dimensional framework and an associated model for the hybrid mining approach are provided with a demonstrated example for validation. It is expected that by adopting suitable returns policies, benefits can be created and shared by both e-retailers and customers.

Returns policies have been adopted across various industries such as computer, publication, and pharmaceutical industries (Davis et al., 1998, Eduardo and Andres, 2004, Hoffman et al., 2002, Longo, 1995 and Padmanabhan and Png, 1997). Researchers have pointed out that the returns policy is a critical yet controversial issue in the planning and implementation of supply chain and marketing strategies (Lau et al., 2000, Mantrala and Raman, 1999, Padmanabhan and Png, 1995 and Yao et al., 2005). Up-stream manufacturers or distributors may offer returns policies to the down-stream retailers or customers to increase order and sales quantities. The most generous returns policy offers unconditional refund of wholesale/retail price for returned products, while the less generous returns policy accepts no returns at all or imposes some types of restrictions for returning (Hahn et al., 2004, Mukhopadhyay and Setoputro, 2005 and Webster and Weng, 2000). It is noted that the adoption of returns policies may substantially affect product sales and operating costs. A loose returns policy can stimulate customers’ buying decisions to leverage sales volume; nevertheless, it can also increase the number of return transactions that incur more handling and logistic costs. In the e-commerce era with more direct channels, customized products, and online shoppings, the returns policy has become an even more important strategic action for e-business to sustain competitiveness and profits. In the research literature, previous works regarding returns policy tend to formulate this problem as a mathematical model in which sales profits is the objective function to be optimized and the buyback price for returned products is the major decision variable. Among these researches, Padmanabhan and Png (1997) concern the effect of a returns policy on pricing and stocking in a competitive retail sector. They show that manufacturers should accept returns if production costs are sufficiently low and demand uncertainty is not too great. Choi, Li, and Yan (2004) investigate the optimal returns policy (also called a buyback policy) for supply chain with e-marketplace in which returned product can be sold with a higher price. Mukhopadhyay and Setoputro (2005) develop a manufacturer’s profit maximization model to jointly consider level of returns policy (buyback price for returned products) and level of modularity in product design for build-to-order products. Although these optimization approaches for dealing with returns policies do provide partial solutions to the strategic problem, many key factors are still missing in the model set up. For instance, customers’ demographic and transaction-based characteristics such as gender, income level, frequency of buying, average monetary of transactions, as well as return patterns are critical for properly classifying customers and selecting returns policies. In addition, whether product types and complexity of operation are critical factors that would influence the likelihood of customers’ return transactions? Are there any associations between customer classes, product types, and return patterns? These questions are no doubt crucial in making decisions related to the adoption and implementation of returns policies. Therefore, in order to make optimal decisions for returns policy, multiple factors from customer, product, and supply chain dimensions must be taken into account. It is reasonable to start from analyzing the return patterns of customers and products. As the business realizes more about the return patterns, they can offer to their customers better returns policies that can not only increase product sales but also decrease the probability of returns as well as associated handling costs. As a result, adopting returns policies can eventually become a win–win strategic move to benefit the supply chain businesses and customers as well. The goal of this paper is to first propose a multi-dimensional framework for illustrating the key factors of the returns policies, and then to use a hybrid mining approach for analyzing return patterns, classifying customers and products with return ratios, as well as directing suitable returns policies and marketing strategies to associated customer and product classes (Hsieh, 2005 and Kuo et al., 2002). The rest of this paper is organized as follows. The framework and hybrid mining model for returns policies are provided in Section 2. Section 3 demonstrates the mining process using an example with simulated data. Relating returns policies and marketing strategies is also discussed in this section. The final section is a conclusion and directions of future works.

We should regard returns as happy returns and making money in reverse (Stock, Speh, & Shear, 2002). Adopting proper returns policies has been a common but critical issue for retailers to gain higher customer satisfaction and profits. In this paper, we propose a multi-dimensional framework and a hybrid data mining approach to deal with this problem. Through two stages of the mining process, we generate clusters and classes for the customer and product dimensions, and derive some cross-dimensional association rules. By using the classification and association rules, better returns policies and marketing strategies can be adopted for labeled classes to increase sales and decrease returns. This proposed hybrid mining approach can be extended and applied to the entire supply chain. In addition, when returned products can be sold through other channels such as the e-marketplace, products’ lifecycles and values can be extended and preserved. Future research works will focus on extending the proposed model and approach to e-supply chain, as well as on conducting this hybrid mining process using some real world data to validate the effectiveness of the framework and process.