Issues in reverse supply chains, part II: reverse distribution issues – an overview

Abstract

In recent years, due to growing environmental concern, there has been an increasing attention to product take‐back, product recovery and the re‐distribution of end‐of‐life products. Reverse logistics (RL), which refers to the distribution activities involved in product returns, has recently received much attention because many companies are using it as a strategic tool to serve their customers; and can generate good revenue. The management of return flow usually requires a specialised infrastructure with special information systems for tracking and dedicated equipment for the processing of returns. An efficient reverse distribution structure may lead to a significant return on investment as well as a significantly increased competitiveness in the market. Therefore, the time seems to be right for a systematic overview of various issues that are arising in the context of reverse distribution. The main purpose of this paper is to review the literature on RL and suggest a classification based on reverse distribution issues. The result of this research provides a better understanding of the RL concept and outlines some future directions of research on modelling and analysis.

Keywords:

• reverse logistics
• closed loop supply chain
• vehicle routing problem
• decision‐making
• third‐party reverse logistics

1. Introduction

Reverse logistics (RL) has gained increasing interest among researchers and practitioners of operations and supply chain management. Possible cost reductions, more rigid environmental legislation and increasing environmental concerns of consumers have led to this increasing attention to RL. However, more than 30 years ago, Ginter and Starling (1978) anticipated that reverse channels of distribution would become central to business activities and time has proven them to be right. In view of this interest, a literature review was proposed to analyse the existing literature that has been published on reverse distribution. The existence, effectiveness and efficiency of service management activities such as repair services and value recovery depend heavily on effective RL operations. Effective RL focuses on the backward flow of materials from customer to supplier (or alternate disposition) with the goals of maximising value from the returned item or minimising the total RL cost. Rogers and Tibben‐Lembke (1999) define RL as ‘the process of planning, implementing, and controlling the efficient, cost‐effective flow of raw materials, in‐process inventory, finished goods and related information from the point of consumption to the point of origin for the purpose of recapturing value or proper disposal’. Tibben‐Lembke and Rogers (2002) in their examination of forward versus RL systems identified the various costs associated with RL and compared them to forward logistics.

Carter and Ellram (1998) pointed out the significance of specifying a well grounded conceptual framework for RL management. They emphasised more on the environmental aspect of RL and defined it as the ‘process whereby companies could become more environmentally efficient through recycling, reusing, and reducing the amount of materials used’. Bloemhof‐Ruwaard et al. (1999) examined distribution issues such as location of collection points in a reverse logistic system. Fleischmann et al. (2000) reviewed and analysed recent case studies on logistics network design issues in the context of recovery networks and devised a framework of three typical RL network structures namely RL networks for bulk recycling, remanufacturing and reuse. Dowlatshahi (2000) in his work defined five categories of the literature on RL:

1. studies wherein the authors attempted to provide the basic concepts and a general summary of RL (De Brito and Dekker 2002, Kopicki et al. 1993, Rogers and Tibben‐Lembke 1999, Stock 1992, 1998);

2. scholarly works addressing quantitative approaches (Fleischmann et al. 1997, Fleischmann et al. 2000, Minner 2001);

3. works dealing with more specific logistical issues such as distribution, warehousing, and transportation (Jahre 1995, Pohlen and Farris 1992);

4. examination of company profiles illustrating that some manufacturing technologies had a critical role in the performance of RL systems (Thierry et al. 1995);

5. research into applications of RL in goods produced, for instance, plastics, papers, metals and other materials (Kroon and Vrijens 1995).

Ferguson and Browne (2001) examined the emerging issues in RL, in particular the information requirements for RL within the extended enterprise. They addressed the initial development of possible distribution channels, their key operational decisions and supporting information systems for the recycling of end of life products. Murthy and Djamaludin (2002) reviewed the literature about different aspects of warranties for new products. De Brito et al. (2002) discussed over 60 case studies of RL activities and, based on the studies, they indicated the critical factor for the practice of RL. They presented the case studies according to the following: decision‐making focus; network structures; relationships; inventory management; and planning and control. Dekker et al. (2004) reviewed multi‐echelon RL network models for closed‐loop supply chains (CLSCs). Beullens (2004) described some important frameworks, models and insights of RL that had been developed in recent years. Mitra (2005) provided a detailed review of the literature on vehicle routing problem with backhauling (VRPB). Prahinski and Kocabasoglu (2006) reviewed the literature in reverse supply chains and developed 10 research propositions to be studied using empirical research methods. They found that opportunities exist to use survey‐based research methods to explain current practices, predominant and critical issues and managerial techniques used to manage the reverse supply chain. Srivastava (2007) classified the green supply chain management (GrSCM) literature into three broad categories on the problem context:

1. literature highlighting the importance of GrSCM;

2. literature on green design;

3. literature on green operations.

Rubio et al. (2008) analysed the main characteristics of articles on RL published between 1995 and 2005 in the production and operations management field. They did not cover adequately all the aspects of reverse distribution.

The specific objectives of this paper are:

1. to suggest a classification of available literature in the field of reverse distribution;

2. to identify critical issues on each classification;

3. to identify emerging trends in the field of reverse distribution;

4. based on above, to suggest directions for future researchers in this field;

5. as far as possible to consolidate all available literature on reverse distribution.

The organisation of this paper is as follows. After a brief introduction, the details of the research methodology are presented in Section 2. In Section 3, the existing literature has been classified based on the reverse distribution issues. Detailed discussion on these classifications along with critical issues on each is undertaken in Section 4. Finally, Section 5 closes the paper by offering conclusions and an attempt to provide some perspectives on future research.

Objectives and format

This paper is the second of three reviews by the authors that consider issues in reverse supply chain management. The third and final paper, to be published in a forthcoming issue of the Journal, will consider the classification and simple analysis of reverse supply chains. In addition to the environmental regulation, product recovery and inventory management issues of Part I, Sasikumar and Kannan (2008), issues related to reverse distribution are addressed here. The specific objectives of this paper are:

1. to suggest a classification of available literature in the field of reverse distribution issues;

2. to identify critical issues for each classification;

3. to identify emerging trends in the field of reverse distribution;

4. to suggest directions for future researchers in this field;

5. as far as possible, to consolidate the available literature on reverse distribution issues.

The paper is organised as follows: after a brief introduction, the details of the research methodology are presented in Section 2. In Section 3, the existing literature has been classified based on reverse distribution issues. Detailed discussion and the identification of critical issues are undertaken in Section 4. Finally, Section 5 closes the paper by offering conclusions and some suggestions for future research.

2. Research methodology

A literature survey was employed as the research methodology in this study to develop a framework for reverse distribution. The literature on RL was collected from the journals in the areas of operations management, supply chain, material recovery, operations research, environmental engineering and information systems. This survey will be useful to those researchers who are interested in the modelling and analysis of reverse distribution networks. The literature search included journals published by numerous publishers, in particular Elsevier, Emerald, Inder Science, Springer link, and Taylor and Francis. As we pointed out earlier, our aim was to analyse those articles that directly related to reverse distribution.

3. Classification based on the reverse distribution issues

The literature available on RL is reviewed here with a focus on distribution issues in RL. The literature on reverse distribution issues is classified into five categories:

1. network design for RL;

2. third‐party RL (3PRL);

3. vehicle routing problem;

4. decision‐making models;

5. role of information technology in RL.

4. Reverse distribution issues

Reverse distribution can take place through the forward channel, through a separate reverse channel, or through a combination of both forward and reverse channels. The possible functions in the reverse distribution channel are: collection, testing, sorting, transportation and processing (Pohlen and Farris 1992). It is necessary to design a distribution network to perform these functions. The important aspect in a reverse distribution network is the degree of uncertainty in terms of both quantity and quality of used products returned by the consumers. Both are important determinants for a suitable network structure. Many authors addressed the design of reverse distribution networks in a product recovery context. Stuart et al. (2005) addressed the problem of reducing the inefficiency and excess material handling in returns processing at a fashion catalogue distributor.

4.1 Network design for reverse logistics

In conventional supply chains, logistics network design was commonly recognised as a strategic issue of prime importance. The location of production facilities, storage concepts, and transportation strategies are major determinants of supply chain performance. Reverse logistics should also be taken into account during the design of the support network such as location and capacity of warehouses, plants, choice of outsourcing vendors, distribution channel and supporting technology. Efficient implementation requires appropriate logistics structures to be set up for the arising goods flow from users to producers. The design of such logistics networks is investigated. In almost all of the case studies, mixed integer linear programming (MILP) location‐allocation models have been proposed to support network design.

4.1.1 Open‐loop reverse supply chain

Recycling can often be described as an open‐loop system, because the products are not returned to the original producer but can be used in other industries. Min (1989) developed a multiple objective mixed integer program to select the most desirable shipping options (direct versus consolidated) and transportation modes for product recall. Although he considered a trade‐off between transportation time and cost associated with RL, his model could not handle multi‐modal situations. Zografos and Samara (1989) presented a location routing model to study the hazardous waste transportation and disposal problem by considering the criteria:

1. minimise disposal risk;

2. minimise routing risk;

3. minimise travel time.

Kroon and Vrijens (1995) reported a case study concerning the design of a logistics system for reusable transportation packages. The authors proposed the use of MILP model in which the major decisions were the determination of the number of containers, the number of container depots and their locations, and the service, distribution and collection fees. The proposed model did not consider the capacity limits of the depots. Jayaraman and Srivastava (1995) developed a model for the multiple facility, multiple levels of equipment problem called the multiple equipment multiple cover facility location‐allocation problem. This considered maximal coverage of demand, including multiple coverage of demand by equipment of a specific type, given that each equipment had a certain probability of being unavailable. Del Castillo and Cochran (1996) presented a pair of linear programs (one aggregated and another disaggregated) and a simulation model to optimally configure the RL network involving the return of reusable containers so that the number of reusable containers was maximised. The authors did not take into account transportation issues related to RL.

Marin and Pelegrin (1998) modelled a special type of location problem (return plant location problem) to determine the plants to be opened, the amount of primary product required by each customer (from plant) and the amount of secondary product returned from each customer and described heuristic and exact algorithm. Min et al. (1998) reviewed the existing location routing problem literature with respect to its algorithmic development and added realities. Barros et al. (1998) proposed a two‐level location model for the sand recycling problem and considered its optimisation using heuristic procedures. They formulated a MILP model to minimise the total cost of the network. Krikke et al. (1999) also proposed MILP model for the multi‐echelon product recovery network design which focused on the remanufacturing of photocopiers. Nagel and Meyer (1999) discussed a novel method for systematically modelling end‐of‐life networks and showed ways of improving the existing and new systems with ecological and economical concerns. Louwers et al. (1999) presented a non‐linear facility location‐allocation model for the collection and reprocessing of carpet waste. Duhaime et al. (2000) discussed the collection and distribution of returnable containers for Canada post.

Fleischmann (2001) examined reverse supply networks consisting of multiple collection points and central facilities where returned products were inspected, sorted and refurbished. Shih (2001) utilised a mixed integer programming model to optimise the infrastructure design and the reverse network flow for the recovery of electrical appliances and computers. Knemeyer et al. (2002) utilised a qualitative methodology to examine the feasibility of designing an RL system to recycle and/or refurbish end‐of‐life computers that were deemed no longer useful by their owners. Hu et al. (2002) formulated a linear analytical cost‐minimisation model for a multi‐time‐step, multi‐type hazardous wastes RL system. De Brito and Dekker (2002) provided a decision framework for RL in terms of strategic, tactical and operational aspects of the problem. Ammons et al. (2002) addressed the network design problem of uncertainty by a solution methodology using an upper and lower‐bounding scheme on the robust objective function, instead of examining the results from a finite set of scenarios. Pochampally and Gupta (2003) utilised a three‐phase mathematical programming approach to effectively design an efficient reverse supply chain network. In Phase I, they selected the most economical product to reprocess from a set of different types of used‐products, using a mixed‐integer mathematical programming model. Phase II implemented the analytic hierarchy process to identify potential facilities in a set of candidate recovery facilities. Phase III solved a single time‐period discrete location model to achieve transportation of the right mix and quantities of goods across the reverse supply chain network. Jayaraman et al. (2003) formulated a MILP model to determine an efficient strategy for the RL operations of hazardous products. The objective of the model was to find the optimal number and location of collection and refurbishing facilities with the corresponding flow of the hazardous products. Bogataj and Bogataj (2004) presented a facility location problem to achieve optimal response time in the lead time study in the frequency space, where different perturbations could appear. Realff et al. (1999, 2000, 2004) proposed a robust approach extension to a design model for carpet recycling previously considered in Ammons et al. (1999). Savaskan and Van Wassenhove (2006) examined how the allocation of product collection to retailers impacts their strategic behaviour in the product market, and discussed the economic trade‐offs the manufacturer faced while choosing an optimal reverse channel structure. Bautista and Pereira (2006) focused on the problem of locating collection areas for urban waste where different types of waste (glass, paper, plastic, organic material) were stored in special refuse bins. Min et al. (2006a) proposed a MILP model and a genetic algorithm to solve the RL problem involving product returns with the objective of minimising the total RL cost. A MILP model was also developed by Sharma et al. (2007) to facilitate better leasing and logistics decisions (including end‐of‐life disposal options) from the perspective of an electronic equipment leasing company. Pati et al. (2006a, 2006b) proposed a linear programming optimisation model to minimise the supply chain cost for the Indian paper industry with two different sources of raw materials (wood and wastepaper).

Kara et al. (2007) presented a simulation model of an RL network for collecting end‐of‐life appliances in the Sydney metropolitan area and calculated the collection cost in a predictable manner. Salema et al. (2007) proposed a generalised model for the design of an RL network where capacity limits, multi‐product management and uncertainty on product demands and returns were considered. The model was based on the recovery network model (RNM) proposed by Fleischmann et al. (2001). Lieckens and Vandaele (2007) extended a facility‐location MILP model in an RL context with queuing relationships in order to incorporate a product's cycle time and inventory holding costs, as well as to deal with the higher degree of uncertainty and congestion, typical characteristics of these networks. Min et al. (2008) proposed a mixed‐integer programming model and a genetic algorithm that can solve the RL problem involving consolidation of returned products. Srivastava (2008) formulated a multi‐echelon, multi‐product, multi‐period MILP model (conceptual model) as a bi‐level optimisation problem. Pati et al. (2008) formulated a mixed integer goal programming model to assist in proper management of the paper recycling logistics system in India and studied the inter‐relationship between multiple objectives (with changing priorities) of a recycled paper distribution network. The proposed model also assists in determining the facility location, route and flow of different varieties of recyclable waste paper in multi‐item, multi‐echelon and multi‐facility decision‐making framework. Aras and Aksen (2008) addressed the problem of locating collection centres of a company for distance and incentive‐dependent returns. They formulated a mixed integer non‐linear facility location‐allocation model to determine both the optimal locations of the collection centres and the optimal incentive values for each return type to maximise the profit from the returns. Neto et al. (2008) developed a framework for the design and evaluation of sustainable logistic networks, in which profitability and environmental impacts are balanced. They introduced a new methodology based on the properties shared by multi‐objective programming and data envelopment analysis (DEA). Mutha and Pokharel (2008) proposed a mathematical model for the design of an RL network handling product returns. The model considered modular product structure with different disposal and recycling fractions for each module of each product.

4.1.2 Closed‐loop supply chain

A major issue in reverse distribution is integration of the forward and reverse channels. Returns information captured should be integrated with forward supply chain information to achieve optimum planning and reduction of costs. The whole support network can then be designed in such a way that it can service both the forward and RL processes efficiently. This is in line with the concept of a CLSC design. Remanufacturing and reuse often lead to closed‐loop systems since the product or packages are returned to the original producer. French and LaForge (2006) investigated re‐use issues and practices related to CLSCs in process industry firms from the producer's perspective with the objective of identifying important issues in the field. Talbot et al. (2007) analysed empirical evidence from a sample of 205 environmentally responsive SMEs operating in the fabricated metal products and electric/electronic products industries. A coherent research model was developed which classified the CLSC activities along two dimensions, the forward and reverse supply chains.

Clendenin (1997) provided an overview of a reengineering approach to RL and presented the methodologies which might be helpful to management and the issues associated with business problem benchmarking as an input to reengineering. Jayaraman et al. (1999) developed a MILP model for a closed‐loop logistics system. In their model, only remanufactured products constitute the forward and reverse flows. Fleischmann et al. (2001) considered the integration of forward and reverse distribution, and gave a generic integer programming formulation. They took two cases of photocopier remanufacturing and paper recycling, and showed that there was potential for cost savings if one undertook an integrated view rather than a sequential design of the forward and reverse distribution networks. Similarly Fleischmann et al. (2003) considered the integration of CLSCs and spare parts management at IBM. Schultmann et al. (2003) developed a hybrid approach to establish a CLSC for spent batteries that combines an optimisation model for planning a reverse‐supply network and a flow‐sheeting process model that enables a simulation tailored to potential recycling options for spent batteries in the steelmaking industry. Krikke et al. (2003) developed a quantitative modelling approach to support decision‐making concerning both the design structure of a product (Krikke et al. 2004) and the design structure of the logistics network and the model was applied to a CLSC design problem for refrigerators using the real life R&D data of a Japanese consumer electronics company. Guide et al. (2003) investigated different remanufacturing strategies and the associated product and process characteristics. The US Navy aviation depots were used as a typical make‐to‐order (MTO) example in remanufacturing. Beamon and Fernandes (2004) developed a multi‐period integer programming model to study a CLSC in which manufacturers produce new products and remanufacture used products. They used the present‐worth method to jointly analyse investment and operational costs. A sensitivity analysis of the model was performed, and conclusions were made regarding model behaviour and performance. Dyckho et al. (2004) dealt with the expansion of a supply chain to closed loop systems and analysed the material flow in the automotive cycle. Seitz and Peattie (2004) investigated a case of remanufacturing engines and addressed the issue of acquiring cores from this installed base. The issues concerning customer relationship in the CLSC were also discussed. Savaskan et al. (2004) addressed the problem of choosing a suitable channel structure for the collection of end‐of‐life returns from customers. They considered a manufacturer with three options for collecting such products:

1. undertaking the collection effort by himself;

2. providing suitable rewards to the retailer to undertake the collection effort;

3. subcontracting the collection effort to a third party.

Min et al. (2005) presented a non‐linear integer program for solving the multi‐echelon, multi‐commodity closed loop network design problem involving product returns. However, their models did not consider temporal consolidation issues in a multiple planning horizon. Chouinard et al. (2005) dealt with problems related to the integration of RL activities into the regular supply chain and to the coordination of the information system. Inderfurth (2005) focused on a product recovery system where, in the context of extended product responsibility, a manufacturer of original products was also engaged in remanufacturing used products taken back from its customers. For this type of CLSC the optimal recovery and production policy was evaluated. The analysis was restricted to stationary demand and return patterns. Sheu et al. (2005) formulated a linear multi objective programming model to optimise the operations of both integrated logistics and corresponding used‐product RL in a given green‐supply chain. Factors such as the used‐product return ratio and corresponding subsidies from governmental organisation for RL were considered in the model formulation. Jayaraman (2006) presented an analytical approach towards production planning and control for CLSCs with product recovery and reuse. This approach consists of a mathematical programming model, remanufacturing aggregate production planning, for aggregate production planning and control. The model was designed to aid operational decision‐makers in an intermediate to long‐range planning environment and also served as a focal point for developing formal systems for production planning, inventory control, and other tactical decision‐making. Kumar and Malegeant (2006) showed that a manufacturer could create value by implementing a partnership with an eco‐non‐profit community organisation in the collection process of used products for the CLSC. The study focused on the reuse‐a‐shoe program of Nike and the creation of Throwplace.com to point out the benefits of strategic alliances between manufacturers and eco‐non‐profit organisations. Srivastava and Srivastava (2006) developed a conceptual model and an integrated modelling which utilised product ownership data, average life cycle of products, past sales, forecasted demand and likely impact of environmental policy measures to manage product returns for RL by focusing on estimation of returns for select categories of products in the Indian context. Min et al. (2006b) proposed a mixed integer non‐linear programming model to minimise the total RL costs for the RL problem involving both spatial and temporal consolidation of returned products.

Vlachos et al. (2007) tackled the development of efficient capacity planning policies for remanufacturing facilities in reverse supply chains, taking into account not only economic but also environmental issues, such as the take‐back obligation imposed by legislation and the ‘green image’ effect on customer demand. The behaviour of the generic system under study was analysed through a simulation model based on the principles of the system dynamics methodology. Similarly, Kumar and Yamaoka (2007) proposed system dynamics (SD) modelling methodology (Spengler and Schroter 2003) to analyse the CLSC design for the Japanese car industry. Jun et al. (2007) introduced the research issues on closed‐loop product life‐cycle management where product information flow was closed due to emerging technologies. Morana and Seuring (2007) presented a classification of products for end‐of‐life acquisition based on the marginal value of time and the product life‐time. Hammond and Beullens (2007) formulated a model with the intent of examining issues surrounding the recent European Union directive regarding waste electric and electronic equipment (WEEE). They expanded the work of Nagurney et al. (2002) and Nagurney and Toyasaki (2005) and presented a variational inequality, CLSC formulation that allows for such cases. Kocabasoglu et al. (2007) investigated whether four dimensions of business uncertainty, namely munificence, dynamism, hostility and heterogeneity, influenced investments in both forward and reverse supply chains. Listes and Dekker (2005) and Listes (2007) presented a generic stochastic model for the design of networks comprising both supply and return channels, organised in a closed loop system. The authors described a decomposition approach to the model, based on the branch‐and‐cut procedure known as the integer L‐shaped method. Seitz (2007) focused on in‐depth case studies within the remanufacturing facilities of a major European vehicle manufacturer. Webster and Mitra (2007) examined the impact of take‐back laws within a manufacturer/remanufacturer competitive framework. They developed a general two‐period model to investigate questions of interest to policy‐makers in government and managers in industry. Lu and Bostel (2007) proposed a two (0,1) level mixed integer programming model, in which simultaneously forward and reverse flows and their mutual interactions were considered. The problem was formulated as an uncapacitated facility location model and an algorithm based on Lagrangian heuristics was developed.

Lee and Dong (2008) discussed the logistics network design for end‐of‐lease computer products and developed a deterministic programming model for systematically managing forward and RL flows with the objective of minimising the total cost in the logistics network. Mitra and Webster (2008) analysed a two‐period model of a manufacturer who makes and sells a new product and a remanufacturer who competes with the manufacturer in the second period and examined the effects of government subsidies as a means to promote remanufacturing activity. Chung et al. (2008) proposed a multi‐echelon inventory system with remanufacturing capability. They developed a CLSC inventory model that maximised the joint profits of the supplier, the manufacturer, the third‐party recycle dealer and the retailer under contractual design. Fuente et al. (2008) proposed an integrated model for supply chain management in which the operation of the reverse chain was built based on the existing processes of the forward chain. Finally the proposed model was validated in a company from the metal‐mechanic sector. Kusumastuti et al. (2008) developed a facility location‐allocation model for redesigning a closed‐loop service network at a computer manufacturer. The model considered the possibility of having the network span across several countries and multi‐period planning horizons. Sheu (2007) presented a coordinated RL management system for multi‐source hazardous wastes in a high‐technology manufacturing region. A linear multi‐objective analytical model was formulated to minimise both the total RL costs and the corresponding risks. Sheu (2008) formulated a linear multi‐objective optimisation model to optimise the operations of both nuclear power generation and the corresponding induced waste RL. Factors such as the operational risks induced in both the power generation and RL processes were considered in the model formulation. Gu and Ji (2008) presented a cost‐minimisation operational model for the integrated logistics of a remanufacturing/manufacturing (R/M) system based on the consumer market. The functionality of the proposed operational model focused mainly on production planning and distribution planning of the renew/new products of the period T, collection planning and distribution planning of the used‐products of the period T‐2, disassembly planning and distribution planning of the disassembled products of the period T‐1. Kannan et al. (2008b) designed an integrated forward logistics multi‐echelon distribution inventory supply chain model and closed loop multi‐echelon distribution inventory supply chain model for the built‐to‐order environment using genetic algorithm and particle swarm optimisation and the proposed model was validated by considering two case studies: one for a tyre manufacturer and the other for a plastic goods manufacturer.

4.2 Third‐party reverse logistics (3PRL)

Reverse logistics may take place either through the original network or through specialised logistical providers (third‐party logistical providers). When considering outsourcing decisions (Serrato et al. 2007) for RL, the fundamental factor to consider is whether there is a viable third‐party reverse logistics provider (3PRLP) for the type of RL network required.

Krumwiede and Sheu (2002) developed a RL decision‐making model to guide the process of examining the feasibility of implementing RL in third‐party providers such as transportation companies. Spicer and Johnson (2004) proposed the concept of third party demanufacturing. It was defined as an extended producer responsibility approach in which private companies take up end‐of‐life responsibility for products on behalf of the original equipment manufacturer (OEM). Mukhopadhyay and Setoputro (2006) proposed the use of fourth party logistics (4PL) as a return service provider. They presented a profit‐maximisation model to jointly obtain the optimal policies for the seller and the 4PL through the use of Stackelberg‐like game theory, where the seller acts as the leader and the 4PL acts as the follower. Ko and Evans (2007) developed an optimisation model and associated algorithm to design an integrated logistics network for 3PLs. Min and Ko (2008) proposed a mixed‐integer programming model and a genetic algorithm to solve the RL problem involving the location and allocation of repair facilities for 3PLs. Du and Evans (2008) addressed a problem involving a manufacturer outsourcing to a third‐party logistics (3PL) provider for its post‐sale service. They developed a bi‐objective, mixed integer programming optimisation model for the RL networks that dealt with the returns requiring repair service.

4.3 Vehicle routing problem (VRP)

The vehicle routing problem in RL system design is the problem of routing vehicles to serve a set of locations for both delivery and backhaul of products. The Tabu Search (TS) method has been proved to cope sufficiently with VRP problems.

Jennings and Scholar (1984) formulated the regional hazardous waste management system as simply a vehicle routing problem in an attempt to accomplish the goal of either minimum cost or minimum risk. In ReVelle et al. (1991), a synthesised linear programming method was proposed to manage the RL flows of spent nuclear fuel. Chang et al. (1997) suggested that short‐term planning of vehicle routing and scheduling problems would be a valuable subsequent analysis after the completion of long‐term regional planning for solid waste management. Dethloff (2001) investigated the relationship between the vehicle routing problem with simultaneous delivery and pick‐up and other vehicle routing problems, and suggested a heuristic construction procedure to a real life problem. Beullens et al. (2003) provided an excellent survey of sector design models in RL and discussed the vehicle routing issues in an RL systems. Süral and Bookbinder (2003) solved exactly small instances of a particular case of problem in which each customer had a pickup or a delivery demand, but not both, and hence was visited only once. Nearest neighbour or sweep constructive procedures, as well as improvement procedures making customer relocations were proposed. The single vehicle routing problem with deliveries and selective pickups arises in a number of RL contexts in which customers returned goods (e.g. empty containers) to the depot. Such a problem, encountered in the Quebec City area, was recently analysed by Privé et al. (2006). It involved the delivery of bottled water and soft drinks to convenience stores and the collection of empty bottles and aluminium cans. Mourao and Amado (2005) presented a new heuristic method to find high quality solutions for a mixed capacitated arc routing problem, inspired by the household refuse collection problem in Lisbon. Schultmann et al. (2006) modelled RL aspects with vehicle routing planning. The objective was to generate a tour schedule with minimum cost.

McKinnon and Ge (2006) examined the recent trend in empty running by trucks in the UK and assessed the potential for a further reduction in empty running in the food supply chain using a new technique. Alshamrani et al. (2007) examined an RL problem, motivated by blood distribution for the American Red Cross, where containers in which products were delivered by a single vehicle from a central processing point to customers (stop) in one period were available for return to the central point in the following period. Krikke et al. (2008) presented an approach (MUST and CAN orders) to optimise the collection (transportation) of dismantled materials from end‐of‐life vehicles in real life cases for auto recycling and they discussed an application of online monitoring of inventory levels in RL to improve the collection efficiency of the mandatory collection of dismantled materials. The routing problems were solved by a combination of route generation and set partitioning. Aras et al. (2008) addressed the collection centre location problem (CCLP) of a company that aims to collect used products from product holders. They formulated a mixed‐integer non‐linear facility location‐allocation model to find both the optimal locations of a predetermined number of collection centres and the optimal incentive values for different return types. Since the problem is NP‐hard, they developed a Tabu Search based heuristic called TS‐CCLP which incorporated a simplex search procedure as a subroutine. Gribkovskaia et al. (2008) formulated a mixed integer, linear programming model for single vehicle routing problem with deliveries and selective pickups. The objective was to design a vehicle route of minimum net cost, visiting each customer, performing all deliveries, and a subset of the pickups.

4.4 Decision‐making models

Multi‐criteria decision‐making (MCDM) is one of the most widely used decision methodologies in the sciences, business, government and engineering worlds. MCDM methods can help to improve the quality of decisions by making the decision‐making process more explicit, rational and efficient.

Meade and Sarkis (2002) developed a model for selecting and evaluating 3PRLP using analytic network process (ANP). However, their model did not represent a tool for determining whether or not to outsource RL activities, but it helped in the decision of selecting a 3PRLP once the outsourcing strategy was chosen by the firm. Guide and Pentico (2003) developed a hierarchical decision model for remanufacturing and reuse. Haas and Murphy (2003) presented three techniques along with the basic Charnes, Cooper and Rhodes (CCR) model to compensate for non‐homogeneity in decision‐making units in DEA and the results showed that none of the adjustment mechanisms are clearly superior to the unadjusted CRR model. Bufardi et al. (2004) proposed a multi‐criteria decision‐aid approach to aid the decision‐maker in selecting the best compromise EOL alternative on the basis of his/her preferences and the performances of EOL alternatives with respect to the relevant environmental, social and economic criteria. A case study was provided to illustrate how the proposed ELECTRE III could be used for product EOL alternative selection in real‐world applications. Srivastava and Srivastava (2005) developed a hierarchical decision‐making framework to find the feasibility of profit‐driven RL networks. Ravi et al. (2005a) presented an ANP‐based decision model to structure the problem of conduct of RL for EOL computers in a hierarchical form and linked the determinants, dimensions, and enablers of the RL and the alternatives available to the decision‐maker for a computer industry. A combination of balanced scorecard and ANP‐based approach provided a more realistic and accurate representation of the problem for conducting RL operations. Ravi et al. (2005b) employed an interpretive structural modelling (ISM) based approach to model the RL variables typically found in computer hardware supply chains. These variables had been categorised under ‘enablers’ and ‘results’. The main objectives were: to identify and rank the variables of RL activities in the computer hardware industry; to find out the interaction among identified variables; and to understand the managerial implications of this research. Ravi and Shankar (2005) analysed the interactions between those major barriers that prevent the application of RL in automobile industries using ISM methodology. Tan and Kumar (2006) presented a decision‐making model for manufacturers to maximise their profits in RL operations. A system dynamic model was developed to complement prior models and was validated using data collected from a computer manufacturer handling returns with volumes transacted over a period of two years. Bian and Yu (2006) analysed various countries in the Asia pacific region to determine their suitability for carrying out RL operations for an international electrical manufacturer using the analytical hierarchy process (AHP). Kuo et al. (2006) presented an efficient green fuzzy analysis method to allow designers to evaluate different design alternatives. Fuzzy analysis was used to reduce the bias caused by the weighting factor, design attributes and people. They constructed the hierarchical structure of environmentally‐conscious design indices by using AHP with four aspects, namely, inventory analysis, impact analysis, life‐cycle cost analysis and improvement analysis.

Staikos and Rahimifard (2007) applied AHP as a decision‐making model to identify the most appropriate reuse, recovery and recycling option for post‐consumer shoes. Lu et al. (2007) presented a multi‐objective decision‐making process for green supply chain management to measure and evaluate the supplier's performance using AHP modified by a fuzzy logic process. Araz et al. (2007) proposed a methodology for a outsourcing management utilising information obtained from outsourcing selection process. The proposed methodology was based on preference ranking organisation METHod for enrichment evaluations (PROMETHEE) and fuzzy goal programming. Similarly, Queiruga et al. (2008) applied PROMETHEE as a multi‐criteria decision‐making method for the selection of good alternatives for potential locations of WEEE recycling plants in Spain. Shi et al. (2008) applied AHP to examine and prioritise underlying barriers to adoption of cleaner production by SMEs in China from the perspectives of government, industry and expert stakeholder groups. In Ravi et al. (2008), a combination of ANP and zero one goal programming was used as a solution methodology to deal with a problem related to the selection of feasible RL for EOL. Gomes et al. (2008) presented two cases where the decision‐makers had different preferences. In the first case, social agents required an evaluation of different disposal alternatives for plastic waste. In the second case, existing construction and demolition waste recycling facilities required a performance evaluation. With the help of a multi‐criteria decision aiding hybrid algorithm, a multi‐criteria decision support system under development at COPPE/UFRJ and CASNAV, a consistent hierarchy of the possible alternatives was achieved. Shankar et al. (2008) employed an SD approach to model RL variables (enablers, results and inhibitors) in an automobile company, and the interactions among the variables affecting the system was investigated and a causal relationship developed. Efendigil et al. (2008) proposed a method for selecting appropriate and desirable third‐party vendors taking the subjective requirements of the company into account. The proposed methodology outlined the development of a conceptual framework integrating fuzzy logic and artificial neural networks as a tool while including the environmental factors in the design of an RL provider selection. Kannan et al. (2008a) analysed the interaction of criteria that was used to select the green suppliers who addressed the environmental performance using ISM and AHP and the effectiveness of the model was illustrated using an automobile company. Kannan et al. (2008c) proposed a multi‐criteria decision‐making model for selecting the collecting centre location in the RL supply chain model using a analytical hierarchy process and fuzzy analytical hierarchy process. Wadhwa et al. (2008) made an attempt to bring fuzzy‐based flexible MCDM and RL together as a well‐suited group decision support tool for alternative selections.

4.5 Role of information technology in reverse distribution

One of the most serious problems that companies face in the execution of an RL operation is the dearth of good information systems. In order to overcome this threat, a flexible RL information system is required. Dhanda and Hill (2005) showed that information systems and information technology play an important role in the support of an RL process. Reverse logistics is typically a boundary‐spanning process between the companies or business units of the same company, thus developing systems that have to work across boundaries adds additional complexity to the problem. As an example for a retailer, a system that tracks returns at store level is desirable. The system should create a database at store level so that the retailer can begin tracking returned product and follow it all the way back through the supply chain. Information systems should also include detailed information programs about important RL measurements, such as return rates, recovery rates, and returns inventory turnover, etc. For many companies, current information systems do not allow them to monitor the status of their returns. Additionally, useful tools such as radio frequency are helpful. New innovations such as two‐dimensional bar codes and radio frequency identification license plates (RFID) may soon be widely used.

Sarkis et al. (2004) discussed reverse e‐logistics issues and practices associated with the execution and management of the reverse e‐logistics function with respect to the natural environment. Information and communication technology can help companies realise new, innovative business opportunities in the area of CLSCs (van Nunen and Zuidwijk 2004). Mukhopadhyay and Setoputro (2004) examined the return policy in the context of the e‐business setting where the manufacturer or a retailer was selling directly to customers via the Internet. They developed a profit‐maximisation model to obtain optimal policies for price and the return policy in terms of certain market reaction parameters. Daugherty et al. (2005) provided an insight about how RL performance can be influenced by key strategic decisions by taking a survey of businesses in the automobile aftermarket industry. Finally, they suggested that information support for authorising, tracking, and handling returns can positively impact both economic and service quality‐related performance. Radio frequency identification has become an emerging technology (Saar and Thomas 2003) in the fields of supply chain management, manufacturing and logistics. A typical RFID system consists of tags and readers, application software, computing hardware and middleware. Wager et al. (2005) analysed the potential impact of the broad application of RFID labels in municipal solid waste and discussed the results from the perspective of the precautionary principle. Parlikad and McFarlane (2007) discussed how RFID‐based product identification technologies can be employed to provide the necessary information in product recovery decisions.

5. Conclusion and research implications

Implementation of legislation, social responsibility, corporate imaging, environmental concern, economic benefits and customer awareness are forcing OEMs not only to provide more environmentally‐friendly products but also to take back their used, end‐of‐lease or end‐of‐life products, or products under warranty to minimise waste and conserve resources. Therefore, OEMs have turned to a better design of their products for maximum reuse and recycling and to the retrieval of used products through a network for reuse, remanufacture, recycling or disposal, so that maximum value can be achieved from their used products. Reverse logistics, referring to the distribution activities involved in product returns, plays an important role in achieving ‘green supply chains’ by providing customers with the opportunity to return the warranted and/or defective products to the manufacturer. In view of this focus, we have analysed the literature that has been published in various scientific journals on reverse distribution, and as far as the methodology employed in the articles is concerned, we grouped the articles based on the content issues and the critical issues are addressed on each classification. In accordance with our findings from the literature review, the following conclusions are presented.

	It is observed that more research work has recently been carried out on the network design issue, particularly in CLSCs. Network design for heterogeneous products recovery with uncertain return rate is to be studied (Lee and Dong 2008) in the future.
	All the multiobjective programming models discussed are bi‐objective only. So, there is a research scope for considering more than two objectives.
	Radio frequency identification is an emerging technology in the information and communication field. So, there is a large research potential to make use of this technology in RL.
	Lieckens and Vandaele (2007) only used a differential evolution algorithm for problem solving and there is a research opportunity to make use of a non‐dominated sorting genetic algorithm and sheep flocks heredity model algorithm as a problem‐solving tool in the RL field.
	In vehicle routing problems, time‐varying road traffic conditions can be considered by incorporating the queuing effects caused by transportation.
	More statistical analysis can be conducted to identify the sensitive parameters of the network model.
	From the literature, it is identified that there is a research potential to make use of VIKOR, ELECTRE, SAW, SIR, UTA, TACTIC, MAVT and Performance value analysis as a multi‐criteria decision‐making tool for the selection of 3PRL providers and facility locations (collecting centre location, recycling and remanufacturing facility location, and disposal site, etc.).