Journal of the Operational Research Society (2005), 1–15 r 2005 Operational Research Society Ltd. All rights reserved. 0160-5682/05 $30.00 www.palgrave-journals.com/jors A warehouse-based design model for reverse logistics MI Salema1, APB Póvoa2,* and AQ Novais3 1 Centro de Matemática e Aplicac¸ões, FCT-UNL, Caparica, Portugal; 2Centro de Estudos de Gestão, IST, Lisboa, Portugal; and 3Dep. Modelac¸ão e Simulac¸ão, INETI, Lisboa, Portugal In this paper, an MILP formulation is proposed for the design of a reverse logistics network based on a warehouse location–allocation model, which optimizes, simultaneously, the forward and reverse networks. A single product model with unlimited capacity is first defined. Subsequently, the model is extended to a multi-product capacitated recovery network model, where capacity limitations and a multi-product system can be considered. The proposed model is compared to published work in the field, where different model assumptions have been proposed. Two cases are described so as to gain a better insight into the model and allow a comparative analysis. Journal of the Operational Research Society advance online publication, 9 November 2005 doi:10.1057/palgrave.jors.2602035 Keywords: reserve logistics; optimization; model performance; warehouse location Introduction In the past recent years, a growing environmental consciousness has been shaping the way society looks on waste. A challenging question has emerged: what can be done to products that until now have been considered as waste? Society’s attitude towards environmental issues has been changing and different kinds of influential factors can be identified: (1) legislative initiatives by governments, (2) increasing awareness from consumers and (3) companies’ perception on new business opportunities. Thus, many companies have been adapting their practices to meet this new reality. This change has a profound reflex in the logistic chain, since the supply chain no longer finishes at the point where products reach the final customers, but it now includes their return as well. Returned products can range from disposed products that are to be recycled/ remanufactured to products that are sent back due to costumer’s dissatisfaction. Reverse logistics addresses this modified supply chain concept and all its related aspects and, in terms of operations, covers a wide range of areas: distribution, production planning, scheduling and stocking, among others. It can be said that reverse logistics has an impact in all fields that are covered by the traditional logistics operations (forward logistics). However, there are some differences: (1) in forward logistics, uncertainty concerns clients’ demand, whereas in the reverse logistics uncertainty is mostly concerned with *Correspondence: APB Póvoa, Centro de Estudos de Gestão, IST, Av. Rovisco Pais, Lisboa 1049-101, Portugal. E-mail: apovoa@ist.utl.pt clients’ supply to the reverse network; (2) in terms of the number of suppliers and customers, the two chains also differ: the forward chain has, usually, few suppliers and a large number of customers, while the opposite applies to the reverse chain, which might involve a large number of suppliers and a small number of customers. In spite of these differences, many aspects are common and thus an integrated analysis should be performed. Its relevance, in a macro-perspective, is related to whether or not there are advantages to optimize these two structures simultaneously, or if, on the other hand, it is better to treat them as two separate and independent networks. In the last decade, the interest in this subject has grown, both in terms of public and academia. And while the number of scientific publications has increased considerably, the work reported is essentially case study based, leading to models difficult to be applied to a wide variety of problems. The need for general purpose models can therefore be identified as an important issue. Below, some of this work will be analysed in detail and it will be seen that just two papers deal with the simultaneous design of the forward and reverse logistics chains. Barros et al (1998) study the problem of recycling sand from demolition waste and propose a model for the sand reverse distribution network. This model is based on the multi-level facility location model with capacity constraints, where the ‘raw-material’ is sand from demolition waste. A heuristic based on linear relaxation was specially developed for this case study. In the same year, Marı́n and Pelegrin (1998) propose a factory location model. The authors assume that there is a