Multi-product, Multi-supplier Order Assignment and Routing for an e-Commerce Application in the Retail Sector Louis Rivi` ere 1 , Christian Artigues 1 , Azeddine Cheref 1 , Nicolas Jozefowiez 2 , Marie-Jos´ e Huguet 1 , Sandra U. Ngueveu 1 and Vincent Charvillat 3,4 1 CNRS, LAAS-CNRS, Universit´ e de Toulouse, INSA, INP, France 2 LCOMS EA 7306, Universit´ e de Lorraine, Metz 57000, France 3 Universit´ e de Toulouse, IRIT, INP, France 4 U Devatics, Toulouse, France Keywords: e-Commerce, Retail Market, Order Assignment, Vehicle Routing, Genetic Algorithms. Abstract: With the rise of virtualization, the share of e-commerce in the retail market continues to grow in an omnichan- nel context. We consider an existing software tool, developed by the Devatics company, for pooling inventories in stores to meet online orders. The problem which arises therefore consists in seeking the optimal allocation of a set of customers to stores. In this paper we consider a variant of the offline problem corresponding to an evolution of the existing software, consisting of assigning a set of predefined orders when the transportation cost depends on a delivery tour to the customer locations. We show that the problem corresponds to a vehicle routing problem with additional but standard attributes. A mixed-integer linear programming formulation is given and several heuristics are proposed : a giant tour-based genetic algorithm, a simple cluster-first, route second heuristic and an assignment-based genetic algorithm. Preliminary computational results on a set of realistic problem instances suggest that the assignment-based genetic algorithm better scales as the problem size increases. 1 INTRODUCTION With the rise of virtualization, the share of e- commerce in the retail market continues to grow in an omnichannel context. One of the services offered by the Devatics company is Onestock 1 , a tool for pool- ing inventories to meet online orders. The problem which arises therefore consists in seeking the opti- mal allocation of a set of customers to stores. Two modes of assignment are possible. Indeed, we can consider the ”Online” assignment mode, consisting of the allocation of each order as they are declared, and the ”Offline” assignment which consists of as- signing all the orders in a single large block. In this paper we consider the offline problem. The Onestock software solves a variant where the transport cost are fixed. In this paper we consider an evolution of the problem towards a variant where the transport costs depend on delivery tours to the customer locations. We propose an mixed-integer linear programming for- mulation (MILP) of the problem. On realistic data 1 https://www.onestock-retail.com/ instances, we compare several heuristics and meta- heuristics. We show that a cluster-first, route-second based genetic algorithm obtains the best results. The problem formulation is given in Section 2. The real- istic data extraction approach that we use to generate the data instances is then presented in Section 3. Sec- tion 4 first gives a quick state of the art review of effi- cient metaheuristics for vehicle routing problems and propose adaptations for the considered E-commerce problem. Section 5 provides a computational compar- ison of the proposed exact and heuristic approaches. Concluding remarks are drawn in Section 6. 2 MODELING THE PROBLEM The problem considers a fixed set of online orders D for a set of products P. Each order d ∈ D asks for an amount q dp of product p ∈ P. We have a set M of stores, and each store m ∈ M has a stock s mp of prod- uct p ∈ P. The problem is then to meet the demand in products of each order by using the store stocks 438 Rivière, L., Artigues, C., Cheref, A., Jozefowiez, N., Huguet, M., Ngueveu, S. and Charvillat, V. Multi-product, Multi-supplier Order Assignment and Routing for an e-Commerce Application in the Retail Sector. DOI: 10.5220/0010318304380445 In Proceedings of the 10th International Conference on Operations Research and Enterprise Systems (ICORES 2021), pages 438-445 ISBN: 978-989-758-485-5 Copyright c  2021 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved