Research Seminars 2016/2017

Emanuele BORGONOVO - Bocconi University, Italy

May 22th 2017

Sensitivity Analysis in the Management Sciences: A Review

Abstract: The solution of several operations research problems requires the creation of a quantitative model. Sensitivity analysis is a crucial step in the model building and result communication process. Through sensitivity analysis, we gain essential insights on model behavior, on its structure and on its response to changes in the model inputs. Several interrogations are possible and several sensitivity analysis methods have been developed, giving rise to a vast and growing literature. We present an overview of available methods, structuring them into local and global methods. For local methods, we discuss Tornado diagrams, one way sensitivity functions, differentiation-based methods and scenario decomposition through finite change sensitivity indices, providing a unified view of the associated sensitivity measures. We then analyze global sensitivity methods, first discussing screening methods such as sequential bifurcation and the Morris method. We then address variance-based, moment-independent and value of information-based sensitivity methods. We discuss their formalization in a common rationale and present recent results that permit the estimation of global sensitivity measures by post-processing the sample generated by a traditional Monte Carlo simulation. We then investigate in detail the methodological issues concerning the crucial step of correctly interpreting the results of a sensitivity analysis. A classical example is worked out to illustrate some of the approaches. Joint work with Elmar Plischke, Clausthal University of Technology.

Jamal OUENNICHE - University of Edinburg, United Kingdom

April 10th 2017

A Dual Local Search Framework for Combinatorial Optimization Problems with TSP Application

Abstract: In practice, solving realistically sized combinatorial optimization problems to optimality is often too time consuming to be affordable; therefore, heuristics are typically implemented within most applications software. A specific category of heuristics has attracted considerable attention, namely local search methods. Most local search methods are primal in nature; that is, they start the search with a feasible solution and explore the feasible space for better feasible solutions. In this research, we propose a dual local search method and customize it to solve the traveling salesman problem (TSP); that is, a search method that starts with an infeasible solution, explores the dual space—each time reducing infeasibility, and lands in the primal space to deliver a feasible solution. The proposed design aims to replicate the designs of optimal solution methodologies in a heuristic way. To be more specific, we solve a combinatorial relaxation of a TSP formulation, design a neighborhood structure to repair such an infeasible starting solution, and improve components of intermediate dual solutions locally. Sample-based evidence along with statistically significant t-tests support the superiority of this dual design compared to its primal design counterpart.

Zeynep AKSIN - Koç University, Turkey

March 27th 2017

How Experienced Waits Drive Queue Behavior in the Lab

Abstract: Using laboratory experiments, we study join and quit decisions by subjects from a single server, observable, first come first served queue. In a set-up that incentivizes decisions that would maximize an expected utility function that is linear in waiting costs, we explore the role queue length and encountered service time experience plays on these decisions. We show that both the probability of quitting a queue and the survival time in a queue are affected by the queue length as well as experienced service times, for the same total waiting times. We further find that subjects are less inclined to join queues with random service times relative to a benchmark queue with deterministic service times. The implications of the results in terms of queue design and delay announcements in queues are discussed, along with ongoing experiments that explore the role of a waiting time announcement upon entry. Joint work with Busra Gencer, Evrim Gunes, Ozge Pala.

Jean-Charles Chebat, HEC Montreal, Canada

February 20th 2017

Counterproductive Effects of Commonsensical Marketing Strategies in Services Marketing

Abstract: Customers are increasingly violent toward frontline employees. Service corporations developed three commonsensical strategies to deal with consumers, that is, "customer is king", "service with a smile", and "corporation as a family". Our empirical data (some 500 service employees) show that such strategies bring about paradoxical negative consequences on the employees' behaviors toward the service corporation, especially in terms of reduced commitment to the employer and deviant behavior. Employees show an increased level of anger toward the corporation, emotional exhaustion. Managerial conclusions are drawn from the findings.

Moritz FLEISCHMANN - University of Mannheim, Germany

January 30th 2017

Strategic Grading in the Product Acquisition Process of a Reverse Supply Chain

Abstract: Most recommerce providers are applying a quality-dependent process for the acquisition of used products. They acquire the products via websites at which product holders submit upfront quality statements and receive quality-dependent acquisition prices for their used devices.This presentation is based on two papers that are motivated by this development of reverse logistics practice and aim to analyse the product assessment process of a recommerce provider in detail. We first propose a sequential bargaining model with complete information which captures the individual behaviour of the recommerce provider and the product holder. We determine the optimal strategies of the product holder and the recommerce provider in this game. We find that the resulting strategies lead to an efficient allocation, although the recommerce provider can absorb most of the bargaining potential due to his last mover advantage. We then relax the assumption of complete information and include uncertainty about the product holder's valuation of the product. We show the trade-off underlying the recommerce provider's optimal counteroffer decision and analyse the optimal strategy, using a logistic regression approach on a real-life data set of nearly 6,000 product submissions. The results reveal a significant improvement potential, compared to the currently applied strategy. The second paper takes the analysis of incomplete information further and derives the equilibrium strategies if both players are facing some uncertainties. At its core, it then addresses the recommerce provider’s optimization of the price-quality menu. Finally, we propose an alternative acquisition process that overcomes some observed deficiencies of the current process.

Bernard FORTZ - Université Libre de Bruxelles, Belgium

December 15th 2016

Computational Strategies for a Multi-Period Network Design and Routing Problem

Abstract: The conventional multicommodity capacitated network design problem deals with the simultaneous optimization of capacity installation and traffic flow routing, where a fixed cost is incurred for opening a link and a linear routing cost is paid for sending traffic flow over a link. The routing decision must be performed such that traffic flows remain bounded by the installed capacities. In this talk, we generalize this problem over multiple time periods using an increasing convex cost function which takes into account congestion (number of routing paths per edge) and delay (routing path length). We propose a compact Mixed Integer Linear Program (MILP) formulation for this problem, based on the aggregation of traffic flows by destination following the per-destination routing decision process underlying packet networks. We observe that the resolution with realistic topologies and traffic demands becomes rapidly intractable with state-of-the-art solvers due to the weak linear programming bound of the proposed MILP formulation. We also introduce an extended formulation where traffic flows are disaggregated by source-destination pairs, while keeping the requirement of destination-based routing decisions. This extended formulation provides for all evaluated topologies stronger linear programming lower bounds than the base formulation. However, this formulation still suffers from the large size of the resulting variables and constraints sets; hence, solving the linear relaxation of the problem becomes intractable when the network size increases. In this talk, we investigate different computational strategies to overcome the computational limits of the formulations. We propose different branch-and-cut strategies and a Lagrangian relaxation approach. Joint work with Enrico Gorgone (ULB) and Dimitri Papadimitriou (Nokia - Bell Labs).

Bernard GENDRON - CIRRELT, Canada

November 21th 2016

Branch-and-Price-and-Cut for Multicommodity Network Design

Abstract: We consider a mixed-integer programming model that represents a large number of network design applications in transportation and logistics. We consider several alternatives for solving this model, in particular a column-and-row generation approach, recently introduced by Frangioni and Gendron (2013) under the name "Structured Dantzig-Wolfe Decomposition". We present preliminary computational results that show comparisons of the different variants on a set of large-scale network design instances.

Elena BELAVINA - University of Chicago Booth School of Business, USA

October 13th 2016

Grocery Access, Market Structure and Food Waste

Abstract: This paper studies how access to grocery stores, and the extent and nature of competition in the grocery retail market influences food waste. Access to grocery, or how dense is the network of retail stores in a neighborhood, varies extensively as a result of zoning laws and other city government initiatives. Similarly, some markets are dominated by one chain, while others have a high degree of competition with a lot of independent grocery stores. And finally consumers in some markets are more price-sensitive, while in others the degree of product availability or service level is the key competitive variable. We build a multi-echelon arborescent supply chain model that includes heterogeneous customers each making optimal perishable inventory replenishment timing and level decisions in the face of demand uncertainty, as the lowest tiers. Competing grocery stores are the next tier, their demand arises as the superposition of the stochastic order processes of customers, and they themselves manage store inventories. We use these model to compute food waste and its dependence on store density and market structure. My analysis reveals that, independent of the market structure, denser grocery store networks result in higher food waste at the store level, but lower consumer food waste, in contrast with the conventional wisdom that higher level of price competition would lead to higher consumer food waste due to resulting lower price of groceries. The conventional logic does not take into account the reduction in food waste due to increased convenience of grocery shopping. Overall, denser grocery store networks have lower food waste as consumer-side waste is substantially higher than waste at the store level. Further, keeping store density fixed, when price is main competitive dimension, market structures with low degree of competition (a single dominant chain) are more environmentally friendly than one with many individual retailers. On the other hand, when the main competitive dimension is service level, a higher degree of competition is preferred. That is, even without changing grocery store density by instilling the “right” competitive structure city governments can influence food waste levels.

Pietro DE GIOVANNI - ESSEC Business School, France

October 3rd 2016

Environmental Collaboration and Process Innovationin Supply Chain Management

Abstract: This paper investigates a dynamic supply chain model in which a manufacturer decides some process innovation investments and a retailer sets the price. The process innovation investments contribute to the development of the environmental performace, which represents our state variable. The environmental performance is damaged by some negative externalities implied by the demand. This generates an interesting operational trade-off between sales and environmental damages that both players seek to solve. We show the overall bene…ts that cooperation in a process innovation program generates for all supply chain members.