- Wednesday 28 Aug 2019, 09:15 - Thursday 29 Aug 2019, 13:00
- Spoken Language
- Santander (M2-11)
- Van der Goot Building
- Campus Woudestein
Dinner in Restaurant 'In de Rustwat' https://idrw.nl/
This workshop programme will provide basic and advanced methodologies for revenue management problems. The course aims at integrating different disciplines to solve both theoretical and practical concerns that arise in the design and development of systems for network revenue management, pricing, and operation.
The lectures will cover operation research methods, optimization algorithms and econometric models and machine learning. Keynotes will be given on timely topics by well-known researchers in the field. This workshop is open to Master, PhD. and post-doctoral students, as well as researcher and practitioners. This opportunity will facilitate professional networking and exchange of ideas about the theory and practice of research in revenue management.
- Network Revenue Management
- Pricing models
- Choice-based optimization
- Discrete and large-scale optimization
- Decision support systems for revenue management
- Case studies
- Dr. Kalyan Talluri
- Dr. Ilker Birbil
- Dr. Arnoud den Boer
- Dr. Shadi Sharif Azadeh
- Dr. Luce Brotcorne
- Dr. Christiane Barz
- Dr. Catherine Cleophas
- Dr. Arnoud den Boer
- Dr. Joern Meissner
- Dr. Shadi Sharif Azadeh
Send us your abstract
There are four time slots available for students’ presentations. We highly encourage PhD students or postdocs who are working on the related topics to present their research in this workshop. Interested students should send their abstracts (1 page, Times New Roman with 1.5 space between lines), and titles to firstname.lastname@example.org.
Submission deadline is 14 August 2019.
|09:30-09:45||Welcome and introduction|
|09:45-10:45||Dr. Kalyan Talluri|
|12:00-13:30||Lunch on Campus|
|14:45-15:45||Dr. Shadi Sharif Azadeh|
|19:00||Dinner venue (In Restaurant In de Rustwat)|
|09:00-10:00||Dr. Ilker Birbil|
|10:15-11:15||Dr. Arnoud den Boer|
|11:30-12:30||Dr. Alba V. Olivares-Nadal|
Capacity sharing is arguably one of the best approaches to obtain sustainable and cost-effective use of resources. There exist various mathematical programming tools for optimal resource allocation. However, we still need to convince multiple parties to agree upon sharing their capacities. Even if they give their consent for collaboration, they also rightfully raise their concerns regarding the privacy of their sensitive data used in optimization models. In this work, we present the first model that considers the data privacy in bid-price control for network revenue management.
Our analysis makes use of several previous privacy studies based on random transformations of the problem data. However, our focus on the bid-prices allows us to present new results about the privacy of the dual solutions. It is well-known that the transformed problem takes a long time to solve due to the loss of sparsity structure of the original problem. To overcome this issue, we propose a new solution approach that produces a random transformation which is likely to result in a sparse problem.
Finally, we support our results with a computational study on a set of revenue management problems where the network structure is taken from a major European firm.
Cookie-Cutter Competition? Non-Price Strategies of Multiproduct Firms under Uniform Pricing
In this paper, we study how multiproduct firms compete using non-price strategies - namely, quantity, promotions, and rationing - in an industry where all firms charge the same price. Prior literature observes uniform pricing in retail chains (Gentzkow and DellaVigna, 2017), online music (Shiller and Waldfogel, 2011), movie industry (Orbach and Einav, 2007), soft drinks (McMillan 2007), and rental cars (Cho and Rust, 2010). Under classical uniform pricing, prices do not vary across regional markets. But significant price variation exists between firms in a market (Gentzkow and DellaVigna, 2017). We observe uniform pricing to be not a merely within-firm phenomenon but an industry-level one, which, to our knowledge, has not been examined in the previous literature.
Our study is set in India’s biscuit industry. We use Nielsen’s monthly data on the prices, quantities, sales revenues, and promotions of nearly 15,000 SKUs and 800 firms during April 2014 to March 2015. The data are disaggregated into 12 market segments - such as Cream, Crackers, and Marie - and 40 urban and rural regional markets. We augment these SKU-level data with CMIE Prowess data on firm financials. The richness of our data allow us to examine non-price strategies in detail.
We find that products with one standard deviation higher productivity offer, on average, 1.5% more quantity for the same price. Firms also compete by offering volume promotions for more productive products.
We find greater levels of product availability and productivity-induced competition in urban areas compared to rural areas, implying uneven welfare effects. We show that deviating from uniform pricing can improve welfare of rural consumers. Overall, our results indicate that using non-price strategies, more productive products appear to gain market share, implying that competition thrives under the veil of uniform pricing. In the paper we propose a quantity-based measure of product-level productivity that controls for the biases related to input measurement, simultaneity and product scope of the firm. Our results are robust to alternative methods of estimating product-level productivity (De Loecker et al., 2016; Dhyne et al., 2017).
Our paper primarily contributes to the marketing literature in emerging economies, where differences in culture, socioeconomic conditions, institutions, and infrastructure result in novel marketing strategies (Prahalad, 2005; Narasimhan et al., 2015; Sudhir and Talukdar, 2015; Qian et al., 2015). Uniform pricing is usually motivated by consumer’s perceived fairness (Kahneman et al., 1986a and 1986b), and can, in theory, increase producer surplus (Chen and Cui, 2013). The potential negative welfare effects for rural consumers relative to urban consumers indicate that bottom-of-the-pyramid strategies combined with fairness-based pricing strategies do not always benefit the poor.
Co-Author : Ajay Bhaskarabhatla
A demand-based optimization approach to find market equilibria in oligopolies
Oligopolistic competition occurs often in transportation as well as in other markets due to reasons such as barriers to entry, limited capacity of the infrastructure and external regulations. In transport oligopolies, suppliers are profit maximizers and take decisions that are influenced both by the preferences of the customers, who want to purchase one of the services on the market, and by the strategies of the competitors.
In our work, the preferences of the customers are modelled at a disaggregate level using discrete choice models and are embedded in each operator's optimization problem. Using a disaggregate approach that accounts for heterogeneous demand allows to better model supply-demand interactions. Competition among market players is modelled explicitly as a non-cooperative game. The result is a multi-leader-follower game. We present a MIP model inspired by the fixed-point iteration algorithm that find Nash equilibrium solutions of finite games. Numerical experiments show that the computational performance of the model depends on the type of decision variables used to model supply strategies and on the discrete choice model used to describe demand. The nonlinear formulation is non-convex and becomes intractable when many discrete variables are introduced, while the linearized formulation is convex but combinatorial due to the nature of the proposed simulation framework.
We propose an algorithmic framework in which candidate equilibrium solutions are first found by means of heuristic blocks and constitute the initial restricted sets of strategies in the fixed-point MIP model, which is used to find subgame equilibrium solutions. Subgame equilibria are checked against the original game by solving best-response problems, and new candidate strategies can be added to the next iterations of the fixed-point MIP model using a column-generation-like approach.
Considering outlier detection to identify extraordinary demand events for quantity-based revenue management
Most quantity-based airline revenue management systems rely on forecasting expected demand to prepare revenue-optimal capacity allocations. Inaccurate demand forecasts result in non-optimal allocations and hence, hurt revenue. The majority of airline revenue management systems also allow room for analysts to compare the accumulated bookings against the forecasts and to intervene, if they deem the demand forecast to be inaccurate. However, human analysts are fallible and do not always intervene correctly. Therefore, increasing decision support for analysts is needed. Building automated procedures for detecting outliers, where demand differs from expectations due to systematic market shifts, is an open challenge.
In a set of controlled simulation scenarios, we let demand systematically deviate from the general level. After transforming demand into bookings via a minimal revenue management simulation, we apply a variety of outlier detection techniques to determine whether the number of bookings can be classified as either normal or abnormal. We show that standard approaches to outlier detection, including a Euclidean distance-based approach, K-means clustering, and tolerance intervals, often fail to correctly identify instances where demand differs from forecasts. This motivates the need for developing and extending new approaches. We propose a novel method which combines methodology from functional data analysis for outlier detection with forecasting techniques, in order to be able to identify outlying demand in both historical and online settings. We show that identifying instances of outlier demand using our methodology, and adjusting the forecast in a timely fashion, has the potential to significantly increase revenue.
Co-Authors: Catherine Cleophad, Adam Sykulski and Florian Dost
Exponential Approximation in Dynamic Bid Pricing
The notion of bid-price control can be used in revenue management set-tings where the supply of resources is ﬁxed and customer requests arrive over a ﬁnite time horizon to consume various resource conﬁgurations. The arriving requests must either be accepted or rejected, with the objective of maximizing expected proﬁt over the time horizon. The basic idea of bid-price control is simple: Accept the request if the revenue earned exceeds the value of the resources consumed as measured by bid prices. In practice, major airlines have used bid-price control policies for deciding when to open and close customer fare classes for sale.
Although the system under control is dynamic, tradional models only compute static bid prices. Recently, dynamic bid pricing has been addresed by approximating the value function by a weighted sum of given basis func-tions. Aﬃne, piecewise linear, or quadratic functions have been suggested as basis, but the aﬃne case is by far the most widespread approach used. However, there is no reason to assume those choices will lead to a tight ap-proximation of the value function. In our work we use a sum of weighted exponentials, which ensures the convergence to the real value function. As a consequence, our method will provide a tighter upper bound than previous approximations. In order to estimate the parameters of our basis functions, a highly dimensional problem must be solved. To deal with this issue, we propose a row-generation algorithm that involves repeatedly solving non-convex integer programs.
Choice Driven Dial-A-Ride Problem for On-Demand Mobility Systems
This paper presents a choice driven dial-a-ride problem. For the first time, we introduce a unique mathematical model for the DARP that benefits from assortment optimization and dynamic pricing to better allocate resources to demand. We measure the attractiveness of each ride option with a utility function. The optimization problem is then defined by constructing a travel menu that maximizes the expected profit. The model is evaluated both from revenue performance and transportation point of view.
Reviews, Biases and Six-sigma
Reviews for products and services written by previous consumers have become an influential input to the purchase decision of customers. Many service businesses monitor the reviews closely, for feedback as well as detecting service flaws, and they have become part of the performance review for service managers with rewards tied to improvement in the aggregate rating. It is therefore of great importance to understand how much the public ratings reflect true quality of the product or service. Many empirical papers have documented a bias in the aggregate ratings, arising due to various causes: an inherent self-selection in writing reviews, as well as customers' bounded rationality in evaluating previous reviews. While there is a vast empirical literature analyzing reviews, theoretical models that try to isolate and explain the ratings bias are relatively few, and most are based on a rational Bayesian learning assumption on the part of consumers. However, writing a review requires some effort in itself and it seems unlikely that consumers would make the considerable effort to do a Bayesian update of their beliefs before making purchases. Assuming consumers simply substitute the average rating they see as a proxy for quality, we give a precise characterization of the effect of biases in the ratings arising from two sources (i) an acquisition bias, when consumers confound ex-ante innate preferences for a product or service with ex-post experience and service quality and do not separate the two, and (ii) an under-reporting bias, when consumers with extreme positive or negative ratings are more likely to write reviews than consumers with moderate product ratings. We develop a parsimonious choice model for consumer purchase decisions and show that both sources lead to an upward bias. Based on our theoretical characterization, we give two important practical applications for a service firm: (a) estimation of true process quality and its variability for the firm's internal performance, Six sigma and quality programs (b) effect on pricing and assortment decisions of the firm, when potential customers purchase based on the biased ratings. Our results give insight into how quality, prices and customer feedback are intricately tied together for service firms.
Dynamic Pricing with Demand Learning and Reference Effects
We consider a seller’s dynamic pricing problem with reference effects: the phenomenon that sales is not only influenced by the current price, but also by a so-called reference price constructed in the minds of potential customers based on the seller’s price history. There is substantial empirical evidence that customers are loss averse: that means that the demand reduction when the selling price exceeds the reference price is larger than the demand increase when the selling price falls behind the reference price by the same amount. Consequently, the expected demand as a function of price has a time-varying “kink” and is not differentiable everywhere. The seller neither knows the underlying demand function nor observes the time-varying reference prices. In this setting, we show that neglecting the reference effect can be very costly. We design and analyze a policy that (i) changes the selling price very slowly to control the evolution of the reference prices, and (ii) gradually accumulates sales data to balance the trade-off between learning and earning. We prove that, under a variety of reference-price updating mechanisms, our policy is asymptotically optimal; i.e., its T-period revenue loss relative to a clairvoyant who knows the demand function and the reference-price updating mechanism grows at the smallest possible rate in T. We also extend our analysis to the case of gain-seeking customers, and show that, surprisingly, the `difficulty of the learning problem’, measured by the asymptotically optimal growth rate of the regret, is parameter-dependent.
Ilker Birbil is a faculty member in Erasmus University Rotterdam at the Econometric Institute, where he serves as an endowed professor for the Chair in Data Science and Optimization. He received his PhD degree from North Carolina State University, Raleigh, USA. He then worked for two years as a postdoctoral research fellow in Erasmus Research Institute of Management, Rotterdam, The Netherlands.
His research interests include parallel and distributed optimization in machine learning, algorithm development for large-scale optimization problems, data science, revenue management, stochastic dynamic programming. Lately, he is very much interested in data privacy in decision making.
Venue on 28 August 2019
There is a fee of 40 euros, and registration is necessary for participation. The registration is closed. It is no longer possible to register for the workshop.
Bank Transfer: reference: 12020001.001.111 Ectrie – Workshop Revenue Management & Pricing and your name.
Erasmus Universiteit, Erasmus School of Economics
P.O. Box 949, 3000 DD Rotterdam, the Netherlands
Interested students should send their abstracts and titles to email@example.com no later than 1 August 2019.