Revenue Management under Competition and Uncertainty Tao Yao Assistant Professor The Harold and Inge Marcus Department of Industrial and Manufacturing Engineering Penn State University taoyao@psu.edu
Outline Self-Introduction Project summary: Revenue Management Proposed Research 2007-06-18 2
Self-Introduction Education PhD, Management Science and Engineering, Stanford University, 2005 MS, Engineering Economics System & Operations Research, Stanford University MS, Mathematics, UCLA BS, Mathematics, Peking University Research Interests Methodology: stochastic models, optimization, and game theory Applications: decision making under uncertainty, real options, manufacturing, service and supply chain operations, information and technology management, energy and environment Hobby Soccer 2007-06-18 3
Revenue Management Service packages Revenue maximization Dynamic pricing Demand Management Features Uncertainty Competition Application Fashion, Retail, Air Travel, Hotel, Seasonal Products etc. 2007-06-18 4
Project Summary: Revenue Management under Uncertainty and Competition Strength: Computability Dynamic Game: Variational method Rules of thumb Real time decision support Proposed Work: Forecasting and Optimization Robust optimization Data-driven, Risk Customer Contact Center 2007-06-18 5
Related Fields Dynamic Optimization and Game Control Theory Mathematical Programming Simulation Risk Management 2007-06-18 6
Research Goals, Impacts Comprehensive models, computational techniques Real time data, realistic assumption Application General approach, whole service engineering community Managerial insights, revenue management, randomness, competition 2007-06-18 7
Revenue Management Review McGill and van Ryzin 1999 Transportation Science Bitran and Caldentey 2003 Manufacuting & Service Operations Mangament Boyd and Bilegan 2003 Management Science Elmaghraby and Keskinocak 2003 Management Science Chiang, Chen and Xu 2007 International Journal of Revenue Management Book Talluri and van Ryzin 2004 The Theory and Practice of Revenue Managment Phillips 2005 Pricing and Revenue Optimization 2007-06-18 8
Revenue Management Pricing Auctions Capacity control (inventory) Overbooking Forecasting Economics Customer behavior and perception Techniques Competition and alliance 2007-06-18 9
Classical example: newsvendor Profit maximization, stock decision, uncertain demand Porteus 1990 Petruzzi and Dada 1999 2007-06-18 10
Seller 1 π 1 Seller i π 2 Seller f π f Market Period 1 Period t Period T 2007-06-18 11
Assumptions Perfect information (imperfect information) Demand is deterministic (uncertainty via learning, robust optimization, data driven) Single product (network) Single resource (network) Single period (multiple period, continuous time) Sellers optimization, pricing (resource allocation) Single seller (game) Complete market, risk neutral (incomplete market, risk averse) 2007-06-18 12
Questions How should sellers price the product and allocate resource with competition? What are the equilibrium prices in the market? How to handle demand uncertainty? How to handle demand learning? How to handle risk preference? 2007-06-18 13
Proposed Research: Revenue Management under Uncertainty and Competition Forecasting and Optimization Robust optimization Data-driven, Risk Management 2007-06-18 14
Goals Modeling of competition and dynamics Modeling of demand uncertainty and robust optimization Modeling of primitive data and risk attitude Numerical study Application 2007-06-18 15
Forecasting and optimization Demand uncertainty Demand distribution and parameters Revealed over time High frequency of data Information technology Rich information Simultaneously forecast the demand and optimize the pricing strategy. 2007-06-18 16
Literature Summary Articles Decisions Competition (game) Multi period (dynamic) Stochas -ticity Single / multi service Learning Features Berstimas & de Boer (2005) Price & Allocation No Yes Direct Multi No Monopoly Perakis & Sood (2005) Price & Allocation Yes Yes Indirect (robust opt) Multi Yes Linear demand, Learning algorithm; quasi-vi formulation Kachani, Perakis & Simon, 2004 Price only Yes Yes Direct Single Yes Linear demand, MPEC Friesz, Mookherjee, Rigdon, (2005) Price & Allocation Yes Yes Direct Multi No Optimal control/vi perspective Kwon, Friesz, Mookherjee, Yao, Feng (2006) Price & Allocation Yes Yes Direct Multi Yes Discrete time Kalman filter, Optimal control/vi perspective This work (2007) Price & Allocation Yes Yes Direct Multi Yes Continuous time Kalman filter, Multiplicative demand, Simultaneous forecasting and optimization, Optimal control/vi perspective 2007-06-18 17
Literature Review Dynamic Optimization and Game Differential Variational Inequalities. Pang and Stewart (2003), Friesa et al. (2006) Friesz et al. (2005) consider joint pricing and resource allocation in network revenue management markdown optimization with known demand dynamics and parameters. Kalman Filter Kwon, Friesz, Mookherjee, Yao and Feng(2006) present discrete Kalman-Filter model to forecast the demand and a differential variational inequality model for pricing the service. We also propose an algorithm based on a gap function to efficient computing the optimal pricing strategies. 2007-06-18 18
Sellers Decentralized Problem Sellers Maximize profit over whole time horizon Decisions Prices Allocation of capacity Constraint Demand dynamics Learning dynamics Bounds on price Bounds on capacity Bounds on demand 2007-06-18 19
Proposed Research Our objective in this research is to develop pricing models which simultaneously forecast demand and optimize the pricing strategy under uncertainty. More specifically, we propose a continuous time estimation of parameters using Kalman Filter and markdown dynamic pricing optimization model. 2007-06-18 20
Current Status We present a differential variational inequality model and an algorithm based on a gap function. We have described the dynamics of demand as a continuous time differential equation based on an evolutionary game theory perspective. Realized sales data are refined on a discrete time scale and used to obtain estimates of parameters that govern the evolution of demand. 2007-06-18 21
Numerical Example (Competition) Revenue Changes 2007-06-18 22
Looking Forward Combine markdown optimization with continuous time parameter estimation. Develop and experiment nonlinear estimation for continuous time system with discrete measure by extending Kalman Filter method for complex revenue management models using insights gained from the discrete time of the model. Fright Service Network Service lever guarantee, product differentiation Future contract market and spot market Both standard analytical approach and computational (numerical) approach 2007-06-18 23
Robust Optimization Literature: Soyster (1973) Ben-Tal and Nemirovski (1998, 1999, 2000) Bertsimas and Sim (2002) Unknown data distribution, parameters uncertain, within an uncertainty set. Stochastic dynamic programming, curse of dimensionality Robust optimization, optimal payoff, robust within the uncertainty set (Bertsimas and Thiele 2004) 2007-06-18 24
Robust Optimization Robust optimization Uncertainty Competition Perakis and Sood 2006 Adida and Perakis 2006 2007-06-18 25
Risk Management, Data-Driven Imperfect market, Risk preference Review, Van Mieghem 2003 Utility, Lau 1980, Eeckhoudt et al. 1995 Data-driven, Bertsimas and Thiele 2005 Risk Measure, Brown, Ben-Tal, & Bertsimas 06 Data-driven, profit maximization, risk management, network revenue management, competition Risk averse Dynamic game Data information 2007-06-18 26
Risk Management, Data-Driven Imperfect market, Risk preference Review, Van Mieghem 2003 Utility, Lau 1980, Eeckhoudt et al. 1995 Data-driven, Bertsimas and Thiele 2005 Data-driven, profit maximization, risk management, network revenue management, competition Risk averse Dynamic game Data information 2007-06-18 27
Customer Contact Center Penn State SEE Board, Industry Data, Call Center Naren Gursahaney, president of Tyco EPS John J. Brennan, chairman and CEO of ICT group Service Engineering Telephone Call Venters: Tutorial, Review and Research Prospects (Gans, koole, Mandelbaum, 2003) Statistical Analysis, Queueing Science (Brown, Gans, Mandelbaum, Sakov, Shen, Zeltyn, Zhao, 2005) Ticket Queue (Xu, Gao, Ou, 2006) 2007-06-18 28
Contributions Model for competitive pricing and resource allocating in a multi-period, oligopolistic market DVI framework Extended to demand learning Discrete time Kalman filter Continuous time estimation Extended to robust optimization Extended to data driven, risk management Form of equilibrium pricing rules. 2007-06-18 29
Conclusion Project Summary: Revenue Management under Uncertainty and Competition Strength: Computability Rules of thumb Real time decision support Proposed work: Forecasting and Optimization Robust optimization Data-driven, Risk Customer Contact Center 2007-06-18 30
Thank you! 2007-06-18 31
Overview of Service Science, Management and Engineering The scope of Service: Service sector is very crucial in our lives. It includes transportation, information, banking, insurance, real-estate, medical, education, government, wholesale and retail trade, etc. The importance of Service to economics: The service industry has grown to dominate developed economies. In USA, 80% of GDP was from the service sector in 2005. 2007-06-18 32
Overview of Service Science, Management and Engineering Service Science, Management, and Engineering (SSME) is the application of science, management, and engineering disciplines to service sector. Service is the least-studied part of the economy. SSME calls for actions from academia. 2007-06-18 33
Robust Optimization Linear demand Robust Constraint 2007-06-18 34