This dissertation consists of three papers in revenue management: on-line assortment optimization with reusable resources, spatial distribution of surge price under incentive compatible assignment for drivers and optimal price rebates for demand response under power flow constraints. In Chapter 2, we study an on-line assortment optimization problem of substitutable products with fixed reusable capacities. At any time, a potential user with her preference model (possibly adversarially chosen) arrives to the selling platform and the platform offers a subset of products from the available set of products to the user. The user selects a product with probability given by her preference model, uses it for a random duration, which is distributed according to a distribution that only depends on the product selected, and generates revenue to the seller. The revenue contribution depends on the product selected and the actual usage time of this user. The goal of the seller is to find a policy for determining the assortment offered to each arrival to maximize the expected cumulative revenue over a time horizon. We find that a simple myopic policy offering the available assortment that maximizes the expected revenue from a single user at her arrival time provides a good approximation for the problem. In particular, we show that the myopic policy is $1/2$-competitive, i.e., the expected cumulative revenue of the myopic policy is at least $1/2$ times the expected cumulative revenue of an optimal clairvoyant policy that has full information about the adversarially chosen user sequence, including their preference models and arrival epochs. The proof is based on partitioning the expected revenue of optimal clairvoyant policy into two parts and a coupling argument that allows us to bound the two parts in terms of the expected revenue of the myopic policy. In Chapter 3, we study the surge pricing problem on a ride sharing platform when there is a demand shock to the traffic network. The goal of the platform is to maximize the revenue by setting the prices over the network and the assignments between drivers and riders. In particular, we model the city as a continuous two dimensional network with exogenous arrivals of baseline riders, available drivers and demand shocks. We consider the demand shock only exists in a short time scale, so the rider chooses to request the ride or not depending on their willingness to pay and the price quoted to them, and the driver accepts any price to provide service. Since drivers can see the price distribution on driver app, they only accept the assignment from the locations that are incentive compatible for them. Thus, the price change at one location may affect the operations over the network and the platform must consider the incentive of drivers when assigning them. We develop a model for this surge pricing problem and show the structural properties of an optimal solution. Once the prices at the location with demand shock is determined, we can determine the optimal prices on other part of the network. Then, the optimal assignments between riders and drivers can be determined analytically. The surge pricing problem reduces to one that only depends on the price at the location with demand shock. We then extend our model by including strategic behavior of riders, using throughput as objective, dealing with multiple demand shocks, un-constraining the price and considering movement time. We also conduct numerical experiments to study the properties of the model which can not be explored analytically. In Chapter 4, we study the demand response problem of computing price rebates to offer to the customers to reduce the consumption in the presence of power flow constraints and transmission losses on the distribution grid. In particular, we employ alternating current power flow model for the power flow constraints with transmission loss. However, the demand response problem with alternating current power flow constraints is known as a non-convex problem, which is in-tractable to solve. To overcome this, we apply a semi-definite relaxation of alternating current power flow model to obtain a convex approximation for the problem. At the same time, to handle the uncertainty in the power reduction of customers, we use sample average to approach the expected cost and linear injection approximation to estimate the impact of uncertainty in the power reduction. Based on these relaxations and approximations, we propose an efficient iterative heuristic to solve the near-optimal offer price under alternating current power flow constraints and transmission losses. We conduct a substantial amount of numerical tests on our heuristic and compare its performance with other popular models. The result shows that our iterative heuristic leads to a significant reduction in the rebates that one needs to offer to shed a certain demand than the solution which does not consider full transmission loss in its model.

One fundamental problem in revenue management that arises in many settings including retail and display-based advertising is assortment planning. Here, the focus is on understanding how consumers select from a large number of substitutable items and identifying the optimal offer set to maximize revenues. Typically, for tractability, we assume a model that captures consumer preferences and focus on computing the optimal offer set. A significant challenge here is the lack of knowledge on consumer preferences. In this thesis, we consider the multinomial logit choice model, the most popular model for this application domain and develop tractable robust algorithms for assortment planning under uncertainty. We also quantify the fundamental performance limits from both computational and information theoretic perspectives for such problems. The existing methods for the dynamic problem follow ``estimate, then optimize'' paradigm, which require knowledge of certain parameters that are not readily available, thereby limiting their applicability in practice. We address this gap between theory and practice by developing new theoretical tools which will aid in designing algorithms that judiciously combine exploration and exploitation to maximize revenues. We first present an algorithm based on the principle of ``optimism under uncertainty'' that is simultaneously robust and adaptive to instance complexity. We then leverage this theory to develop a Thompson Sampling (TS) based framework with theoretical guarantees for the dynamic problem. This is primarily motivated by the growing popularity of TS approaches in practice due to their attractive empirical properties. We also indicate how to generalize the TS framework to design scalable dynamic learning algorithms for high-dimensional data and discuss empirical gains of such approaches from preliminary implementations on Flipkart, a large e-commerce firm in India.

Revenue management (RM) is the science of selling the right product, to the right person, at the right price. A key to the success of RM, which now spans a broad array of industries, is its grounding in mathematical modeling and analytics. This dissertation contributes to the development of new RM tools by: (1) exploring some fundamental tradeoffs underlying any RM problems, and (2) designing efficient algorithms for some RM applications. Another underlying theme of this dissertation is the modeling of customer preferences, a key component of any RM problem. The first chapters of this dissertation focus on the model selection problem: many demand models are available but picking the right model is a challenging task. In particular, we explore the tension between the richness of a model and its tractability. To quantify this tradeoff, we focus on the assortment optimization problem, a very general and core RM problem. To capture customer preferences in this context, we use choice models, a particular type of demand model. In Chapters 1, 2, 3 and 4 we design efficient algorithms for the assortment optimization problem under different choice models. By assessing the strengths and weaknesses of different choice models, we can quantify the cost in tractability one has to pay for better predictive power. This in turn leads to a better understanding of the tradeoffs underlying the model selection problem. In Chapter 5, we focus on a different question underlying any RM problem: choos- ing how to sell a given product. We illustrate this tradeoff by focusing on the problem of selling ad impressions via Internet display advertising platforms. In particular, we study how the presence of risk-averse buyers affects the desire for reservation con- tracts over real time buy via a second-price auction. In order to capture the risk aversion of buyers, we study different utility models.

No description available.

This dissertation analyzes the economic incentives involved in three distinct supply chain and revenue management decision environments. The first study examines the adoption of the percent deviation contract in a supply chain to induce the buyer to share some of the demand risk in an environment in which the buyer would typically place her order when she has full knowledge of the customer demand levels. The subgame-perfect Nash Equilibrium decisions are characterized, and the percent deviation is shown to achieve full supply chain channel coordination in cases where a simpler contract cannot. Pareto-improving examples based on industry demand data are presented and discussed. The second section considers a revenue management problem for sports and entertainment organizations. Given that the organization starts the selling season by offering ticket packages exclusively, the optimal time during the selling season for the organization to begin selling individual-event tickets is derived. Extensions of the base model are developed to include multiple ticket packages and heterogeneous ticket packages. The model is illustrated using empirical data sets obtained from the Georgia Tech Athletic Department and the Atlanta Symphony Orchestra. The third section develops a model of vendor-controlled category management in which vendors are in charge of the stocking and assortment decisions for a given amount of shelf space at a vendor when the retailer retains control over the retail price. The subgame-perfect Nash Equilibrium strategies for two vendors and a single retailer are analyzed, and a revenue-sharing contract is shown to coordinate the channel when the vendors can produce multiple brands in a given product category and shelf space is sufficiently large or small.

This dissertation includes four independent essays. Essay one (chapter two) considers a two-echelon, two-supply chain (SC) system in which manufacturers supply a generic product to their exclusive retailers, who then use service level and retail price to compete for heterogeneous consumers. We question: how do varied consumer preferences get reflected not only in differentiated products/services, but through them to the choice of SC structure that delivers them? We find that SCs can strategically manipulate the product/service strategy and SC structure to hedge themselves from horizontal competition. The key finding is that in a market where consumers have stronger diminishing marginal utility on service, then less differentiated products/services will be observed, and only decentralized supply chains can be the market equilibrium. This is in contrast to the well-known result in marketing that choosing vertical integration is always a Nash equilibrium, and that choosing decentralization can only be a Nash equilibrium when product substitutability is high. Essay two (chapter three) explores the classical revenue management problem in a competitive context, with both price and seat inventory competition. The main question is how should management make strategic marketing (pricing) and operational (seat allocation) decisions in such a competitive market? Do the conventional approaches (models and algorithms based on a monopoly market) give us the appropriate strategies? We find that in a market where price competition dominates, managers should set a lower price and safety protection level for full fare customers than in a monopoly or alliance market. In a market where seat inventory competition dominates, managers should set a higher price and safety protection level than a monopoly or alliance would. Interestingly, in a market where the two levels of competition are more evenly matched, managers should set a lower price and a higher safety protection level than a monopoly. We also explore the effect of the degree of competition and the market structure on the strategic decisions, and whether there is a first adopter advantage or second adopter disadvantage with revenue management. Essay three aims to extend the understanding of the Newsvendor model to a competitive framework. In a market with both price and inventory competition, newsvendors can gain customers with price and secure the sales with availability. We find that the newsvendors should adjust their inventory (safety stock or total inventory) and pricing strategies responsively to the nature of the competitive market. The profits of the newsvendors and their suppliers are also different under different competitive contexts. Both the Nash equilibrium strategy and the players' profits are influenced by the demand correlation and variability, but in different ways under different competitive scenarios. These observations provide some theoretical basis for the strategic selection made by newsvendors operating in certain competitive markets. Essay four (chapter five) explores the issue of competitors cooperating. It is a commonplace observation that even the most competitive firms often find it in their best interests to cooperate. An example of cooperation in operations management is when two supply chains agree in advance to transship or 'pool' surplus product for use by another. The alternative is to let their customers switch unsatisfied demand to a competitor. Which is preferable, and how does such a preference depend on the many parameters, prices, the nature of competition, the degree of competition, wholesale prices etc? To get answers, we study a stylized model under three market environments: a market with an exogenous retail price, an endogenous retail price, and with price competition. The summary answer is that strong price competition between substitutable goods should lead to caution in signing transshipment contracts. But with little price competition and particularly where retailers are free to set the transshipment price, then transshipment is probably the way to go. We also address the issue of an optimal transshipment price in each scenario, and compare the Nash equilibrium strategies between competing and transshipping.

Die effiziente Nutzung beschränkter Kapazitäten (z.B. Flugzeugplätze, Hotelzimmer) erweist sich für Anbieter als kritischer Erfolgsfaktor. Zur Steuerung der Buchungsanfragen wird daher Revenue Management angewandt. Um langfristig profitable Kundenbeziehungen aufzubauen, sollten auch kundenwertbezogene Informationen (den langfristigen Wert des Kunden für den Anbieter repräsentierend) einbezogen werden. In der Folge sieht sich der Anbieter vielen Unsicherheiten gegenüber. Da die Berücksichtigung von Unsicherheiten die Effizienz der Steuerungsentscheidungen und damit die Erlöshöhe beeinflusst, widmet sich die Dissertation zunächst der Analyse und Systematisierung der unsicherheitsbasierten Problemfelder und nachfolgend der Erweiterung traditioneller Steuerungsmethoden, die in Simulationsstudien evaluiert werden. Die Intention des Beitrags ist es, das Forschungsvorhaben in seiner Zielstellung und Methodologie nachvollziehbar darzulegen.

Zugl.: Braunschweig, Techn. Univ., Diss., 2006

Zugl.: München, Techn. Univ., Diss., 2007 / Hergestellt on demand

Diss. Univ. St. Gallen, 2008.