Modern marketplaces, enabled by recent advances in information technology, are becoming ever more complex, dynamic and interconnected. Customers are increasingly more informed about the choices available to them and are thus able to make better decisions. For firms, this presents both opportunities and challenges. On the positive side, the firm has a bigger than ever array of tools at its disposal that enable it to better sell to its customers. Such tools are both computational (pricing algorithms, customer targeting data, social networking information) and economical (dynamic mechanisms) in nature. The challenging aspect for the firm is that modern technology also allows consumers to easily learn the opinions of fellow customers and to be strategic in their decision-making. Having such strategic and networked customers is not necessarily a net negative for the firm, but it certainly makes the firm's problem of how to sell its products a more complex one.
This project aims to develop a framework for understanding how to sell goods in markets that are fundamentally dynamic and interconnected. The project will research what mechanisms are revenue-optimal (or near-optimal, while being simple and computationally tractable) in dynamic settings where customers can learn from each other. The PIs will study the effect of network learning on the firm's behavior and whether it forces the firm to share some of the costs associated with consumer learning. The project will also focus on how the social networkstructure affects the optimal mechanism and will try to understand whether such network learning effects lead to lower (or higher) revenue for the firm and aggregate consumer welfare.
Fundamental research in the dynamics of networked markets provides deeper knowledge to a broad audience that includes firms, consumer groups and regulators. Curriculum development at the interface of operations research, information systems and computer science will benefit from this research experience.
This research project led to five papers on the topics of dynamic mechanism design, dynamic pricing, new product launch strategies and the diffusion of information in social networks. The main contributions of these papers are: 1. In "Intertemporal Price Discrimination: Structure and Computation of Optimal Policies," by Omar Besbes and Ilan Lobel, we develop a new framework for determining intertemporal prices that are optimal in the presence of consumers who strategically time their purchases. We provide a polynomial-time algorithm to compute optimal prices for an arbitrary strategic demand model. We also show that complex seasonal pricing structures, of the kind often observed in practice, emerge solely as a consequence of a price discrimination strategy by sellers, even in the absence of seasonal demand or inventory considerations. 2. In "Optimal Long-Term Supply Contracts with Asymmetric Demand Information," by Ilan Lobel and Wenqiang Xiao, we study the dynamic contracting problem faced by a manufacturer and a retailer that has local demand information. We show there exists an optimal long-term dynamic contract that has a very simple structure, with the retailer being charged a payment that is linear in the number of units ordered, plus a one-time initial fee. 3. In "Optimizing Product Launches in the Presence of Strategic Consumers," by Ilan Lobel, Jigar Patel, Gustavo Vulcano and Jiawei Zhang, we study the product launch strategies of firms developing new technology products. We show that the optimal strategy for such a firm is a policy that alternates between long and short release cycles and that such a policy significantly outperforms a simple single cycle. We also develop a mixed-integer programming technique for computing optimal release cycles in the presence of a heterogeneous customer base. 4. In "Information Diffusion in Networks through Social Learning," by Ilan Lobel and Evan Sadler, we propose a new metric for measuring the outcome of a social learning process: information diffusion. This new metric determines whether the information from the strongest private signals eventually gets diffused throughout the entire network and stands in contrast to the stronger classical metric of information aggregation across agents. We show that social learning always diffuses information in networks that satisfy a minimal connectivity condition when the network structure is common knowledge. The same is not true, however, of social networks where the topology is not commonly known. We also offer a technique that can be used to verify whether information is diffused in a general network and use it to show that information diffuses in a network with preferential attachment. 5. In "Preferences, Homophily, and Social Learning,’ by Ilan Lobel and Evan Sadler, we study the role of preference heterogeneity and homophily in the equilibrium outcome of a social learning process. We show that network density is the key determinant of whether preference diversity enables or hinders social learning, with information aggregation in dense networks being helped by heterogeneity of preferences and aggregation in sparse networks being hindered by it. Network density is also the key determinant of whether homophily is useful for information aggregation in social networks, but with homophily playing a positive role in sparse networks instead.
