About

Xuanming Su is a Professor of Operations, Information and Decisions at the Wharton School of the University of Pennsylvania. His research focuses on operations management, with particular emphasis on omnichannel retailing, service operations involving online and offline self-order technologies, and the design of social comparison effects and fairness in supply chain contracts. He has contributed to the understanding of digital experience optimization, consumer behavior in multi-channel environments, and strategic pricing and capacity allocation. Professor Su has received numerous awards for his scholarly work, including the Best Paper Award in Management Science in 2020, the MSOM Practice-Based Research Award in 2017, and the NSF CAREER Award in 2008-2013. He is actively involved in teaching decision processes and related courses at Wharton, and his research has been featured in various publications and media outlets.

Research topics

• Computer Science
• Machine Learning
• Economics
• Business
• Microeconomics
• Artificial Intelligence
• Marketing
• Advertising
• Mathematical economics
• Internet privacy
• Finance
• Computer network
• Commerce

Selected publications

• Food Ordering and Delivery: How Platforms and Restaurants Should Split the Pie

  Management Science · 2025-07-02 · 5 citations

  articleSenior author

  Food ordering and delivery platforms generate online demand for restaurants and deliver food to customers. In return, restaurants pay platforms a commission, typically a percentage of the order amount. Platforms offer partner restaurants the choice of a range of commission rates, rewarding higher commission payments with featured display slots and discounted delivery fees, both of which stimulate demand. Unfortunately, the current environment is grim: Platforms scurry to cover delivery costs, whereas restaurants gripe about excessive commissions. To understand current practice, we develop a game-theoretic model with a platform and multiple restaurants. Our modeling results highlight two existing problems. (1) Platforms, on their apps/websites, feature restaurants that are located too far away. Because these restaurants do not internalize the platform’s delivery costs, they are willing to choose aggressively high commissions to earn featured display. (2) Platforms charge delivery fees that are too high and set delivery boundaries that are too narrow. This is because they bear the entire burden of delivery but earn only a fraction of food revenues. To solve these problems, we propose a simple fix to existing commission contracts: beyond sharing food revenue (currently done but at high commission rates), platforms and restaurants can also split delivery costs and fees (currently not done). We show that our method attains first-best, that is, maximizes the total pie shared by platforms and restaurants. Using data on a representative city, we numerically show that, on average, our coordinating contract lowers commission rates by 33.3%, lowers delivery fees by 40.4%, increases restaurant profit by 25.0%, increases platform profit in 30.9% of the markets, and increases total profit by 13.3%. We discuss the characteristics of markets that enable our coordinating contract to yield a winning outcome for all parties. This paper was accepted by Karan Girotra, operations management. Supplemental Material: The online appendices and data files are available at https://doi.org/10.1287/mnsc.2023.00435 .

  PublisherDOI

• Food Ordering and Delivery: How Platforms and Restaurants Should Split the Pie

  SSRN Electronic Journal · 2023 · 2 citations

  Senior authorCorresponding
  • Computer Science
  • Business
  • Marketing
  PublisherDOI

• Optimal Pricing and Overbooking of Reservations

  Production and Operations Management · 2021 · 14 citations

  Senior authorCorresponding
  • Computer Science
  • Microeconomics
  • Business

  We study the optimal design of reservations for a firm with limited capacity. The firm faces a random number of customers, each of whom has a random valuation for service. The reservation policy has two components: pricing and overbooking. For the former, the firm charges a reservation fee (at the time of reservation) and a service price (at the time of service). For the latter, the firm imposes a booking limit that caps the number of reservations it sells. Given the firm's reservation policy, customers make reservations in advance and later decide whether to show up. Denying service to reservation holders is costly. We obtain the following equilibrium results. First, when demand is small relative to capacity, the firm's pricing structure relies on reservation fees prepaid in advance, but when demand is large relative to capacity, it relies on payment received upon service. Second, when demand is low and/or predictable, the firm accepts all reservation requests, but when demand is high and/or variable, the firm uses a booking limit.

  PublisherDOI

• A Bayesian Level-<i>k</i> Model in <i>n</i>-Person Games

  Management Science · 2020 · 14 citations

  Senior authorCorresponding
  • Computer Science
  • Artificial Intelligence
  • Mathematical economics

  In standard models of iterative thinking, players choose a fixed rule level from a fixed rule hierarchy. Nonequilibrium behavior emerges when players do not perform enough thinking steps. Existing approaches, however, are inherently static. This paper introduces a Bayesian level-k model, in which level-0 players adjust their actions in response to historical game play, whereas higher-level thinkers update their beliefs on opponents’ rule levels and best respond with different rule levels over time. As a consequence, players choose a dynamic rule level (i.e., sophisticated learning) from a varying rule hierarchy (i.e., adaptive learning). We apply our model to existing experimental data on three distinct games: the p-beauty contest, Cournot oligopoly, and private-value auction. We find that both types of learning are significant in p-beauty contest games, but only adaptive learning is significant in the Cournot oligopoly, and only sophisticated learning is significant in the private-value auction. We conclude that it is useful to have a unified framework that incorporates both types of learning to explain dynamic choice behavior across different settings. This paper was accepted by Manel Baucells, decision analysis.

  PublisherDOI

• New Models of Strategic Customers in the Age of Omnichannel Retailing

  Springer series in supply chain management · 2019-12-14 · 3 citations

  book-chapterSenior authorCorresponding
  PublisherDOI

• Optimal Retail Location: Empirical Methodology and Application to Practice

  Manufacturing & Service Operations Management · 2019-01-01 · 110 citations

  articleSenior author

  We empirically study the spatiotemporal location problem motivated by an online retailer that uses the Buy-Online-Pick-Up-In-Store fulfillment method. Customers pick up their orders from trucks parked at specific locations on specific days, and the retailer’s problem is to determine where and when these pickups occur. Customer demand is influenced by the convenience of pickup locations and days. We combine demographic and economic data, business location data, and the retailer’s historical sales and operations data to predict demand at potential locations. We introduce a novel procedure that combines machine learning and econometric techniques. First, we use a fixed effects regression to estimate spatial and temporal cannibalization effects. Then, we use a random forests algorithm to predict demand when a particular location operates in isolation. Based on the predicted demand and cannibalization effects, we solve the spatiotemporal integer program using a quadratic program relaxation to find the optimal pickup location configuration and schedule. We estimate a revenue increase of at least 51% from the improved location configuration and schedule. The online appendices are available at https://doi.org/10.1287/msom.2018.0759 .

  PublisherDOI

• A Model of Rational Retrials in Queues

  Operations Research · 2019-10-29 · 28 citations

  article

  Customers often wait in queues before being served. Because waiting is undesirable, customers may come back later (i.e., retry) when the queue is too long. However, retrial attempts can be costly as a result of transportation fees and service delays. This paper introduces a framework for rational retrial decisions in stationary queues. Our approach accommodates retrials in queues by replicating the Naor's model [ Naor P (1969) The regulation of queue size by levying tolls. Econometrica 37(1):15–24.] repeatedly over time periods. Within each period, we study an observable queue in which customers make rational state-dependent decisions to join, balk, or retry in a future period. We focus on a stationary environment where all arrivals, including new and retrying customers, will face the steady-state distribution of the system in equilibrium. Equilibrium analysis on customers’ decision making is necessary, as they choose optimal strategies corresponding to the stationary queueing dynamics that are in turn determined by their decisions. We characterize the equilibria in both stable and overloaded systems. We find the following: (1) Compared with a system without retrials, the additional option to retry can hurt consumer welfare. (2) Compared with the socially optimal decisions, surprisingly, self-interested customers retry insufficiently (they join overly long queues) when the retrial cost is low and retry too often when the retrial cost is high. (3) Self-interested (retrial) customers can generate positive externalities by smoothing workload over time.

  PublisherDOI

• New Functions of Physical Stores in the Age of Omnichannel Retailing

  Springer series in supply chain management · 2019-01-01 · 15 citations

  book-chapterSenior author
  PublisherDOI

• Omnichannel Service Operations with Online and Offline Self-Order Technologies

  Management Science · 2017-07-24 · 173 citations

  articleSenior author

  Many restaurants have recently implemented self-order technologies across both online and offline channels. Online technology, through websites and mobile apps, allows customers to order and pay before coming to the store; offline technology, such as self-service kiosks, allows store customers to place orders without interacting with a human employee. In this paper, we develop a stylized theoretical model to study the impact of self-order technologies on customer demand, employment levels, and restaurant profits. Our main results follow. First, customers using self-order technologies experience reduced waiting cost and increased demand, and moreover, these benefits may even carry over to customers who do not use these technologies. Second, although public opinion suggests that self-order technologies facilitate job cuts, we find instead that some firms should increase employment levels, and, paradoxically, this recommendation holds for firms with high labor costs. Finally, we find that firms should implement online (offline) self-order technology when customers have high (low) wait sensitivity. The supplementary appendix is available at https://doi.org/10.1287/mnsc.2017.2787 . This paper was accepted by Serguei Netessine, operations management.

  PublisherDOI

• Reservation Policies in Queues: Advance Deposits, Spot Prices, and Capacity Allocation

  Production and Operations Management · 2017-11-17 · 44 citations

  articleSenior authorCorresponding

  At firms such as restaurants, customers either make reservations in advance or join queues on the spot. Reservation holders may not show up, and walk‐ins have to wait. Using a game‐theoretic model between the firm and customers, this paper studies the following: (i) reservation deposits and service prices, and (ii) capacity allocation between reservations and walk‐ins. We have three main results: (i) When reservation no‐shows lead to wasted capacity that cannot be reallocated, the firm should front‐load all charges into the reservation deposit; (ii) The firm should charge a lower service price to reservation‐holders than to walk‐in customers when it decides to serve both; (iii) Less capacity should be allocated for reservations as the potential market size grows; with sufficiently large potential demand, the firm should stop taking reservations. Our results follow from key operational tradeoffs between reservations and queues: reservations permit 100% utilization, but queues operate at less than 100%; however, reservations have constant returns to scale, while queues enjoy increasing returns.

  PublisherDOI

Recent grants

• CAREER: Operations Management with Strategic Consumer Behavior

  NSF · $322k · 2012–2015

• CAREER: Operations Management with Strategic Consumer Behavior

  NSF · $485k · 2008–2012

Frequent coauthors

• Teck‐Hua Ho

  Nanyang Technological University

  12 shared

• Stefanos Zenios

  Stanford University

  8 shared

• Fei Gao

  Indiana University Bloomington

  7 shared

• Jaelynn Oh

  University of Utah

  6 shared

• Fuqiang Zhang

  Washington University in St. Louis

  5 shared

• Glenn M. Chertow

  Stanford Medicine

  4 shared

• Chloe Kim Glaeser

  University of North Carolina at Chapel Hill

  3 shared

• Yaozhong Wu

  National University of Singapore

  2 shared

Awards & honors

• Best Paper Award in Management Science, 2020
• Best Paper Award in Manufacturing & Service Operations Manag…
• Best Paper Award in Management Science, Finalist, 2019
• MSOM Practice-Based Research Award, Finalist, 2017
• MSOM Young Scholar Prize, 2014