Workshop on Platforms and Matching Markets 2025

Speaker: Yehua Wei, Associate Professor of Decision Sciences, Fuqua School of Business, Duke University

Title: Managing Advanced Reservations of Reusable Resources with Continuous Arrivals and Discrete Usage

Abstract: Many service industries, such as hospitality and car rental, involve managing reusable resources through advanced reservations. A distinctive feature in many of these settings is \emph{hybrid temporal resolution}, where reservation requests arrive continuously, while resource usage changes over discrete time units, typically on a daily or weekly basis. Motivated by this, we propose a dynamic model that incorporates hybrid temporal resolution to study the allocation of reusable resources under advanced reservations. Under this model, we develop a dynamic resource allocation algorithm that exhibits significantly stronger guarantees than the existing algorithms for models without hybrid temporal resolution. The algorithm we design is adaptive only with respect to customers reserving a single discrete usage unit, revealing that limited adaptivity is sufficient to achieve strong performance. Our analysis hence demonstrates that, rather than being a modeling complication, the hybrid temporal resolution dynamic can be a source of analytical and operational advantage, enabling the design of more robust and efficient reusable resource management strategies.

Speaker: Mika Sumida, Assistant Professor, Department of Data Sciences & Operations, USC Marshall School of Business

Title: Dynamic Resource Allocation with Recovering Rewards under Non-Stationary Arrivals

Abstract: In many resource allocation settings, the value derived from resources depends on their usage history, with resources that have sufficient recovery or idle periods between allocations often providing greater utility or reward. This paper studies a resource allocation problem in which resource rewards recover over time following each use. Motivated by settings such as content recommendation, service platforms, and renewable energy management, we consider a dynamic matching problem with non-stationary arrivals, where customer types and matching preferences vary over time. Each arriving customer must be immediately and irrevocably matched to a resource or lost. The reward from matching a resource is non-decreasing in the time since its previous use. The goal is to maximize the expected reward collected from all arrivals over a finite time horizon.

Speaker: Farbod Ekbatani, University of Chicago

Title: A Stochastic Growth Model for Online Platforms

Abstract: We investigate the growth trajectory of online platforms operating as market makers in two-sided matching markets. Their expansion, often referred to as the chicken-and-egg dilemma in the literature, poses an inherent challenge; service providers find value in joining the platform due to the presence of customers, while customers are drawn to the platform by the availability of diverse services. This endogenous network effect influences the growth trajectory and propels emerging platforms towards “get big fast” type of strategies to rapidly achieve critical mass on both sides.

Much of the existing literature addresses this expansion challenge through a static market-clearing lens, emphasizing how network externalities shape a platform’s growth trajectory and equilibrium size. However, this approach overlooks the inherent dynamics of gradually recruiting servers and attracting new customers. Embedded in this growth process are compensation schemes and the related intertemporal incentive-compatibility constraints that a platform must satisfy to sustain its expansion.

To explicitly account for the stochastic dynamics of a platform’s gradual growth, we model the system as a Markov process. Service providers join according to a Poisson process whose intensity is modulated by the platform’s compensation structure. Meanwhile, customers are transitory, arriving according to a Poisson process with an intensity that is nondecreasing in the number of active servers. Under mild conditions, we characterize the optimal server compensation schemes the platform should adopt. Furthermore, we show that straightforward compensation schemes, such as uniform payments, do not achieve optimal growth. Instead, an optimal mechanism must account for each server’s "seniority".

Speaker: Sébastian Martin, Assistant Professor of Operations, Kellogg School of Management, Northwestern University

Title: Relative Monte-Carlo for Reinforcement Learning

Abstract: We introduce Relative Monte Carlo (rMC), a new policy gradient algorithm for reinforcement learning that leverages relative returns between a root sample path and counterfactual trajectories induced by alternative actions. This structure yields an unbiased gradient estimator with significantly reduced variance. rMC is compatible with any differentiable policy class—including deep neural networks—and is guaranteed to converge even in infinite-horizon settings. The approach uses common random number coupling to encourage path merging, lowering both variance and simulation cost. rMC is particularly well suited to discrete-event control problems where actions have localized effects, such as in queueing, supply chains, or two-sided platforms. For instance, it can be used to optimize dynamic matching in ride-sharing systems or manage inventory across fulfillment centers for e-commerce platforms. We provide theoretical complexity guarantees for a family of inventory control problems, and empirical results demonstrate that rMC achieves faster convergence, improved policy performance, and greater robustness compared to standard RL baselines, all with minimal hyperparameter tuning.

Speaker: Vahideh Manshadi, Michael H. Jordan Professor of Operations & Research Director for Operations Research, Yale School of Management

Title: Why the Rooney Rule Fumbles: Limitations of Interview-stage Diversity Interventions in Labor Markets

Abstract: Many industries, including the NFL with the Rooney Rule and law firms with the Mansfield Rule, have adopted interview-stage diversity interventions requiring a minimum representation of disadvantaged groups in the interview set. However, the effectiveness of such policies remains inconclusive. In light of this, we develop a framework of a two-stage hiring process, where rational firms, with limited interview and hiring capacities, aim to maximize the match value of their hires. The labor market consists of two equally sized social groups, m and w, with identical ex-post match value distributions. Match values are revealed only post-interview, while interview decisions rely on partially informative pre-interview scores. Pre-interview scores are more informative for group m, while interviews reveal more for group w; as a result, if firms could interview all candidates, both groups would be equally hired. However, due to limited interview capacity and information asymmetry, we show that requiring equal representation in the interview stage does not translate into equal representation in the hiring outcome, even though interviews are more informative for group w. In certain regimes, with or without intervention, a firm may interview more group w candidates, but still hire fewer. At an individual level, we show that strong candidates from both groups benefit from the intervention as the candidate-level competition weakens. For borderline candidates, group w candidates gain at the expense of group m. To understand the impact of non-universal interview-stage interventions on the market, we study a model with two vertically-differentiated firms, where only the top firm adopts the intervention. We characterize the unique equilibrium and demonstrate potentially negative effects: we show that in certain regimes, the lower firm hires less group w candidates due to increased firm-level competition for them, and further find examples where overall fewer group w candidates are hired across the market. At an individual level, while superstar candidates in both groups benefit, surprisingly the impact on borderline candidates may reverse: the lower firm may replace borderline group w candidates with borderline group m candidates in its interview set, effectively reducing the hiring probability of those borderline group w candidates. Overall, our findings highlight challenges in diversifying the labor market at early hiring stages due to information asymmetry, filtering, and competition. Beyond our context, our natural framework of a market with two-stage hiring may be of independent interest.

Speaker: Rad Niazadeh, Associate Professor of Operations Management, Booth School of Business, University of Chicago.

Title: Robust Dynamic Staffing with Predictions: Theory and Applications to Last-Mile Operations

Abstract:  In this talk, motivated by our collaboration with Amazon Last-Mile, I study a natural dynamic staffing problem where, over a finite horizon, we must sequentially hire staff from different workforce pools to meet an unknown target demand. Predictions about the demand arrive over time as uncertainty intervals that become progressively more accurate. Also, the workforce responds less to shorter notices and becomes less available over time. This creates a trade-off: hiring early secures staff but risks overstaffing, while hiring later leverages better predictions but risks understaffing as availability declines.

The main result of our paper is showing how to optimally navigate this trade-off. In particular, I show how to design a simple and computationally efficient online algorithm that minimizes the worst-case staffing imbalance cost against any sequence of prediction intervals, i.e., a minimax-optimal policy. At a high level, the approach relies on identifying a restricted adversary that allows us to characterize the minimax cost through an offline LP, and then emulating that LP solution in the online setting. We also design and analyze a second minimax optimal policy, with improved practical performance, that relies on LP-resolving. If time permits, I will also discuss extensions to multiple target demands under both egalitarian and utilitarian objectives, to reversible hiring with discharge costs, and to the case of inconsistent prediction intervals.

Speaker: Negin Golrezaei, W. Maurice Young (1961) Career Development Associate Professor of Management & Associate Professor of Operations Management, MIT Sloan School of Management

Title: Learning Safe Strategies for Value Maximizing Buyers in Uniform Price Auctions

Abstract: We study the bidding problem in repeated uniform price multi-unit auctions from the perspective of a single value-maximizing buyer who aims to maximize their cumulative value over T rounds while adhering to return-on-investment (RoI) constraints in each round. Buyers adopt m-uniform bidding format, where they submit m bid-quantity pairs (b, q) to demand q units at bid b. We introduce safe bidding strategies as those that satisfy RoI constraints in every auction, regardless of competing bids. We show that these strategies depend only on the bidder's valuation curve, and the bidder can focus on a finite subset of this class without loss of generality. While the number of strategies in this subset is exponential in m, we develop a polynomial-time algorithm to learn the optimal safe strategy that achieves sublinear regret in the online setting, where regret is measured against a clairvoyant benchmark that knows the competing bids a priori and selects a fixed hindsight optimal safe strategy. We then evaluate the performance of safe strategies against a clairvoyant that selects the optimal strategy from a richer class of strategies in the online setting. In this scenario, we compute the richness ratio, α ∈ (0, 1] for the class of strategies chosen by the clairvoyant and show that our algorithm, designed to learn safe strategies, achieves α-approximate sublinear regret against these stronger benchmarks. Experiments on semi-synthetic data from real-world auctions show that safe strategies substantially outperform the derived theoretical bounds, making them quite appealing in practice.

Speaker: Huseyin Topaloglu, Howard & Eleanor Morgan Professor of Operations Research & Information Engineering, Cornell University / Cornell Tech

Title: Revenue Management with Calendar-Aware and Dependent Demands: Asymptotically Tight Fluid Approximations

Abstract: When modeling the demand in revenue management systems, a natural approach is to focus on a canonical interval of time, such as a week, so that we forecast the demand over each week in the selling horizon. Ideally, we would like to use random variables with general distributions to model the demand over each week. The current demand can give a signal for the future demand, so we also would like to capture the dependence between the demands over different weeks. Prevalent demand models in the literature, which are based on a discrete-time approximation to a Poisson process, are not compatible with these needs. In this talk, we focus on revenue management models that are compatible with a natural approach for forecasting the demand. Building such models through dynamic programming is not difficult. We divide the selling horizon into multiple stages, each stage being a canonical interval of time on the calendar. We have random number of customer arrivals in each stage, whose distribution is arbitrary and depends on the number of arrivals in the previous stage. The question we seek to answer is the form of the corresponding fluid approximation. We give the correct fluid approximation in the sense that it yields asymptotically optimal policies. The form of our fluid approximation is surprising as its constraints use expected capacity consumption of a resource up to a certain time period, conditional on the demand in the stage just before the time period in question.

Speaker: Philipp Afèche, Professor of Operations Management & Statistics, Rotman School of Management, University of Toronto

Title: Ride-Hailing Networks with Strategic Drivers: The Effects of Driver Wage Policies and Network Characteristics on Performance

Abstract: Ride-hailing platforms face two important challenges: (i) there are significant spatial demand imbalances that require some repositioning (empty routing) of drivers; (ii) the control of driver supply is partially decentralized in that drivers strategically decide whether to join the network, and if so, whether and where to reposition when not serving riders. We study the following question for such ride-hailing networks: Under decentralized repositioning, how effective are driver wage policies in achieving the optimal centralized performance benchmark? We consider a stationary fluid model of a ride-hailing network in a game-theoretic framework with riders, drivers, and the platform. We show how the effectiveness of driver wage policies under decentralized repositioning depends on the interplay of the network’s spatial (travel time) configuration, driver wage flexibility, and the congestion-sensitivity of travel times: (1) We identify conditions on the travel times for the existence of a driver repositioning equilibrium. (2) For networks with constant travel times, we show that the centrally optimal repositioning flows can be implemented under decentralized repositioning, provided the platform has sufficient wage flexibility, whereas more limited wage flexibility leads to inefficiencies in terms of driver idling. (3) For networks with congestion-sensitive travel times, the centrally optimal repositioning flows can generally not be implemented under decentralized repositioning, even with full wage flexibility, so that decentralized repositioning leads to higher congestion, capacity levels, and driver wage rates compared to centralized repositioning.

Speaker: Francisco Castro, Assistant Professor of Decisions, Operations & Technology Management, UCLA Anderson School of Management

Title: Electric Vehicle Fleet and Charging Infrastructure Planning

Abstract: We study electric vehicle (EV) fleet and charging infrastructure planning in a spatial setting. For a centrally managed fleet that serves customer requests arriving continuously at rate λ throughout the day, we determine the minimum number of vehicles and chargers for a target service level, along with matching and charging policies. While non-EV systems require extra Θ(λ2/3) vehicles due to pickup times, EV systems differ. Charging increases nominal capacity, enabling pickup time reductions and allowing for an extra fleet requirement of only Θ(λ^ν ) for ν ∈ (1/2, 2/3], depending on charging infrastructure and battery pack sizes. We propose the Power-of-d dispatching policy, which achieves this performance by selecting the closest vehicle with the highest battery level from d options. We extend our results to accommodate time-varying demand patterns and discuss conditions for transitioning between EV and non-EV capacity planning. Simulations verify our scaling results, insights, and policy effectiveness. While long-range, fast-charging fleets resemble non-EV systems, short-range, low-cost fleets can still perform competitively---underscoring the need for EV-aware management policies.

Speaker: Yash Kanoria, Merrill Lynch Professor of Workforce Transformation, Columbia Business School

Title: What Is Your AI Agent Buying? Evaluation, Implications and Emerging Questions for Agentic E-Commerce

Abstract:  Online marketplaces will be transformed by autonomous AI agents acting on behalf of consumers. Rather than humans browsing and clicking, vision-language-model (VLM) agents can parse webpages, evaluate products, and transact. This raises a fundamental question: what do AI agents buy---and why? We develop ACES, a sandbox environment that pairs a platform‑agnostic VLM agent with a fully programmable mock marketplace to study this question. We first conduct basic rationality checks in the context of simple tasks, and then, by randomizing product positions, prices, ratings, reviews, sponsored tags and platform endorsements, we obtain causal estimates of how frontier VLMs actually shop. Models show strong but heterogeneous position effects: all favor the top row, yet different models prefer different columns, undermining the assumption of  a universal "top" rank. They penalize sponsored tags and reward endorsements.  Sensitivities to price, ratings, and reviews are directionally human‑like but vary sharply in magnitude across models.  Motivated by scenarios where sellers themselves use AI agents to optimize product listings, we show that a seller‑side agent that makes minor tweaks to product descriptions---targeting AI buyer preferences---can deliver substantial market‑share gains if AI‑mediated shopping dominates. We also find that modal product choices can differ across models and, in some cases, demand may concentrate on a few select products, raising competition questions. Together, our results illuminate how AI agents may behave in e‑commerce settings, and surface concrete seller strategy, platform design, and regulatory questions in an AI‑mediated ecosystem.

Speaker: Ming Hu, Professor of Operations Management & Statistics, Rotman School of Management, University of Toronto

Title: Spatial Staffing

Abstract:  We study the staffing problem for an on-demand vehicle-sharing platform operating over a d-dimensional service region. The platform makes a one-time capacity decision of how many vehicles to employ at the start of an infinite horizon, and dynamically controls vehicle-customer matching and routing over time. The objective is to minimize the long-run average cost, which includes vehicle operations costs and customer waiting costs. We show that the optimal staffing level consists of a nominal load (i.e., the minimum number of vehicles to ensure system stability) plus a safety buffer. This safety level depends on key system parameters, including the dimensionality d, the distributions of customer origins and destinations, and the ride-pooling capacity q (i.e., the maximum number of passengers per vehicle). Specifically, (i) when q=1, the safety level scales with the arrival rate raised to the power of d/(d+1), a result that holds broadly across distributional assumptions and mirrors earlier findings under stylized spatial models. (ii) When q≥2, the scaling becomes 2d/(2d+1), which, to our knowledge, was not discovered in the past. Importantly, we derive these results by analyzing exact vehicle routing without relying on stylized assumptions and by developing provably near-optimal policies for general spatial matching settings beyond those in prior work.