In the healthcare industry, quickly recalling defecting products can be instrumental in avoiding potentially harm to patients. Building on upper echelons theory and personality theory, we argue that neurotic CEOs are more vigilant when they face uncertainty, as they do when making recall decisions, leading to faster recalls. We build on a novel, machine-learning based approach to measure the personalities of 110 CEOs of U.S. healthcare firms and find broad support for our theory in analyses.

We measure the digital impact of the Indian COVID-19 lockdown using a survey coupled with browser history data (n=1,094). Gender differences are present for online leisure, production, YouTube, social media and job search – and not for online learning. Working women sacrificed online leisure while maintaining online production. The gender gap is larger among parents. Fathers self-reported significantly larger childcare time increases, yet browser data and partner reports do not support this.

We analyze prices paid in low-income and high-income areas after three hurricanes. Using a triple difference regression, we isolate the percentage change in prices paid by low-income versus high-income areas in affected versus unaffected areas. Compared to high-income areas, low-income areas experience a larger drop in promotions, higher unit percent price increases, a larger decrease in offer sets of low-priced items, and a larger increase in substitution from low-priced to high-priced items.

Emergency Department (ED) clients lack knowledge about the various elements their personal ED journey comprises. In a field study (N=18,903) we provide "Personalized Information about Operations and Time (PIOT)" – (frequently updated) information about the procedures and anticipated wait times for each patient. We show PIOT improves people’s attitudes, and PIO (operational-only information) reduces patients' abandonment. Providing PIOT offers a novel approach to improving healthcare service.

In this paper, we study how the boarding patients can affect the treatment time in emergency departments (EDs). First, we conduct an empirical analysis using a dataset from eight EDs and show that the relationship has an inverted U-shape. Next, we recognize two mechanisms behind it: (i)the additional workload on ED physicians, (ii) the hospitalist visits triggered by boarding congestion. Finally, we propose two interventions that can jointly mitigate this impact up to 68% in a tertiary hospital.

We introduce a MILP formulation for patient prioritization in the ED that incorporates a breakdown of predicted patient LOS. We propose an SAA based reformulation that can be solved efficiently and provide guarantees on its convergence, stability and sample complexity. A study of 40000 patient visits from a US hospital shows that our solution increases ED throughput by >50% and decreases average waiting time by >75%. Our method displays desirable properties of fairness in patient prioritization.

Motivated by trends in the healthcare industry, we study the integration of Acute Care (AC) and Post-Acute Care (PAC). Patients' outcomes depend on both the AC length of stay and the PAC efforts. A central controller that manages both AC and PAC may choose to discharge a patient earlier and ``compensate'' for this by greater PAC efforts. We characterize the optimal effort integration and its dependencies on patient's operational severity, and the extent to which AC and PAC are complements/substitutes.

We propose a method to enhance post-approval safety surveillance of pharmaceuticals. Clinical trials typically do not provide sufficient evidence for flagging rare safety issues. Our approach combines pre-approval clinical trial results with post-approval surveillance data for common and rare adverse effects to decide whether to flag the rare reaction, weighing type I and type II error costs. We analytically show when our approach is most valuable and numerically demonstrate its effectiveness.

We study the interaction between a seller and a buyer, both of whom face uncertainty related to downstream demand, over a two-period horizon. Making multi-unit purchase decisions before demand from their respective lower tier realizes, both agents seek to minimize their demand mismatch risk as perceived according to their degree of foresight. Focusing on the effect of buyer foresight, we find that with multi-unit purchases sellers benefit from some degree of buyer foresight but not too much.

Behavioral aspects of consumer return policy design and their interaction with other decisions in retailing have not been investigated to date. Leveraging a generalized newsvendor model, we conduct a randomized experiment to assess how subjects jointly make key decision (order quantity, price, and refund amount) and the effect of salvage value on them. We identify time-dependent behavioral regularities that we explain through a process theory, thus providing a new direction for future research.

Actual inventory can be lower than recorded system inventory due to shrinkage or loss. To handle inventory inaccuracy, managers can decide to clean inventory data before placing order quantities. Alternatively, they can deliberately decide to not clean, but to compensate for the inaccuracy in the order decision. The optimal decision depends on the cost of cleaning and the efficiency loss due to the compensation. We present a set of hypotheses on this trade-off and test these in lab experiments.

Modern traceability technologies promise simplified recalls, increased visibility, and verification of sustainable practices. However, the benefits obtained from traceability are conditional on technology adoption throughout a supply chain. Hence, traceability initiatives need to subsidize some early adopters within a network of supply chains to achieve broad diffusion. We address the problem of identifying this "seed set" and describe how the supply chain network structure affects the choice.

We design a behavioral simulation using a blockchain-enabled market for trading suppliers’ capacities among competing retailers. Retailers have different valuations for goods and order before knowing their demand. After demand realization, retailers can trade among themselves. While average initial orders do not differ significantly compared to the control group, significant improvements in inventory allocation and profits are achieved, and trading strategies arise that mitigate demand risk.

This study examines the impact of (a) leasing or selling decisions and (b) alternate supply chain forms (CSC or DSC) on digital piracy and supply chain profits. We develop an analytical model where supply chain members lease or sell digital goods in the presence of pirated goods in a two-period setting. We find that leasing digital goods to buyers has higher supply chain profit than selling in a CSC, and selling provides higher supply chain profit than leasing in a DSC.

Motivated by the setting of Airbnb, we consider a game theoretical setup in which each seller on the platform provides a single-unit product and competes on price. We investigate sellers' optimal pricing decisions and the platform's optimal assortment display policy. Additionally, we incorporate constraints to guarantee a certain degree of seller and customer fairness. Using data from Airbnb, we present a case study to illustrate how our model framework can be applied in practice.

We study the relative merits of different dispute resolution mechanisms in two-sided platforms, in the context of disputes involving malicious reviews and blackmail. We develop a game-theoretic model of the strategic interactions between a seller and a (potentially malicious) consumer. Our results suggest that decentralization, when implemented correctly, may represent a more efficient approach to dispute resolution.

We consider an on-demand service platform (e.g., Uber, Lyft, DiDi) that serves two customer classes with distinct willingness to wait and to pay. We formulate this problem as a Markov decision process in which the platform controls customer admission and service provider allocation. Using our model's structural properties, we show that the optimal policy is characterized by two admission thresholds. In a numerical study, we then compare the optimal policy with several simpler policies.

This study analyses street-turn strategies for empty container repositioning in the hinterland using a double-ended queue model for matching operations. Containers arrive over time at the consignee and the demand for containers arises from the shipper. We prove that the matching time impacts matching proportion, while it marginally influences the consignee's inventory policy and cost per container. Also, the consignee's withholding level is mainly determined by the shipper's production rate.

In relay-based logistics, orders are routed through pit-stops with a different driver assigned to each segment. This paper formulates an integer optimization model to coordinate driver, truck and driver movements. We develop an arc-based column generation algorithm which expands time-space networks iteratively until convergence. Results show that relay operations, combined with our algorithm, can lead to faster deliveries, better driver lifestyles, and a lower environmental footprint.

In online retail, fulfillment optimization is the problem of dynamically determining, for every new order, the fulfillment node that will fulfill every item within the order. To solve the problem, we propose a novel policy function approximation approach based on genetic programming to generate interpretable fulfillment policies. Results show that policies more simple than mathematical programming based ones are able to significantly improve over a myopic assignment.

We study a dynamic pricing problem over multiple periods when consumers develop a habit or satiation from their past consumption. We derive an inter-temporal demand function to capture these two effects. We establish that the profit maximization problem under our demand function is jointly concave and then characterize the trends in the optimal prices over the multi-period horizon. Finally, we provide several extensions including bounds on prices and optimal profit and non-stationary state dependence.

Ride-hailing firms set prices dynamically to match supply and demand. But rapid price changes incentivize riders to wait for low prices. When prices drop, patient customers request en masse, causing a drop in supply and a price increase. We show how dynamic pricing inherently creates such oscillations in supply and prices, that these oscillations in supply levels are inherently inefficient, and that a service model that allows riders to wait in a formal queue overcomes this inefficiency.

We study the pricing problem of an assortment of multiple, substitutable products. We propose two new methods that can support retailers with maximizing their revenues. The first method is a customer choice model based on the Markov Chain Choice model in combination with reservation prices. The second method relies on a linear approximation for the finite inventory, finite time horizon multi-product dynamic pricing problem.

We present our collaboration with OCP, the world’s largest producers of phosphate and phosphate-based products, in support of a green initiative designed to reduce the company’s greenhouse gas emissions. The proposed robust optimization-based methodology guides the company’s investment in solar panels and batteries, which accounts to over one billion US dollars, as well as their day-to-day operations, and is expected to significantly reduce both the company’s emissions and energy bill.

Electrification is as a critical enabler of the decarbonization. It is imperative to study the growth in electric vehicles adoption to plan for this impending transformation. Existing EV adoption studies typically ignore the influence of other green technologies. In this paper, we bridge these critical gaps. We employ a dynamic discrete choice model to study these technologies' diffusion. Our work projects adoption and evaluates counterfactual scenarios.

Firms that already invest in more sustainable technologies as a proactive measure against changing market dynamics, are likely to gain a significant competitive advantage. We analyze a large focal firm's optimal green investment strategy, accounting for the uncertainty in the competitors' actions and the future green market size. Optimal policy is composed of `Retreat’, `Defend’, and `Attack’ strategies, one of which is optimal based on the problem parameters. We provide managerial insights.

Companies operating under gig-contractor models are offering part-time job requiring longer work availability. This ensures quicker delivery & better service quality. Using data from an online retailer we find that workers with similar level of experience perform differently. This is explained through the role of worker availability. At similar levels of experience, workers in high-work group perform better than those in the low-work group after controlling for task and worker characteristics.

New sustainability-marketed firms have emerged in the grocery and consumer packaged goods categories responding to consumers’ rising preferences for sustainable products. Motivated by this trend in the retail industry, we study how the entry of a new firm that sells an assortment of sustainable consumer goods affects the consumers’ channel choices and the existing retailers’ market shares in two different types of product offerings -- packaged and fresh goods.

Slow-moving items constitute a large share of the retail assortment and often result in intermittent orders by the retailer. We estimate retail channel inventories based on prior orders and Point-of-Sales data to predict the timing of future orders. We demonstrate both theoretically and empirically that this an inventory modeling approach outperforms the Croston's method and thus provides a viable alternative to conventional time-series models.

When tree-based models, such as Random Forest or XGBoost, are used in optimization, their formulations are often intractable and unscalable. We propose a scalable approximation of the optimization formulation that can optimize over ensemble tree models in linear time while still capturing over 90% of optimality on a variety of datasets. MOTEM (Method for Optimizing over Tree Ensemble Models) is an algorithm for optimizing an objective function that is determined by an ensemble tree model.

AI systems are increasingly demonstrating their capacity to make better predictions

than humans. Yet, recent empirical studies suggest that experts may doubt the quality of these systems. We explore the extent to which a decision maker (DM) can properly learn whether a machine produces better recommendations, and analyze a dynamic Bayesian model, where a machine performs repeated decision tasks under a DM’s supervision. We fully characterize the conditions at which learning fails and succeeds.

We study data-driven operations management problems with high-dimensional data. The standard approach involves two separate modeling phases: learning a low dimensional statistical model (dimensionality reduction) from data and then optimizing a decision problem with parameters input from said statistical model. We propose a prescriptive dimensionality reduction approach that better aligns the two phases and delivers superior results over standard methods (e.g., principal component analysis).

Increasingly, Consumer Packaged Goods (CPG) companies make use of reusable packaging and manage them via a deposit system. We study a CPG company that offers a product in reusable and disposable packaging, and manages reusable packaging via a deposit system. We formulate a decision model in which the CPG company sets product price and deposit fee under price and deposit sensitive demand while considering packaging durability. We provide analytical conditions for optimality and procedure to solve

Power systems require new approaches to system operations to respond to the increased volatility of solar or wind energy. In this paper, we suggest a novel online demand response program with variable prices. Our program is able to incorporate new information on changing wholesale market conditions while notifying load operators early enough to enable response. We furthermore propose Deep Reinforcement Learning as a tool to identify effective prices.

In this research project, we aim to understand when private households with existing or planned

rooftop solar installations should invest in electricity storage and how these investment decisions

affect their electricity usage patterns as well as the market structure overall. We use a novel household panel datasets to structurally estimate households consumption utility functions and valuations for self-produced solar energy.

I study farmer poverty and productivity with unique data from 2,000 Philippine coconut farms and field work. I find strong evidence that micro-level farming practices account for large productivity differences and that best fertilization practices vary with environment. Supporting organizations should develop customized farming advice and assist farmers with fine details of implementation, an approach not currently preferred, but increasingly available through emerging information technologies.

Firms increasingly recognize the importance of their upstream suppliers’ social responsibility. However, they may fail to heed the unintended negative consequences of their own common practices on the suppliers’ social responsibility decision. Our study shows that cost auditing may undermine the supplier's social responsibility choice, which sheds light on the reluctance of many suppliers to commit to social responsibility programs.

We consider contracting between a supplier of specialty material forgings and a buyer that manufactures airplane components by extensively machining them down. Due to high material removal costs, the buyer prefers forgings to be as similar in geometry and size to the component as possible. We assess the implications of two innovative approaches for improving supply chain performance: (i) Input material reduction incentives via contracting, and (ii) scrap material recycling.

We study the impact of monitor leniency on supplier CSR risk. Using audit data from a global apparel brand, we find that leniency helps to reduce CSR risk. Testing interaction effects with our moderators, we find that greater leniency helps to reduce CSR risk when a facility’s compliance ability is low; when a facility has been audited a small number of times; and when a facility is located in a developing country. Our work provides insight into the relational factors that can influence supplier audits.

Manufacturers are increasingly being held responsible for sustainability violations across their whole supply chain. We examine a three-tier supply chain where a manufacturer and its direct supplier incentivize sub-supplier sustainability through auditing and supplier development. We show that the mechanisms are substitutive for the supplier but can be complementary for the manufacturer. Moreover, the manufacturer delegates the effort in case of very low or very high external pressure.

We propose a conceptual framework for supplier sustainability improvement that we refer to as CARE, based on self-assessments and consisting of Collect, Assess, React, and Enhance phases. CARE is highly scalable, making use of machine learning techniques to understand the link between the general supplier characteristics and their verified sustainability profile, predict the future sustainability levels of even unassessed suppliers, and determine the best plan for improvement.

We present a queueing games framework to investigate how sharing of real-time queue-length information at emergency department (ED), where urgent patients receive priority over nonurgent patients, influences nonurgent patients' decision to enter or balk the ED queue, and how it affects the overall social welfare of patients who visit the ED. We show that under certain conditions, it may be better to partially reveal ED queue-length information rather than making ED queues completely transparent.

In a field experiment in an Emergency Department, we found that providing delay information improves patients' waiting satisfaction by 81%, and decreases their likelihood of reneging by 14%. The announced delay acts as a reference point against which the patients compare the actual delay. Following Prospect Theory, we found that patients are loss-averse that the likelihood of LWBS is much lower when they wait a shorter time than announced than when they wait a longer time than announced.

Standing orders allow triage nurses in EDs to order tests for certain medical conditions before the patient sees a physician, which could reduce the patient’s LOS. Medical literature documents the use of standing orders decreasing average ED LOS for the patient subject to standing orders. We model operational impact of standing orders and introduce a threshold based congestion-sensitive policy which performs well wrt overall average ED LOS across a wide range of scenarios.

In the blood donor management problem, the blood bank incentivizes donors to donate, given blood inventory levels. We propose an optimization model to design donor incentivization schemes that account for the blood inventory dynamics and the donor's donation process. By adopting the Pipeline Queue paradigm, we have a tractable convex reformulation. Numerical results show the advantages of the optimal policy compared with benchmark policies in reducing both shortages and wastage.

Clinical trials are a critical step for the development of new drugs, both in

cost and in terms of elapsed time to bring the potential drug to the market. Since clinical trials are increasingly going global, optimizing the supply chain can bring huge benefits. We use a MDP to model the inventory problem between the central depot and the regional depots that will then supply the investigation sites. The main decision is when to resupply and how much inventory to send to each regional depot.

The Organ Procurement & Transplantation Network (OPTN) is migrating all US organ allocation policies to a novel continuous-distribution model. We introduce a novel analytical framework to illuminate policy tradeoffs and enable exploration of the efficient frontier of policies. Jointly with the OPTN, we applied our framework to the design of a new allocation policy for lungs. Starting in 2023, all deceased-donor lungs in the US will be allocated according to this policy.

How the cost of completing a queue varies with (i) the experience of wait and (ii) the characteristics of the queue are not well understood. In this study, we use the incentive-compatible BDM mechanism to experimentally address these questions. We find that when service speed is slow, experienced wait increases (decreases) the completion cost of impatient (patient) subjects. Also, the length and speed of a queue affect completion costs, but not proportional to their effects on total waiting time

Through a series of experiments, we show that when others are waiting in line, customers accelerate their own service time, sacrificing their own consumption utility. This behavior is driven by concern for others. We show that the negative effect of others queueing on one’s own service time is moderated by the participants' self-wait and visibility between customers in service and those waiting in line.

Contrary to traditional queueing theory, recent field studies in health care and call centers indicate that pooling queues may not lead to operational efficiencies relative to dedicated queues. We use a series of experiments to examine the conditions under which this may be the case and to test servers' customer orientation as a behavioral mechanism that may explain why. We also examine whether higher levels of customer orientation and performance persist across changes in queue configuration.

For many information goods, longer publication cycles are more economical, but often result in less timely information. While the digitalization of publication processes makes shorter cycles more economically viable, in practice, not all competing firms choose to publish more frequently. In this paper, we use a game-theoretic model to determine how information providers choose publication cycles and prices under competition and inform publishers of adaptive strategies for digitalization.

In knowledge-intensive businesses, projects are often initiated by individuals, who may then look for collaborators to push their idea forward. What operating dynamics arise in equilibrium when the decision-making process on project ownership and project execution is decentralized? Using a stylized principal-agent model, we find that principal investigators have a tendency to keep ideas for themselves too much, but when they share the project ownership, they tend to over delegate.

Demographic and professional activity data on 227 Operations Management faculty in 22 U.S. public schools is compared to their base salaries (all 227 faculty) and total compensation (a subset of 15 schools and 150 faculty). A primary factor correlated with pay levels are “A” journal publications. Local school effects (cost of living), willingness to move institutions, taking on administrative duties, and achieving Fellow status in professional societies are all correlated with higher salaries.

Multi-homing gives platforms access to a larger pool of supply; however, it also changes the nature of competition between platforms in a market. It is not clear whether this works to the advantage of platforms or not. In light of this, we ask a fundamental question: When is it better for two-sided platforms to pool their workers? We answer this question through a game-theoretic study. We identify the key trade-offs associated with pooling decision and highlight the key role of scale.

Gig workers prevalently "multihome'' by dynamically allocating their services in real-time between multiple platforms. As a growing number of platforms access the same pool of workers to complete their gigs, the question of how workers choose between competing platforms has grown in salience. In this work, we study gig workers' multihoming decisions by using machine learning methods to estimate a structural model from ride-hailing proprietary data combined with publicly reported trips data.

Using a large panel dataset with repeated choices of riders for both Uber and Lyft, we estimate a structural “consider-then-choose” model to better understand the trade-offs faced by riders. We find that riders' choices are not fully explained by operational factors such as price and waiting time, indicating that riders view the platforms as differentiated services and not as commodities. We also find that the multi-homing behavior of riders is only observed for a small fraction of the rides.

We study capacity flexibility via an innovative business practice: On-Demand Warehousing. In this emerging application, a platform connects independent warehouse providers, who are willing to sell excess capacity, with a firm that requires on-demand capacity. On-demand warehousing does not require long-term commitments, but rather provides flexible warehouse capacity, on-demand. Our results highlight how on-demand warehousing allows a firm to absorb demand fluctuations better.

Enabled through recent advances in technology, coupled with the advent of new system providers and decreased price points, automated and robotic order picking solutions evolved as a surging market. As implementation projects and the variety of solutions are rising, managers face the decision which ones to select for their specific business case. We contribute by proposing a mathematical optimization approach that assigns each stock keeping unit the most suitable solution under space constraints.

Online pickers encounter customer interference while picking orders, affecting productivity due to store traffic and queues, and service quality due to picking errors. Within the day, there are less-busy periods when stores resemble a warehouse. We match similar orders picked during peak vs non-peak periods to establish the value of picking during non-busy hours. Our research has implications for online grocers willing to be productive without incurring additional cost of maintaining dark stores

Dynamic resource allocation problems arise under a variety of settings and have been studied across disciplines such as Operations Research and Computer Science. This work introduces a unifying model for a very large class of dynamic optimization problems. We show that this class encompasses a variety of disparate and classical dynamic optimization problems and we characterize the performance of the fluid certainty equivalent control heuristic for this class of problems.

We consider revenue management problems with heterogeneous resources, each with unit capacity. An arriving customer makes a booking request for a particular interval of days. The goal is to find a policy that determines an assortment to offer each customer to maximize total expected revenue. We show that we can efficiently perform rollout on any static policy. We develop two static policies derived from value function approximations, and give performance guarantees for both policies.

We study the energy commitment, generation, and storage problem for a wind farm co-located with a battery. Modeling this problem as a Markov decision process under wind and price uncertainties, we prove the optimality of a state-dependent threshold policy. Our numerical study with data-calibrated instances has revealed the high efficiency and scalability of our solution method constructed with structural knowledge.

We study the energy generation and storage problem for a pumped hydro energy storage facility integrated with a wind farm. The operator decides in real-time how much water to pump/release, how much wind energy to generate, and how much energy to buy/sell. Modeling the problem as a Markov decision process, we prove the optimality of a state-dependent threshold policy for the upper reservoir water level. Leveraging this result, we develop an efficient solution method for data-calibrated instances.

We explore the effect of physical store presence on purchase decisions in omnichannel retailing. We use geolocated customer-level data from a major shoe retailer and study the differential role of physical proximity (number of stores, distance to closest store), as well as service quality (assortment breadth and service level). We find that, while proximity generally increases sales, the service quality provided by the physical store network increases offline sales but decreases online sales.

We examine transaction-level data of an Italian furniture retailer to study channel-specific effects of fulfillment lead time on demand. We find that the showroom channel makes consumers less sensitive to fulfillment lead time than both online and catalog channels. Niche products and experience goods further accentuate the difference of lead time sensitivity between showroom and non-physical channels. Our study highlights the previously-ignored lead time aspect of the physical store’s value.

An omnichannel grocery retailer can fulfill incoming orders either from the dark store or from a brick-and-mortar store. Customers are offered only those products available in the DS and the B&M store. We develop a new business model that offers customers all products available in the DS or the B&M store. We develop a new decision-making mechanism to determine the fulfilment location for each order. We computationally compare our dynamic policies with the omnisciently optimal solution.

We study a problem of sequential learning and choosing from a group of similar alternatives. The unknown payoff of accepting an alternative depends on a set of common features that enable transfer learning across the group. There is also an idiosyncratic value that needs to be learned by sampling over time. The problem is whether to accept the current alternative, continue sampling, or switch to the next one. We model it as a Bayesian dynamic program and analyze structural properties.

We consider the prototypical problem of sequentially assigning jobs to workers at a large centralized matching platform under an infinite worker supply governed by a fixed distribution; this encapsulates the defining characteristic of large market settings such as online labour marketplaces. The goal is to maximize cumulative payoffs from matches. We resolve several foundational questions pertaining to the complexity of this problem setting and provide novel rate-optimal algorithms and analyses.

In many contexts such as technology and management consulting, clients seek the expertise of providers to find solutions for their business problems. We develop a dynamic game-theoretic model where the provider chooses how to present alternative solutions, and the client chooses which solution to try.

We study if discounts for products at their expiry dates can reduce waste and increase profit. In a Markovian inventory model we obtain combinatorial expressions for the transition rates, but with no informative stationary distribution. In a regime where customer arrivals and order-up-to-level grow large, we obtain via Donsker's theorem expressions for waste and profit. In an MNL setting we prove that optimizing regular prices and discounts always reduces waste compared to not giving discounts.

Consumers often exhibit showrooming behaviour in which they visit a brick-and-mortar (BM) store to gather product information but complete the product purchase in the online channel. Many BM retailers carry exclusive store brand products. We examine how consumer showrooming interacts with a BM retailer's exclusive store brand strategy. Contrary to common notion, our findings reveal that the BM retailer can benefit from consumer showrooming when it carries an exclusive store brand.

We investigate whether horizontal competition drives the increase of the number of product portfolio varieties of self-interested firms that compete for demand through their product portfolio sets. We characterize the equilibrium, in both, the complete information game and the incomplete information game and prove that neither firms have the incentive to go beyond its monopolistic choice. Moreover, we show that proliferation may fail as an entry barrier when the game is played a la Stackelberg.

Many manufacturers frequently introduce new products and retire low-performing SKUs. These portfolio adjustments cause a demand shock for existing products. We study the impact of these demand shocks on the bullwhip effect for existing products. We prove that retiring products always increase the bullwhip effect for existing SKUs while introducing new products does not necessarily lead to this increase. We also study the behavior of the bullwhip effect as function of time remaining to the shock.

We propose a three-step demand forecasting framework that combines the expert's knowledge of the market with the machine learning algorithm's ability to leverage historical information to forecast seasonal demand for rapid innovation products. Using data from Canyon Bicycles, we find a 29% reduction in forecast error (measured by WMAPE) over a purely judgmental forecast.

We present our work with a publicly listed food manufacturer in the UK and a private equity firm that invests in the heavy equipment industry to improve their procurement practice. We used natural language processing and machine learning to organize their vast unstructured procurement data and to classify the suppliers and products into hierarchical categories. With our accompanying decision support tool, we identify the procurement inefficiencies and provides request-for-quote (RFQ) targets.

Strategies for scheduling patients to minimize resource and waiting costs is a key problem in healthcare operation. The provider should decide patients to schedule as well as future appointment times to assign. We consider the multi-class advance scheduling problem with random arrivals. We propose and analyze the optimal policy that determines patients and appointments for the current state, effectively prioritizes the patients, and efficiently balances the workload across the booking window.

We propose a data-driven method for surgery scheduling that combines AI algorithms with stochastic optimization techniques. While existing approaches only consider the surgery type to differentiate between surgeries, our method allows us to consider different surgery-, patient- and physician-related aspects. We find that our method outperforms an existing method which only takes procedure types into account, resulting in higher operating room utilization and lower waiting times.

The simultaneous consideration of dynamic inter-day and intra-day scheduling decisions is an established theoretical and practical problem that has remained open due to its highly stochastic nature, complex structure, and the curse of dimensionality. We develop the first analytical optimization model and theoretical results addressing this joint problem within a computationally tractable optimization framework with theoretical performance guarantees.

We develop a model for a regional decision-maker to analyze the requirement of medical equipment capacity in the early stages of a spread of infections. We use a stochastic differential equation to capture the growth of infections in a community spread and shutdown model. We develop results to determine shutdown time, to show how to compensate for limited medical equipment capacity, and to show how capacity-sharing across regions can deliver a peak-timing benefit beyond traditional risk pooling.

During the early weeks of the COVID-19 pandemic, hospitals managed surge capacities by transferring patients among different hospitals within the same health network, repurposing operating rooms as ICU beds, and cancelling elective surgeries. Using publicly available data we develop an analytical framework to study how these policies can be implemented, individually or in combination, in order to optimize the pandemic response in a region, while delivering care to the uninfected patients.

The classical SIR model and its variants have succeeded in predicting infectious diseases' spread. To better capture the COVID-19 outbreak, we extend the SIR model to impose a testing capacity. We study how to choose the best type of testing method, how to allocate limited testing capacity over time and across symptomatic and asymptomatic people. We use the COVID-19 data and a sliding window method to calibrate our model and point out its public policy implications.

The algorithm-augmented business model promises service companies the benefits of both algorithms and humans. But companies will only realize this promise if their workers rely on algorithms, and there is conflicting evidence about workers’ willingness to do this. We design a laboratory experiment to resolve this conflict, and find that workers are generally unwilling to rely on algorithms but that they become more willing to do so in response to high customer loads and learning interventions.

We conduct a series of online experiments to isolate the pure effects of three types of Relative Performance Feedback, RPF, on middle-ranked workers' performance. We find that providing any type of feedback reduces performance compared to no feedback. Aligned with theory, delivering feedback increases the focal employee's shame and social comparison involvement (SCI), which measures the focal individual’s level of engagement in social comparison while performing the task.

Many human-machine interactions focus on the optimization of the system output yet tend to overlook human behavior. Using an intervention-based field experiment in a semi-automated warehouse, we study the impact of a behavioral intervention that provides humans with more information about the picking process and enables them to choose out of a set of pre-defined goals. We find that human performance is enhanced by 6%. Our insights enrich the discussion on human-machine interactions.

Being operationally transparent is often misaligned with firms’ incentives. In such situations, regulators or industry watchdogs may impose operational transparency requirements on firms. In this study, we examine the impact of operational transparency on R&D novelty. Exploiting an exogenous shock in the US pharma industry, we show that increased transparency leads to less novel R&D bets. Our research not only contributes to operations management theory but also informs policy.

A financial constraint for R&D is an essential issue for a firm's operations yet often ignored in research. We analyze a supply chain consisting of a supplier and a manufacturer, where the latter has an R&D opportunity with limited resources. We show that the latter's profit can decrease even if it can afford more investment, referred to as the R&D conundrum. We investigate operational and information strategies, including upfront R&D investment and keeping its financial budget private.

Concurrent development strategies can enable a firm to gain the first-mover advantage by developing a new product faster. However, they usually entail upfront investments that the firm cannot recoup if the product failed to meet certain quality requirements. Using a two-stage duopoly model, we examine under what conditions a firm will adopt concurrent development strategy in equilibrium in view of uncertain product approval, time-sensitive consumer utility, and asymmetric competition.

Gig economy workers make strategic decisions about where and when to work. We empirically measure two types of strategic behavior: multihoming, an online change between platforms, and repositioning, a physical change between locations. Using a structural model, we show that workers are highly heterogenous in their preferences for both multihoming and repositioning. We provide counterfactual estimates on the effects of proposed firm and regulatory policies aimed at multihoming and repositioning.

Taxis’ and Uber’s matching technologies differ: taxis random-walk in search of curbside pick-ups while Uber centrally dispatches drivers. We study how taxis can defend against Uber encroachment. We find that imitating Uber’s centralized dispatch can improve taxi market share but only when Uber drivers are relatively reluctant to drive. Otherwise, imitating Uber can entice more Uber drivers to drive, leading to an unintended reduction in taxi market share.

Tele-driving refers to a novel concept in which drivers can remotely operate vehicles. Because remote drivers can be operated as a shared resource, tele-driving has the potential to reduce the severity of the spatial mismatch between vehicle supply and customer demand that is often experienced in on-demand mobility services. In this paper, we compare a traditional ride-hailing system with one with tele-drivers, and quantify the potential gains that could be realized by tele-driving.

We study flexible versus dedicated technology choice and capacity investment of a two-product firm under demand uncertainty in the presence of subscription programs. With subscription programs, a proportion of customers allocated to a particular product are allowed to switch to the other product. We analyze how the switching proportion and demand correlation between two subscription demands affect capacity investment and profitability with each technology, and shape optimal technology choice.

High-tech products consist of many modules. We study an assembly system with one module sourced from a supplier with a fixed lead-time and one module produced in-house in a make-to-order (MTO) production system. Since unavailability of modules is costly, it is important to coordinate between the ordering policy for one module and the production of the other. We show optimality of an order policy for the lead-time module with base-stock levels depending on the state of the MTO production system.

The buffer allocation problem is a fundamental optimization problem if flow line planners need to cope with stochastic influences. Additionally, practitioners include spare part planning for manufacturing systems to increase the machine's availability directly. Hence, we tackle this crucial question and are the first to present a joint optimization of buffer capacities and spare part stocks for flow lines of arbitrary length. Among others, we generate new insights on spare part allocations.

The seller wants to sell an item to n buyers such that the buyers' valuation distribution is from a given class (i.i.d., mixtures of i.i.d., affiliated and exchangeable, exchangeable, and all distributions) and the seller minimizes worst-case regret. The first three classes admit the same minimax regret, decreasing in n, while the last two have the same minimax regret equal to that of the case n = 1. Across all settings, the optimal mechanisms are all second-price auctions with random reserve.

We study a new selling mechanism in online display advertising markets that combines both guarantee contracts (GC) and real-time bidding (RTB) and allow advertisers to strategically choose between these two channels. Despite the complexity due to the advertisers’ strategic behaviour and the auctions, we are able to obtain the closed-form solution and show that combining GC and RTB can generate more revenue for the publisher, compared to the case when only RTB or GC is used.

We consider a platform's problem of collecting data from privacy sensitive users to estimate an underlying parameter of interest. We formulate this question as a Bayesian-optimal mechanism design problem, in which an individual can share her (verifiable) data in exchange for a monetary reward or services, but at the same time has a (private) heterogeneous privacy cost which we quantify using differential privacy.

In this paper, we consider the question of how to better manage turnover among independent drivers who transport parcels for last-mile delivery platforms. We collaborate with a last-mile delivery platform to build a structural model that enables us to estimate the effects of key predictors of drivers' decisions to continue or leave the platform. For this estimation, we apply a dynamic discrete-choice framework in a two-step procedure that accounts for unobserved heterogeneity among drivers.

We introduce the `Driver-Aide (DA) Problem', a new mode of service operations in last-mile delivery. The use of a DA can shorten route durations, allowing larger delivery volumes without the need for additional vehicles. However, it is challenging to determine the best way to use a DA (as there are two different ways to use a DA) and evaluate the tradeoffs involved. We develop an optimization-based solution framework and conduct an economic analysis using data provided by an industrial partner.

In this work, we build a structural model to study customers’ preferences on prices and service features when choosing between private taxis and a scheduled shuttle service.

The omnichannel retail evolution has changed the role(s) of the store. Applying a quasi-experimental design to data on store openings by an omnichannel consumer electronics retailer, we explore these new store roles and the value they provide to the retailer. We find that, in contrast to small stores, large experience-centric stores substantially benefit online-first retailers through both customer acquisition and activation mechanisms. Category-level analyses reveal the underlying mechanisms.

This paper studies a monopolistic omnichannel retailer's pricing strategies when customers are strategic in making a channel choice and a purchasing decision in the presence of product value uncertainty. We find that charging a uniform price across the sales channels and disclosing it via the online store is optimal. We study the structural properties of the optimal price and the corresponding profit. Finally, we assess the value of omnichannel retailing relative to single-channel counterparts.

We examine how consumers’ fairness concerns affect a multichannel retailer’s pricing strategy. We find that the retailer should maintain consistent price across channels only when the fraction of unfair-adversed consumers is in an intermediate range, and otherwise should charge different channel prices. Moreover, as the fraction of unfair-adversed consumers increases, the retailer may be better off by strategically enlarging the price gap.

Reinforcement learning is a promising strategy for data-driven sequential decision-making. In many real-world applications, it is desirable to optimize objectives that account for risk in the achieved outcomes. We prove the first regret bounds for reinforcement learning algorithms targeting a broad class of risk-sensitive objectives, including the popular conditional value at risk (CVaR) objective. Our analysis relies on novel characterizations of the risk-sensitive objective and the optimal policy.

Missing information is inevitable in real-world data sets. While imputation is well-suited for statistical inference, its relevance for out-of-sample prediction remains unsettled. We analyze widely used data imputation methods and highlight their key deficiencies in making accurate predictions. Alternatively, we propose adaptive linear regression, a new class of models that can be directly trained and evaluated on partially observed data. We validate our findings on real-world data sets.

We study a newsvendor problem in which the decision-maker only observes historical demands. In contrast to the extant literature, we relax the i.i.d. assumption for past demands and assume instead that they are drawn from distributions within a distance r away from the future demand distribution. We establish an exact characterization of the worst-case regret of Sample Average Approximation. When r is small, we present a near-optimal algorithm which robustifies SAA by using less samples.

We predict that societal culture will have a significant impact on inventory management by (1) causing behavioral biases on individual decision makers and (2) affecting the organizational culture. In this paper, we look at the national inventory levels with a dataset that spans the globe. We find that all national characteristics have a statistically significant impact on managing inventory, some in counter-intuitive ways. We also discuss the impact of development status of nations.

Retailers in e-commerce are facing muti-sources of supply chain complexity, making accurate inventory records increasingly important while greatly challenged. This study systematically explores how supply chain complexity affects IRI using a hierarchical segmentation of the complexity sources in e-commerce. Our research contributes a hierarchical framework for supply chain complexity and complements existing literature regarding IRI by systematically analyzing its causes.

We asymptotically characterize and compare risk pooling approaches. We show that physical pooling dominates information pooling in settings with no additional per-location costs. In the presence of such costs, however, information pooling becomes a viable alternative to physical pooling. Through asymptotic analysis, we also address the grouping problem. The convergence of the expected total costs and the base stock levels are demonstrated through a simple numerical illustration.

We develop an optimization framework for infrastructure development to mitigate the risk of flooding caused by sea level rise and storm surge in a coastal area. Expected flood costs are included using a range of possible sea level rise scenarios, and investment costs are modeled for overall infrastructure development assuming budgetary limitations. Using the City of Boston as a case for this study, our methodology resulted in more than 90% cost reduction compared to a “do nothing” strategy.

We analyze the performance of simple (static) pricing policies for the joint pricing and inventory control problem. Compared to dynamic pricing policies, static pricing policies are more tractable, easier to implement and strategy-proof. We consider a continuous review system with Poisson arrivals of unit demand. We construct simple pricing policies that only increase inventory costs by a constant factor while actually increasing revenue, in comparison with the optimal dynamic pricing policy.

This paper investigates the impact of bilateral relationships on contracting incentives in a supply chain from a behavioral perspective. Our experimental data suggests systematic deviations from the theoretical benchmark and reveal behavioral regularities on contracting behavior. We develop a new behavioral theory where a firm's unfavorable bargaining position inflicts distress to a firm. We show that our behavioral theory explains and predicts supply contract bargaining incentives well.

We study the impact of assortment size on consumer choice decisions in an online recommender system context. Via a field experiment on Alibaba's online retail platforms, we causally show that the both consumers' search and purchase likelihoods first increase and then decrease as the number of options increases. We develop a two-stage consumer choice model and demonstrate that our empirical results are consistent with the predictions of a model that incorporates consumers' anticipated regret.

Appointment scheduling under uncertainty encounters a fundamental trade-off between capacity utilization and waiting times. In contrast to traditional approaches, we maximize capacity utilization while limiting waiting times. We derive a robust mixed-integer linear model, prove NP-hardness for the general problem, and optimality of well-known scheduling rules for special cases. Using real patient data, we show that our approach is a win-win solution to this fundamental trade-off.

Using distributionally robust optimization (DRO) that accounts for service times’ correlation, we study a two-sequential-server appointment scheduling problem. We observe that the optimal schedule can lead to imbalanced waiting times in the two servers, concentrating on the downstream server. To rebalance the waiting times without rescheduling patients, we adopt an idea in the queueing literature to strategically idle (SI) the upstream server. A DRO model is used to find the optimal SI policy.

We present a personalized and probabilistic framework to model the risk of hospital admissions for patients (with and without COVID-19) that attended an Emergency Department (ED) during the pandemic. As predictors, we use patient demographics, measures of ED crowdedness, and the triage information, and investigate if population-level data (such as human mobility, number of COVID-19 cases, vaccination status etc.) could help improve the prediction accuracy of admission risk at the patient level.

For COVID-19 detection, point-of-care tests are cheap and quick but fail policymakers’ accuracy requirements. We propose a robust optimization methodology for optimally combining results from cheap tests for increased diagnostic accuracy. Combining three rapid tests increases area under the curve by 6% compared with the best performing individual test for antigen detection. We demonstrate that robust optimization is a powerful tool to avoid overfitting and improve out-of-sample performance.

Adaptive approval is a novel regulatory program that enables earlier patient access to new drugs for rare diseases. The program has been in place for almost a decade, however, industry participation has been surprisingly low. We present a Stackelberg game-theoretic model to understand why industry participation has been low. Our results inform healthcare policymakers on ways to redesign adaptive approval programs to likely increase industry participation and improve patients' welfare.

In this project, we study the fairness implications of using machine learning to assist a human decision-maker. Relative to a baseline where machine decisions are implemented directly, we show in a formal model that the inclusion of a biased human decision-maker can revert common relationships between accuracy and fairness. Specifically, we document that excluding information about protected groups from the prediction may fail to reduce, and may even increase, ultimate disparities.

In many practical settings, a human reviews recommendations from a decision support algorithm and either approves or adjusts the recommendation. Automation may reduce a ML system's longer-term ability to predict effective adjustments and leads to predictive performance degradation over time. We (empirically) demonstrate this effect and show how to include the loss of learning into the automation decision.

We provide a stylized model in which a principal chooses a classifier, D, with known properties for a Bayesian decision-maker who observes the outcome of D before determining their own label in a binary classification problem. The decision-maker has a utility which deviates from the principal ’s whenever they take an action which contradicts the classifier. We characterize the optimal posterior decision and show how the optimal classifier for assistance depends on the decision-maker's prior.

When entrepreneurs introduce innovations, incumbents must respond, yet may fail to do so in due course. I characterize the incumbent's optimal policy, specifying the conditions under which an incumbent ignores the threat or decides to acquire or imitate the startup. Incumbent inertia may arise if the incumbent waits until the market is ripe or if it is ambivalent about the appropriate response. This second rationale has not previously been identified as a cause for incumbent inertia.

When composing their innovation portfolios, firms can rely on their internal R&D units and invest in projects that are promoted internally; or they can acquire projects that originated outside their boundaries. We ask: How should a firm allocate its scarce resources across the different sources? We investigate this decision by designing a stylized game-theoretic model, and we identify the firm’s optimal resource allocation policy.

Although the participation of firms in cooperative ventures that benefit all firms, such as industry alliances and generic advertising campaigns, has been well-studied in the literature, prior studies have not explored how the firms decide their participation levels when the underlying market conditions change. In this paper, we investigate the impact of boom and bust conditions of the market on two firms' strategic decisions, when they face the prospect of cooperating with their competitor.

We examine the effect of centralization on platforms' pricing decisions in two-sided matching markets. We develop a stylized model that describes the platforms' dynamic matching and pricing process over a continuum of market participants, which is directly motivated by the design of ridehailing platforms. We provide a comprehensive analytical characterisation of the market equilibrium. Next, we develop a simulation-based framework to compare the social welfare under three operating models.

We consider a truckload transportation marketplace in which a platform serves an intermediary to match shippers, who pay for transportation services, with carriers, who get compensation for transporting the loads. The objective of the platform is to design pricing and allocation mechanisms to maximize its long-run average profit. In this paper, we analyze the performance of posted price, auction, and hybrid mechanisms (which combine posted price and auction mechanisms).

The decision to display information of the freight market's current state impacts revenue due to user abandonment. The implications of such a decision are not well understood in platforms for freight exchanges. We study which information design maximizes expected revenue. Queueing models with balking and reneging are applied, and the steady-state behavior of the underlying Markov chains is analyzed. We find that in large markets, showing state information is preferred.

Intermittent demand is difficult to forecast, as many periods have no demand. The time between demands is often not memoryless but –contrary to implicit model assumptions—displays periodicity. Consequently, the time since the last demand is a predictor for future demand. We propose a demand model that accommodates such periodicity and show that the optimal inventory policy is a state-dependent base-stock policy, where the order-up-to-levels depend on the time since the last demand.

We study how managers’ potential personal costs due to shareholder scrutiny affect inventory policies exploiting a quasi-natural experiment. Using a staggered DiD approach, we find that firms incorporated in constituency states increased inventory by 5.2% relative to control firms, indicating a heightened focus on customer service levels. To our best knowledge, our paper is the first to study managerial incentives pertaining to inventory management in a quasi-natural experimental setting.

We study a discrete pricing setting where each customer is offered a contextual price. Often only historical sales records rather than customer valuations are available, where the data is influenced by the previous pricing policy. This introduces difficulties in estimating revenue. We approach this problem using ideas from learning with corrupted labels to formulate loss functions that directly optimize revenue, rather than going through an intermediate demand estimation stage.

Problem definition: This paper investigates the joint estimation of the consumer arrival rate and choice model parameters when ``no-purchasers” (customers who considered the product but did not purchase) are not observable. Estimating demand even with the simplest discrete-choice model such as the MNL becomes challenging as we do not know the fraction that have chosen the outside option (i.e., not purchased).

Mining for heterogeneous response to treatment is a crucial step in data-driven operations. We propose a partitioning algorithm to estimate heterogeneous odds ratio, a popular measure when response to treatment is binary. We integrate an adversarial imputation step to account for partially observed treatments (e.g., if full information is only available for purchased items). We validate our methodology on synthetic data and case studies from political science, medicine, and revenue management.

Disputes can be a common occurrence in online labor platforms due to users' gaming behavior and disagreement over contracting terms. While traditional platforms resolve disputes using a centralized approach, there are emerging platforms that relegate dispute resolution to independent platform users through a voting mechanism. We study when and why this decentralized approach can be better for the platform and the social welfare and how the platform should adjust the dispute fee when adopting it.

We consider the optimal operating policies of a free-to-play game. Accounting for social comparisons between free and premium players, we model the game attracting or losing players. We characterize optimal dynamic pricing and advertising policies and show that the developer should prioritize initial growth through aggressive advertising while postponing the introduction of premium subscription. Surprisingly, the optimal subscription price may start high and gradually decrease.

In many service encounters front-line workers have the discretion to attempt to resolve a customer request or to transfer the customer to an expert service provider. We experimentally examine this decision. Our experiments show that transfers are too low under some incentive systems. However, transfer behavior responds correctly to congestion information. Taken together, these results advance our understanding of cognitive capabilities and rationality limits on human server behavior.

We analyze a retailer’s interactions with a national brand manufacturer (NBM) using a setting in which the retailer makes category management and store brand (SB) sourcing decisions and the NBM strategically determines whether it should produce the retailer’s SB. Our analysis sheds light on different SB strategies observed in practice.

The majority of online sales worldwide take place in online marketplaces that connect many sellers and buyers. Online marketplaces adopt algorithmic tools to curate how the different options in an assortment are presented to customers. This paper focuses on one such tool, the Buybox, that algorithmically chooses one option to be presented prominently to customers. Our analyses show that the Buybox produces benefits for customers, sellers, and the marketplace.

Social commerce integrates user interactions and user-generated content with commercial activities in the context of social media platforms. Examples include the "shop" feature on Instagram. A social media user's on-site purchase decision involves a transformation of the mindset from "social" to "shopping" stimulated by the impulse to purchase. We propose a novel choice model to capture users' shopping behavior on social media sites and examine two strategies to sell through social media.

Decision-makers often face the "many bandits" problem, where one must jointly learn across related but different contextual bandit instances. We study the setting where the unknown parameter in each instance can be decomposed into a global parameter plus a local sparse term. We propose a novel two-stage estimator exploiting this structure efficiently using robust statistics and LASSO. We prove that it improves regret bounds in the context dimension, which is exponential for data-poor instances.

We consider the problem of maximizing charity donations with personalized recommendations and unknown donor preferences. On charity platforms, a donation is observed only when the recommended campaign is selected by the donor, and an eventual donation is made, leading to selection bias issues. We propose the Sample Selection Bandit (SSB) algorithm that uses Heckman's two step estimator with the optimism to resolve the sample selection bias issue.

We explore a new model of bandit experiments where a potentially nonstationary sequence of contexts influences arms' performance. Our main insight is that an algorithm we call deconfounted Thompson sampling strikes a delicate balance between adaptivity and robustness. Its adaptivity leads to optimal efficiency properties in easy stationary instances, but it displays surprising resilience in hard nonstationary ones which cause other adaptive algorithms to fail.

The timeline of product recalls provides corporate insiders with an opportunity to sell stocks before the market reacts to a recall. In this paper, we examine whether such insider trading occurs during the product recall process. Our results show that insider trading is present and that directors but not officers seem to engage in it in the days following defect awareness. Thus, we identify the product recall process as a novel source of information that insiders exploit for personal gains.

How should manufacturers sell network goods through retailers? We study this new supply chain contracting problem, where the retailer can privately observe and control the evolving market conditions. The optimal contract resembles the second-best in the short run, but converges to the first-best in the long run. We guide practice why manufacturers should over supply, mitigate network effects, favor incumbent retailers, and improve retailer information capability, despite information asymmetry.

Insurers have started incentivising customers for buying smart home security products along with home insurance to achieve a reduction in hazard likelihood. In this paper, we study the discounting decision of the insurer and pricing and quality decisions of the smart product manufacturer for offering "smart home insurance" to customers under no-contract, a Wholesale price contract and a Cost-sharing contract, considering (1) equal market power and (2) having a dominant SPM separately.

This work presents a computational study of two-stage stochastic programming and budget-uncertainty robust optimization for capacitated lot-sizing under uncertain demand. To solve the stochastic models, a Benders decomposition approach is tailored to the problem. The tradeoff between computational time and performance on out-of-sample scenarios is investigated. Managerial insights are provided by analyzing the structure of the obtained production plans and the impact of flexibility in planning.

This paper proposes an integrated methodology to optimize tactical production planning and strategic buffer placement in complex capacity constrained high-tech manufacturing supply chains. We introduce a novel multi-stage stochastic programming model that simultaneously tackles demand and lead time uncertainty. For extended planning horizons, we establish a data-driven rolling horizon-based decision framework to derive efficient buffer replenishment policies for varying service levels.

We study frictions in foreign exchange market. We consider two companies with opposite needs of currencies. They can negotiate an exchange between themselves. Forward contract is where negotiation happens prior to randomness resolution, while ad-hoc contract after. The forward contract has a larger potential in increasing quantity decisions due to prior commitment. Ad-hoc contract leads to either unique or continuum of equilibrium(a). Payoff dominance uniquely selects an equilibrium.