A critical decision made by many firms is whether to adopt a responsive supply chain (prioritizing speed) or an efficient supply chain (prioritizing cost). We consider the environmental implications of this choice, and find that firms will have the greatest incentive to invest in responsiveness when it is most detrimental to the environment. We discuss the implications of this for policymakers seeking to encourage firms to use supply chains that generate the least environmental impact.

We investigate a Remanufacturing Original Equipment Manufacturer (ROEM) who can choose

to remanufacture recovered cores either to their original configuration or to current technology. The decay of consumers’ valuations of the products as time passes influences the optimality managerial decisions. Hence, we examine the traditional method of studying this effect and develop a more realistic model that offers new insights into the optimal remanufacturing choices.

This paper extends our research on “bespoke” resilience strategies, by formulating a supply chain model that enhances reported models by adding sustainability. The proposed model examines tradeoffs, constraints, and risks for the extended problem and considers implications for supply chain strategy development. One key question was whether the two concepts are mutually reinforcing or conflicting. Our analysis shows how the answer depends on the features of the supply chain environment.

We consider the problem of designing a sequential clinical trial with a fixed budget in order to find the best treatment as a function of predictive and prognostic patient covariates. We propose computationally tractable heuristics based on the expected value of information that perform well and are asymptotically optimal in the limit of large sample size. We show the benefits of incorporating predictive and prognostic covariates in allocation policies for learning the best treatment strategy.

We exmaine potential learning at the manufacturing facility level. Considering drug shortages as a manufacturing disruption, we investigate the two sources of learning: (1) experience of recovery from disruptions in the past and (2) experience of recovery from on-going disruptions. In addition to these learning effects, we also examine whether the two learning sources are substitutes or complements, and how the diversity of disruption resolution experience moderate these learning effects.

Many US states have adopted telehealth pay-parity policies requiring payers to reimburse healthcare providers equally for telehealth and office visits. But telehealth may require a duplicate visit for a physical exam. We analyze a three-stage game to study the impact of telehealth reimbursement on provider's operational decisions, where patients choose strategically between telehealth and office channels. We find that pay parity can decrease patient access and discuss its policy implications.

We present a Markov decision model for simultaneously managing the treatment of multiple chronic conditions. We first provide a general framework, then demonstrate under which conditions model complexity can be dramatically reduced--including to that of an index optimal policy. We present a case study on patients with age related macular degeneration, and demonstrate that following our framework could reduce symptoms by 38% and direct medical costs by 23%.

Mobile labs are a promising approach to improving access to health. Although there is a variety of use cases for mobile labs, their usage has been primarily in emergency deployment. This means mobile labs are at risk of being idle if not used in non-emergency settings. We analyse operational models for non-emergency mobile lab deployment. Our results show substantial impact can be generated and help inform decision-making regarding pathogen prioritization and operational models.

Supply chain risks, despite being a well-developed theoretical concept, are difficult to empirically quantify. I develop a firm-level measure of supply chain risk exposure using textual analysis on a novel source of unstructured data: managers' discussions during earnings conference calls. Economically validated, the measure provides a credible quantification of firm-level exposure to supply chain risks, and can be reliably utilized as outcome or explanatory variables in empirical research.

Weather has a significant impact on the demand of various products. To improve the performance of the supply chain of weather sensitive products the main objective of the proposed research is to design new classes of contract that can outperform the widely used contracts, such as wholesale price, buyback, and revenue sharing contracts. . A firm’s objective under risk is measured by using the Conditional Value at Risk. Results show that the designed contracts outperform the traditional contracts.

Gathering information on risks in a supply chain is still a significant challenge for firms. However, firms are requested to disclose their material risks in 10-K reports, leading to a substantial amount of information on their risk status. It remains an unanswered question if the information is informative. Using the natural experiment presented by the COVID-19 pandemic and an event study methodology, our analysis provides evidence that risk disclosures are informative.

Crowdsourced delivery platforms have made use of two types of couriers: ad-hoc couriers, who are more flexible, and committed couriers, who are more reliable. In this paper we show that by designing a system that intelligently utilizes order partitioning between the two delivery channels (e.g., makes routing and partitioning decisions jointly), the delivery platform can exploit the benefits of each courier base to improve customer service and reduce the total cost when compared to order pooling.

On two-sided matching platforms such as VolunteerMatch (VM), a sizable fraction of website traffic arrives via an external link, bypassing the platform's recommendation algorithm. We study how platforms can account for this, given the goal of maximizing successful matches. We model the problem as a variant of online matching and introduce an algorithm providing near-optimal guarantees in certain parameter regimes. We also show our algorithm’s strong performance in a case study based on VM data.

We propose a dynamic car relocation policy for a car-sharing network with centralized control and uncertain, unbalanced demand. The policy is derived from a reformulation of the fluid model approximation of the dynamic problem. We project the full-dimensional fluid approximation onto the lower-dimensional space of relocations only. Our policy exploits these gradients to make dynamic car relocation decisions. We close the optimality gap on average by 30% in static and time-varying settings.

In bilateral asymmetric information sharing, a retailer has private demand and a supplier private capacity information. We propose non-financial information-sharing mechanisms without power structure and examine analytically how one’s sharing affects the other’s sharing for cooperative, sequential sharing, and under risk aversion. We find that the retailer shares if demand is higher than a threshold. The supplier shares if capacity cost is within a range of upper and lower cost thresholds.

We consider inventory allocation of multiple products, across a network of warehouses. We propose a multi-period, multi-product newsvendor formulation over a network of capacitated warehouses with depth constraints, minimizing the e-tailer’s shipment cost. The efficient algorithm we propose balances the tradeoff between overage and underage costs across periods. We establish its rate of convergence and in collaboration with a fashion e-tailer, we perform a study showing a cost reduction of 9%.

Despite the potential of a battery energy storage system (BESS) to electrical grids, most standalone use of BESS is not economical due to its high upfront cost and batteries' limited lifespan. Energy stacking, a strategy providing multiple services simultaneously, has been of great interest to improve profitability. However, some key questions remain unanswered. We show that there exists cost-saving synergy, which enables stacking to double the profit of the best standalone service.

Solar power has risen as a sustainable & inexpensive option, but its generation is variable during the day and non-existent at night. Thanks to recent technological advances, a combination of solar+storage holds the promise of cheaper, greener, and more reliable off-grid power. Our work sheds light on this question by developing a model of strategic capacity investment in renewable generation and storage to match demand with supply in off-grid use-cases, while relying on fossil fuel as backup.

We examine whether and how renewable energy adoption affects energy efficiency (EE) improvement. Using site-level data from an industrial conglomerate, we find that using renewables to meet 10% more of a site's energy demand led to an additional 2.0% improvement in EE. This effect is heterogeneous in sourcing strategy where outside purchases led to gains, but on-site generation had no effect. Analysis of the mechanism suggests greater managerial focus on EE due to the costs of outside purchases.

We study the problem of designing experiments that are conducted on a set of units, such as users in an online marketplace, for multiple time periods. We first study the optimal design of experiments, to most precisely estimate the instantaneous and lagged effects, post-experiment, when treatment decisions are made before the experiment starts. Next, we study the design of sequential experiments, where adaptive decisions are allowed, and the experiments can potentially be stopped early.

While observational data from queueing system is of great interest for statistical inference of arrival and service processes, the queueing dynamics and the absence of distributional information render queue estimation remarkably challenging. To this end, we propose a robust optimization based framework for inferring service times from waiting time observations. We provide conditions under which our framework produces statistically consistent estimators and present its managerial insights.

We consider an online retailer and characterize policies, which specify the options (selling price and delivery lead time) to be offered to customers. In the static M-option policy M options are set at the beginning of the planning horizon and the decisions on when to offer them are made dynamically. In dynamic single-option policy the retailer offers a single dynamic option. We propose algorithms to optimize the policies and evaluate their benefits.

Online customers like to receive baskets of grocery orders at a confirmed time. Online retailers increasingly offer customers a choice of leadtime, while actively backordering missing items from the baskets. This fundamentally changes strategic inventory management. We develop new allocation policies that commit to an order upon arrival rather than at the moment the order is due. We give analytical results for the performance of these policies and evaluate them with e-tailer data.

Recently, several omnichannel retailers face the growth of online sales through in-store picking. We tackle a new relevant problem where a picker picks online orders while minimizing customer encounters. The problem is modelled as a Markov decision process and solved using a Q-learning approach. Results on a real retail store suggest that retailers should scale in-store picking without jeopardizing offline customers' experience. However, choosing simplistic picking policies is not sufficient.

We study the problem of transforming offline algorithms to their online counterparts, focusing on offline combinatorial problems that are amenable to a constant factor approximation using a greedy algorithm that is robust to local errors. We provide a general offline-to-online framework using Blackwell approachability, producing T^1/2 regret under the full information setting and T^2/3 regret in the bandit setting. We apply our framework to operations problems and produce improved regret bounds.

In this work, we discuss Programmable Actor Reinforcement Learning (PARL), a policy iteration method that uses techniques from integer programming and sample average approximation. We numerically benchmark the algorithm in complex supply chain settings where optimal solution is intractable and show its performs comparable to, and sometimes better than, state-of-the-art RL and commonly used inventory management benchmarks.

Motivated by the successes of deep learning, we propose a class of neural network-based discrete choice models, called RUMnets, which is inspired by the random utility maximization (RUM) framework. This model formulates the agents' random utility function using the sample average approximation (SAA) method. We show that RUMnets sharply approximates the class of RUM discrete choice models. We provide analytical and empirical evidence of the predictive power of RUMnets.

The Operating Rooms (ORs) are some of the most expensive and resource intensive areas of healthcare delivery. We developed and implemented a unique method to overbook surgeon blocks, i.e. assigning more OR block time than we have OR capacity to improve OR access. Using Mixed Integer Quadratic Programming, we spread the risk of overbooking equitably and enable block overbooking through an automated process for the front-line to manage and sustain easily.

To efficiently schedule operating rooms (ORs) in complex inpatient surgical departments, we consider separating the more predictable elective surgeries from the less predictable elective and non-elective surgeries. We solve this problem by heuristics and simulation. Using the data from a university hospital, we find that the partitioning can considerably reduce the cancellation rate and can fairly reduce the elective access times without much damaging the non-elective access times.

Admission and discharge decisions play important roles in hospital intensive care unit (ICU) bed capacity management. In this model, we formulate the readmission of patients as an endogenous process that relies on previous discharge decisions. We develop a model to jointly consider early discharge decisions and admission control, including emergency diversion and elective scheduling. By applying the riskiness index, we can reformulate the problem and solve it efficiently.

This work examines how health insurance companies (HICs) can exert effort and offer cash rewards to nudge patients towards cost-effective providers. We build a stylized model to analyze the HIC’s optimal effort and reward, individually and jointly, under different cost-share structures. We find that neither a reward-only nor an effort-only approach uniformly outperforms the other, and HIC strictly benefits most from the joint approach when the price difference is modest.

Patients with cystic fibrosis disease experience frequent pulmonary exacerbation and require antibacterial treatments. Intravenous aminoglycosides are the primary choice but they cause hearing loss. To detect possible hearing loss, there are several hearing assessment methods available. The overarching aim of this research is to design cost-effective strategies to monitor pediatric patients with CF disease to detect potential hearing loss and improve their quality of life using a hearing aid.

We study incarceration diversion decisions with community corrections as an alternative to jailing to alleviate the prominent issue of jail overcrowding. We formulate an MDP model to optimize incarceration diversions for individuals of different risks. To tackle the curse-of-dimensionality caused by non-memoryless, we develop a novel two-timescale approximation embedded in an actor-critic policy gradient algorithm. We provide structured insights for diversion decisions and service capacities.

Personalized intervention management in healthcare has received a rapidly growing interest. A key challenge for personalization is data scarcity. In this research, we develop data-pooling technique in the reinforcement learning (RL) context to address the small sample issue. We develop a novel data-pooling estimator and establish theoretical performance guarantee. We demonstrate its empirical success on a real hospital dataset with an application to reduce 30-day hospital readmission rate.

Using 50K observations of purchase tasks, we examine the mechanism behind purchasers’ decisions on the timing of the orders. Our analysis focuses on the impact of exposure to suppliers and rare events throughout the history of experiences of the purchasers. We find that exposure to specific suppliers increases the safety time for the current order; rare events in the prior orders are increasing the safety time. We also observe that purchasers rely on self-experienced orders rather than peers.

In recent applications of Markov decision processes, it is common to estimate model parameters from data. When data are collected from a population, one faces a modeling question of whether to estimate different models for subpopulations. This work provides theoretical results and empirical methods for making the decision of whether to model subpopulations separately or not. We also present how to use our results to select the best stratification and empirical results using various instances.

Informal recycling of used lead acid batteries causes tremendous environmental damage, especially to children’s physical and mental developments in Bangladesh. The problem is further exacerbated as lead extracted from the process is used to produce low-quality batteries that require frequent replacements. We study public policy interventions that give incentives to extend battery lives and promote formal recycling under strong environment production, to reduce the circulation of informally-recycled lead.

We are designing a novel business model to extend the life of lead acid batteries used in electric three wheelers in Bangladesh. Through a randomized control trial, we are testing the impact of the business model on battery life and performance, recycling rate and lead emissions, energy consumption and CO2 emissions, and income and profit of battery users. The business model can be generalized to other durable goods and geographies where products fail fast because of market inefficiencies.

The initial landing location of a refugee has implications on their long-term success. We propose two new dynamic algorithms to match refugees to localities within a host country. The performance of the proposed methods is illustrated on real US refugee resettlement data. The first algorithm maximizes the average employment level, and is currently deployed in a pilot in Switzerland. The second algorithm balances employment with the desire for an even allocation to the localities over time.

Current clinical guidelines for hypertension provide physicians with general suggestions for first-line pharmacologic treatment, but do not consider patient-specific characteristics. We utilize electronic health record data to develop personalized predictions and prescription models for hypertensive patients. We demonstrate a 15.87% improvement over the standard of care and propose a novel interactive dashboard to facilitate the deployment of the derived models in the clinical practice.

Rapid and accurate detection of local outbreaks is critical to tackle resurgent waves of COVID-19. A fundamental challenge in case growth rate estimation, a key epidemiological parameter, is balancing the accuracy vs. speed tradeoff for small sample sizes of counties. We present “Transfer Learning for Generalized Random Forests” (TLGRF), a novel framework which uses relevant features affecting the disease spread across time and counties to obtain more robust and timelier county-level estimates.

We address a set of problems that occur in the development of liquid biopsies via the lens of active sequential hypothesis testing (ASHT). Motivated by applications in which the number of hypotheses or actions is massive, we propose efficient algorithms and provide the first approximation guarantees for ASHT, under two types of adaptivity. We numerically evaluate the performance of our algorithms using both synthetic and real-world DNA mutation data.

Motivated by recent development in predictive analytics, we study how to utilize future demand information to design optimal routing strategies when facing demand surges. We consider a multi-class multi-pool parallel server system with partial flexibility, where overflowing a customer to a non-primary server-pool can be associated with efficiency loss and other costs. Our results explicitly characterize how to incorporate future demand into routing decisions and quantify the benefit of doing so.

Motivated by the need for decision support tools for workload prediction and capacity planning in hospitals under the COVID-19 pandemic, we consider a queueing network consisting of two many-server stations, which models the flow of patients between medical/surgical and intensive care unit wards. We approximate the steady-state customer count using the stationary distribution of the associated diffusion model, which can be computed efficiently and be used as a real-time decision support tool.

Motivated by the challenges in census prediction in data-driven settings for hospital resource management, this paper introduces and studies the patient census calibration problem for hospital networks operating in a random environment. We use the expectation-maximization method to efficiently solve the calibration problem. We present simulation results that demonstrate the efficiency and accuracy of the method, and theoretical analyses that provide large-sample statistical guarantees/

An action associated with a small cost is commonly taken by humans for preventing a potential big loss in a wide range of domains. For example, technicians decide whether to do preventive maintenance now or to wait for the next updated machine status information. We model such decisions as a dynamic cost-loss game. We analytically show that using a probability threshold is an optimal policy. We next conduct a controlled laboratory observing how human decision makers behave in this environment.

Resource-based contracts fail to motivate suppliers to provide reliable services as they are paid for after-sales services. Performance-based contracts shifts much of the downtime risk to the supplier by making him responsible for machine uptime, but then customers might reduce care efforts. We propose the full-care contract to achieve both high reliability and care, and show that the FCC not only the best contract analytically but also can motivate risk averse suppliers to bear the chain risk.

Using a behavioral game of supply rationing, we investigate the consequences of sharing information about competing retailers’ historical orders on the Bullwhip Effect. We find that decision makers act more strategically and closer to Nash equilibrium predictions if information about competitors’ historical orders is shared; however, sharing the entire order history does not accelerate the convergence to theoretical Nash equilibrium as much as sharing orders from the last period only.

We examine the interplay between operational effort to increase supply chain resilience and financial arrangements such as buyer intermediated financing, buyer direct financing, and advance payments. We characterize conditions when those arrangements complement operational effort and when they are substitutes. We study when those arrangements lead to coordination or at least efficiency gains for buyers and their upstream supply chain.

Fixed-price freight contracts between firms and transportation providers are non-binding and often fail due to supply and demand uncertainties. We propose indexed pricing to reduce frequent unanticipated costs and supplier performance degradation. We show how to design these contracts for a Pareto improvement over traditional contracts, conduct an experiment with an agricultural firm to validate the models, and quantify the causal effect of indexed contracts on costs and performance.

Modeling of resilience and viability became crucial in case of supply chain disruptions. We discuss research-practice paradoxes and show that the categories of probability, adaptability, and time are major determinants of resilience and viability modeling. We stress the importance of reliable suppliers, disruption probabilities, disruption time and ripple effect estimation, value-creation perspective of resilience, and viability of intertwined supply networks.

Investments in PV solar systems in the residential sector in Egypt are not thriving as expected despite the rapid decrease of the capital cost, electricity tariff reforming, and recent regulations for grid connection. This research targeted the residential sector in Egypt as it consumes 47% from the nation's electricity (IRENA, 2018), and explored the decision-making factors that affect the adoption of solar energy technology by the upscale residential sector.

Many developing countries lack the health-emergency infrastructure of the developed world. In this context, our industry partner Flare (operating in Nairobi, Kenya) coordinates existing ambulance providers by operating a platform. We study the operational challenges for such platforms as they often lack the knowledge about all ambulances' future availability and their location at a tactical level, and typically do not fully control these ambulances.

From cloud computing to Covid quarantines, requests for service can arrive in batches. How should this impact the service's staffing? Here, we find that there is no economy of scale as batches grow large, a stark contrast with classical square root rules. By consequence, the queue length is not asymptotically normal; in fact, the fluid and diffusion-type limits coincide. When arrivals are both quick and in batches, an economy of scale can exist, but we show that it is weaker than expected.

Motivated by hotel housekeeping, we study shift construction decisions for room attendants amid uncertainty about customer arrival and departure times. We provide analytical results using the framework of M-convexity. A numerical case study for one hotel suggests that reallocating a small number of workers to later shifts can effectively eliminate guest waiting after the posted check-in time. We also identify alternate optimal solutions that can be useful for recruiting and retaining workers.

We consider the multi-period problem of jointly allocating resources to J supply nodes and assigning jobs of I different demand origins to the nodes. The goal is to maximize rewards for matching or minimize costs of waiting and assignment. We introduce a distributive decision rule, which represents the proportion of jobs served by each of the supply nodes. We test against benchmark models developed specifically for allocation or assignment decisions only and record 1-15% reductions in costs.

We study a general model of reusable resource allocation problems. Customers of different types arrive according to a stationary stochastic process. The firm's goal is to maximize multiple kinds of rewards generated by customers. We develop an online policy for deciding on actions to take without knowing the distribution of customer types and show that when the usage duration is small compared with the length of the planning horizon, our policy achieves a near optimal performance.

We study the multisecretary problem with a capacity to hire up to B out of T candidates with their values drawn i.i.d from a distribution F on [0,1]. We investigate the achievable regret performance where our benchmark is the offline optimal policy. We establish the insufficiency of the common certainty equivalent heuristic for distributions with many types and gaps. We devise a new algorithmic principle called "Conservatism wrt Gaps" and use this to derive near-optimal regret scaling.

Prophet inequalities are a useful tool for designing online allocation procedures and comparing their performance to the optimal offline allocation. In this paper we derive the best-known guarantee for $k$-unit prophet inequalities for all $k>1$. We also provide a tight resolution to the related Magician's problem. Finally, we improve the guarantee from 0.2 to 0.319 for online knapsack, and from 0.321 to 0.3557 for unit-density online knapsack.

A residential construction firm's product mix and land investment decisions are highly complex due to the need for long-term planning. We study these decisions using a 3-stage capacity-constrained stochastic optimization model with a heterogeneous consumer base under willingness-to-pay distribution uncertainty. Among others, we find that the land investment increases with uncertainty. We also discuss the impact of competition and housing affordability regulations and find non-intuitive results.

We empirically analyze price bargaining for houses between the company and individual buyers. In each bargain, the seller and buyer take turns to make concessions until one of them terminates the bargain by accepting the other’s offer or by exiting it. We relate concessions to compromises via reservation prices and then measure bargaining powers through compromises. We identify house and buyer types and obtain type-dependent estimates of reservation prices and bargaining powers.

Visitor experience in museums is complex, where the utility procured by an artwork depends on multiple artistic, layout-related, and environmental factors. In this paper, we build a framework that analyses sequential choices made by the incumbent when the options are made available in a physical space. By applying it to the context of museums, we are able to study the effect of the curatorial decisions on visitor engagement and build counterfactuals for identifying better layout configurations.

Combining the insights from our survey of academic research and interviews with practitioners, we provide directions for future academic research that take advantage of the availability of big data. Future research on retail analytics can contribute to existing work by: (i) studying new decisions, (ii) using more advanced analytics, (iii) leveraging new data sources, or (iv) applying more sophisticated methods.

Apparel retailers have generally followed one of two supply chain approaches: the traditional approach (lacks quick response capabilities and produces more durable products) or the fast-fashion approach (has quick response capabilities and produces less durable products). Using an infinite horizon game theoretic model, we compare the profitability and environmental impact of the two approaches. We characterize win-win scenarios (higher profit and lower environmental impact) for both approaches.

We conduct a field experiment to examine whether algorithms should be automated or be used to augment human decision-makers. Unlike the common practice of allowing managers to override the output of algorithms, we allow retail merchants to override the inputs in order to capture the private information they possess. Our results show that the input augmentation model increases profitability by nearly 4% compared to the automation model where merchants were not involved.

A contingent free shipping (CFS) policy offers free shipment of an order only if it satisfies a pre-specified threshold amount. Our study empirically documents a novel determinant of optimal CFS terms: ease-of-return experience. To reflect its impact on the CFS policy’s embedded trade-offs, a manager shall apply the following counterintuitive adjustment; set lenient (resp. stringent) CFS terms when the customer return process is convenient (resp. inconvenient).

We present our empirical findings on how bundle promotions affect consumer purchase and return behavior compared to markdowns, using a large apparel brand’s in-store purchase and return panel data. In this work, we show that bundle promotions increase the incidence and decrease the return probability of each product compared to products sold with markdowns, controlling for price, discount depth, and item characteristics.

Customers use on-line product reviews more frequently. We explore the impact of product reviews on customer’s valuation uncertainty for an experience product and how that in turn affects a monopolist retailer’s pricing and refund decisions. In a duopolistic competition, the overall sentiment of the reviews are influenced by both retailers. We show that the retailers make their returns more lenient in the presence of product reviews in both settings.

We rigorously investigate whether and to what extent variations in the (i) the physical context where training is administered (i.e., onsite vs. offsite training); (ii) the familiarity of the trainer who administers the training (i.e., familiar vs. unfamiliar trainer); and (iii) the timing of the week when training is administered (i.e., early-week vs. mid-week vs. late-week training) impact the learning outcomes of health workers and subsequently the likelihood of health commodity stock-outs.

We investigate how the Degree of Standardization across service chain-belonging units impacts performance outcomes. We find that nursing homes that customize service delivery and standardize customer mix tend to experience improved financial performance; those that standardize customer mix tend to experience improved clinical outcomes, and those that customize service delivery tend to experience enhanced resident welfare.

We develop a queueing model to represent a clinic with two types of strategic patients who choose between making an appointment, incurring type-dependent indirect wait cost, and walking in, bearing an inconvenience cost and a risk of being rejected. We focus on the clinics' decisions to allocate slots to walk-ins and appointments to maximize their revenues. The system is analyzed under observable and unobservable settings. The model assumptions and results are examined via a simulation platform.

This study improves our understanding of how workload affects service speed by analyzing patient flow through the ED at a high resolution. We exploit a novel dataset and a nonparametric ML method to track multiple dimensions of workload in realtime. We find the service rate resembles the hazard of a log-normal distribution, nurse load has a greater impact on service speed than physician load, which in turn has a greater impact than system load, and a clearer picture of the system workload.

A key challenge to deploying reinforcement learning in practice is safe exploration. We propose a reinforcement learning algorithm that provably satisfies a safety constraint where it uniformly improves performance at each iteration while achieving sublinear regret. We experimentally validate our results on a sepsis treatment task and an HIV treatment task, demonstrating that our algorithm can learn while ensuring good performance compared to the baseline policy for every patient.

Our study is the first to empirically study the impact of the minimum wage on firms’ scheduling practices. Using a highly granular dataset from a chain of fashion retail stores, we estimate that a $1 increase in the minimum wage, while having a negligible impact on the total labor hours used by the stores, leads to a 27.7% increase in the number of workers scheduled per week, but a 19.4% reduction in weekly hours per worker, and less consistent schedules, which substantially hurt worker welfare.

Under current legislation, multinational companies are at risk of having imports into the U.S. blocked due to alleged use of forced labor in their supply chains. Using a game theoretic model, we study the equilibrium interactions between firms, who may exert costly 'responsibility effort,' and enforcement organizations that allocate scarce resources to investigate multiple firms. We characterize policies that incentivize greater supply chain responsibility based on firm and industry parameters.

How shall gig workers be classified? Compared to the benchmark of contractors, we show that uniform classifications (employees, contractors+) suffer issues of worker’s being undercut and overjoining and will not always make vulnerable workers better off. To classify workers according to their needs, or operationally prioritizing vulnerable workers can Pareto improve over uniform classifications. Our work highlights the importance of worker-specific regulations in the on-demand economy.

We enhance the multi-armed bandit model by considering endogenously non-stationary rewards – specifically rewards that are parametric functions of policy histories (learning). We further incorporate queueing costs, fairness, and arm correlation. We propose the L-UCB, FL-UCB, and QFL-UCB algorithms to solve our model, prove its logarithmic regret, and apply it to split-liver transplantation.

We adapt online learning techniques to scenarios with time-to-event data, where there is a delay between choosing an arm and observing feedback that is endogenous to the quality of the arm. We posit a multi-armed bandit algorithm with a cox-proportional hazards estimator, prove guarantees on the regret under this algorithm, and analyze its performance on a dataset of metastatic breast cancer clinical trials, comparing it to that of other adaptive allocation schemes.

Lack of treatment adherence significant barrier to reducing the global disease burden of tuberculosis (TB). We study the design of targeted interventions for a treatment adherence support platform running in Kenya, whose goal is to help patients on TB treatment. We show empirically that there is large heterogeneity in treatment effects of interventions, and we devise a novel online learning policy based on Thompson Sampling that significantly outperforms the currently employed policy.

Capitation payment models have been increasingly adopted by the payers in the U.S. healthcare market during the past decade. Through a Difference-in-Difference (DID) design, this paper empirically demonstrates that Medicare Advantage (MA), the largest healthcare capitation program in the U.S., inadvertently incentivizes MA health plans to reallocate parts of the capitation payments from the sick to cross subsidize the healthy, a practice to which we refer as reverse cross subsidization.

Emergency transportation systems in developing countries often lack the capacity and coordination to serve patients.

We study the market entrance decision of an entrepreneur into an ETS in a region where independent ambulance providers compete for demand. The entrepreneur may decide (i) to acquire own ambulances to become a competing service provider, (ii) to operate a pure platform that exclusively coordinates existing providers, or (iii) to coordinate and compete by combining both strategies.

Contact centers face operational challenges - proxies for customer experience are subject to uncertainty. A main source is silent abandonment customers (Sab). Sab leave while in queue with no indication. As a result, capacity is wasted. We develop methodologies to identify Sab and to estimate patience. We show how accounting for Sab in a queueing model improves the estimation accuracy of key measures of performance. Finally, we suggest strategies to operationally cope with Sab.

In the service industry, it is common that companies send agents to perform tasks at customer locations. Thereby, many companies rely on their agent's ability to construct tours manually. These manual tours are often non-optimal, causing high travel cost. In our work, we developed a new agent-to-customer assignment approach, designed to promote the manual construction of distance minimal tours. In our experimental study we investigate its effect on the human routing performance.

We study how consumers decide to pay with loyalty points or money by developing a model and estimating it on airline loyalty program data. We find that mental accounting, subjective perceived value, and reference exchange rate of points play important roles, and consumers’ primary points earning source and total earning level are jointly associated with their attitudes toward points and money. We show how a firm can implement pricing policies to efficiently influence consumers’ payment choices.

We compare two microfinancing structures in developing countries where an entrepreneur with zero initial budget borrows a loan to start a business. The entrepreneur faces a Newsvendor problem with finance and effort considerations. We characterize conditions under which a community bank, which can apply social pressure on the entrepreneur to pay all of its debt back, improves individual and social welfare in comparison with a social bank, which has no such mechanism.

Loans can cause bankruptcy risk in capital constrained businesses. This study presents three data-driven contracts that enable us to capture the trade credit and bank credit risks by trade credit insurance and payment protection insurance, respectively. Analyses underscore the significance of using insurance services as risk hedging tools and ensuring the business continuity of supply chain players. Moreover, retailer prefers to receive trade credit if supplier purchase insurance services.

Reverse factoring is increasingly relevant in the industry and examined in academia. We complement the extant analytical studies on reverse factoring with empirical evidence and determine whether theoretical predictions of those models are consistent with industry practice. Some of our corresponding hypotheses are indeed supported, whereas in other cases, we find the opposite direction significant. Based on our analysis, we derive several implications for managers and researchers.

We will use data from a structured online discussion forum to understand what the key engagement drivers in online discussions are, and how we can leverage these drivers to improve the operations and performance of online discussion platforms. We will present both empirical and theoretical results characterizing strategies to guide discussions optimally - a crucial feature in an age where communications in both business and educational settings are increasingly moving to online settings.

Online dating apps are the most common way for couples to meet. Many of these dating apps use subscription based pricing (SP), where subscriptions to the app are sold at a fixed price. In online dating (SP) is controversial as it misaligns the incentives of the platform and its users. Another strategy is contract pricing (CP), where the dating app is contracted at a one time price. We study the profit and welfare trade-offs associated with either pricing strategy for online dating platforms.

We study the herding and learning effects on the incentives for online reviews. We model the evolution of sales and reviews for a seller by a generalized Polya urn process and evaluate the profit for three incentive policies: incentivize before purchase, after purchase, or only if they write positive, possibly fake, reviews. We obtain conditions that each type of incentive is profitable and optimal. The results imply that platforms can curb fake reviews if allowing pre-purchase incentives.

We consider a multi-class multi-server queueing system, in which customers of different types have heterogenous preferences over the many servers available. A service provider designs a menu of service classes that balances maximizing the customers’ average service reward and minimizing customers’ average waiting time. Customers act as rational self-interested utility maximizing agents when choosing which service class to join. We study the problem under heavy traffic conditions.

We consider a first-come-first-serve single-server system in which heterogeneous customers. Customers are both payment and lead-time sensitive and are heterogeneous in both immediate service valuation and lead-time sensitivity. The service provider considers incentive-compatible price/lead-time menus based on the system congestion to maximize revenue. The optimal policy suggests the timing to offer a state-independent inflated lead-time option which may be easy to implement in practice.

Recent empirical observations find that the service time at the non-dedicated service provider is significantly prolonged compared with that at the dedicated service provider. This study models these empirical observations by developing stochastic models to assess the impact of the prolonged service time and other parameters (system workload and asymmetry level) on flexibility configurations. We obtain conditions characterizing the optimal flexibility design in the parameter space.

We examine whether mere adoption of fulfillment services offered by

platforms distorts competition by using data from a leading online retailing marketplace to empirically evaluate the effect on upstream

supply echelons. We find that evidence for regulatory views as the surplus welfare is absorbed by the platform. Smaller merchants with lower margin, are forced to increase price to remain profitable with platform fulfillment, leading to a price disadvantage compared to the bigger suppliers.

Capacity-constrained asset providers (APs) often compete over prices when they trade their transport capacities with logistics service providers (LSPs) via spot markets. To circumvent demand uncertainty, an AP and an LSP can negotiate a contract to secure sales and capacity, respectively. We propose a two-stage game theoretical model to study the trade-off of balancing the contract and spot market by characterizing the contracting and pricing strategies under competition and demand uncertainty.

We study the impact of promotion display for online retail platforms where customers search. Utilizing a dataset set which contains detailed behavior information, we estimate a search and purchase model. Accurate estimation also enables us to evaluate different promotion display schemes and design policies that can improve the revenue. Through counterfactual analysis, we demonstrate that our policies can improve the revenue for some product categories by 2-4%.

Nanostore shopkeepers face complex inventory decisions how to manage their limited cash to acquire products from multiple suppliers. We conduct two studies to understand the drivers of shopkeepers’ behavior. In an empirical study, we find nanostore orders to be severely impacted by supplier visit frequency. In a laboratory experiment, we find that shopkeepers diversify their supply by deviating from optimal replenishment decisions.

Millions of nanostores sell basic items to bottom of the pyramid consumers in emerging markets. Their suppliers struggle with high operational costs because of shopkeepers’ cash constraints. We investigate the impact of trade credits on supply chain performance using difference-in-differences with matching. We find that nanostores receiving a credit transact more often, place larger orders and reject less orders. Trade credits create efficiency gains for suppliers that justify their extension.

We model the competition between online grocery retailers leading to extra emissions from home delivery fleets as a war of attrition. By analyzing the equilibrium strategies of retailers, we estimate the emissions attributable to competition. Our numerical study using data from an online grocery retailer in London shows significant potential for CO2 reduction. Our results can inform policies to reduce the negative environmental impacts of competition in the online grocery retailing sector.

To limit food waste, retailers require suppliers to only send products with a remaining shelf life of at least a minimum life on receipt (MLOR). Such agreements may, however, substantially increase waste at suppliers. We analyze two scenarios of (1) coordinating the supply chain (SC) and (2) collaborating on setting the MLOR level. We show coordinating is neither the only nor always the best way to reduce waste. In some cases, just collaboration can be an excellent way to reduce waste in the SC.

A significant amount of fresh produce is wasted in upstream of the food supply chain due to the high cosmetic standard set by retailers. We examine the economic incentives for retailers to adopt such high standards and their impact on food loss. We show how the retailer’s cosmetic standard decision as well as food loss are affected by rejection rate due to high cosmetic standards and consumers’ willing-to-pay for cosmetic-pleasing products.

Grocery retailers target high inventory levels to avoid out-of-stock situations. A side effect thereof is an undesirable customer picking behavior for the freshest or rearmost item. Products with shorter expiration dates remain at the shelf and convert into food waste over time. Prevailing literature related to food waste in retail neglects this impact. Our paper fills this gap by revealing customer picking behavior in retail stores and by connecting this phenomenon to food waste occurrence.

This paper proposes a multiple-choice rank list model to extend the classic discrete choice model by allowing customers to choose multiple distinct alternatives. In addition, we propose a framework to extract customers' preferences from the clickstream data. The corresponding assortment optimization and personalization problems can be solved by mixed-integer linear programs. Numerical experiments based on Cainiao Network showcase the predictive power of the proposed model.

We propose and analyze a sequential design of price experimentation that balances the learning and earning trade-off in revenue management. Assuming the demand function belongs to a parametric family with an unknown parameter value, we derive a closed-form stopping rule based on the observed Fisher information. The decision maker adaptively stops learning and optimizes a price based on the cumulative information and there is no need to find an optimal “fixed” sample size a priori.

A broker matches suppliers with a single buyer in an industrial market by submitting bids to procure goods to the suppliers. After bids are evaluated, the broker learns the quantities procured and ships the goods to the buyer to meet its demand. Modeling the broker’s problem as a new type of newsvendor network problem, we study the effect of problem parameters and uncertainty on the optimal bids as well as conditions under which it is optimal for the broker to bid at multiple supply locations.

Ridesharing platforms have traditionally implemented dynamic walking, which asks passengers to walk a little towards the car in order to achieve more efficient matches. Using novel models and extremely detailed Lyft data, we propose the new paradigm of static walking, which communicates a predetermined pickup location to the rider.

Discovery program is an affirmative action policy used by NYC Department of Education to increase the number of disadvantaged students at specialized high schools. We show that the discovery program suffer many drawbacks both in practice and in theory, and explore possible replacements. We propose a minimal yet powerful modification of the current implementation via the joint seat allocation mechanism, which we show would improve the welfare of disadvantaged students maximally.

In this paper, we study how schools work around insufficient government funding with supplemental resources from nonprofits. We ask: (i) How do resource-supplementing workarounds differ across schools with different socioeconomic advantage? (ii) What policies can ensure workarounds do not exacerbate educational inequities? We answer thee questions by applying Little's Law with validation from 62 interviews from six strategically sampled schools with different levels of socioeconomic advantage.

Smallholder farmers find themselves at the crossroads of the global efforts to reduce worldwide poverty and hunger, as well as the urgent need to prevent deforestation and the associated environmental consequences. In this work, we study how the smallholder farmers' welfare can be improved while preventing deforestation. For this, we propose a detailed operational model of a farmer’s dynamic decisions of land-clearing and production, under liquidity constraints and random cost and yield shocks.

Many multinational buyers of agricultural commodities have made commitments to halt deforestation and improve farmer livelihoods in their supply chains. We propose group incentives conditional on forest protection requirements as a feasible mechanism for achieving this. We develop an analytical model and characterize the cases when group incentives dominate individual incentives, and use data collected from field research in Indonesia to assess the effectiveness of the approach.

Right-to-Repair regulations require producers to supply necessary information and parts for consumers to independently undertake repairs. These regulations aim to prolong product lifetimes through repairs, but they may have adverse consequences such as cloning. This may encourage producers to reconsider their business model choices between ownership and non-ownership models. We analyze the effect of RTR on business model choice, and the implications for producers, consumers, and the environment.

This paper studies a multi-channel healthcare system with both a telemedicine and an in-person channel. Using a DID identification strategy, we find that introducing multiple channels has significant impact on the in-person channel and system level, including an increase in case complexity and planned consultation time by 20% for in-person consultations. In addition, we observe an increase in system capacity, a reduction in wait time for in-person appointments by 37.5%.

In this paper, we provide an empirical investigation into the role of horizontal and vertical integration of orthopaedic surgeons in driving heterogeneity in the impact of the Comprehensive Joint Replacement (CJR) program. Using a difference-in-differences approach, we find hospitals with high horizontal and vertical integration see an increase in both hospital costs and complication rates of 3.17% 1.17, respectively, while others see either a decrease or no change in these measures.

Comorbid depression could lead to a 100% increase in the cost of care for diabetes. Clinical trials have demonstrated that depression care through care managers in a primary care setting (called Collaborative Care) leads to faster depression remission. We present a mathematical modeling approach that determines the optimal allocation of care managers' time to enrolled patients towards improving clinic revenue and patient health outcomes.

In partnership with IU Health, the largest health system in Indiana with 16 hospitals, we jointly developed a suite of advanced data and decision analytics to support a novel internal travel nursing program. This work addresses a long-standing gap in healthcare between state-of-art data decision support analytics and operational processes. Four months after implementation of our integrated machine learning and optimization tool demonstrated 5% lower understaffing and annualized savings of $900K.

Given the increase in the demand for temporary healthcare workers and their additional cost burden, we propose a two-stage stochastic optimization framework to inform recruitment decision making for a provider facing a period of highly uncertain demand. The optimal recruitment decisions are analytically characterized under a general setting. A case study is conducted to illustrate the application of our framework in an inpatient ward. We also show potential savings from adoption of our model.

This paper develops a queueing game model to investigate the impact of virtual triage in the acute care setting. We find that, when virtual triage excessively recommends emergency (primary) care, it could bring about a decrease in ED (GP) visits. Another finding is for arbitrary self-triage accuracy, the adoption of informative virtual triage can worsen system performance. To unlock the operational benefits, we characterize the optimal virtual triage accuracy subjective to the ROC curve.

Agile project management techniques have become commonplace in many organizations. We experimentally examine how these techniques affect performance in two innovation settings: (1) a design setting and (2) a search setting. Our results caution against uniform adoption of the Agile approach, and suggest that the choice of the approach should depend on the nature of the project and on the risk appetite of the project manager.

We study multiple parallel innovation contests where contest organizers elicit solutions to a set of problems from solvers. Prior theoretical work shows that if problem solution requires less novelty then organizers should restrict participation with one contest. We experimentally show that this may not hold true in practice due to the heterogenity of solver efforts and encouraging solvers to participate in multiple contests can yield a larger organizer profit even for less novel problems.

In this article, we investigate how the working capital management practice of a firm impacts its upstream and downstream supply chain partners.

We propose a Cash Conversion Distance (CCD) metric that identifies the degree with which a firm practices aggressive (or lax) working capital management.

We use secondary empirical data to show how firm's profitability as well as that of its partners varies as a function of this measure and its `importance' in the network.

Online platforms screen users who wish to benefit from their marketplaces. We show that when a platform transitions from private to public ownership, it will drop its screening standards, thus admitting otherwise unqualified users. Dropping standards ahead of an initial public offering allows platforms to increase their user base, leading stock investors to overvalue the stock (while imposing a cost on their users). We substantiate this hypothesis with data from p2p lending platforms.

We study advance selling and upgrading in a priority queue setting that emerges in the amusement park industry. Waiting sensitive customers who purchase cheaper advance tickets may suffer from demand uncertainty when consuming the service. Customers can purchase regular tickets in advance and upgrade to fast-track tickets on-site. We find if there are some offline customers who cannot purchase in advance, then allowing upgrading generates more revenue for the service provider.

We consider a system where customers have to go through two service stages. We study the fully-observable model, in which queue-length information of both queues is available at arrival: customers observe the state of the entire system and decide whether to join or not. To learn the value of information we compare the fully observable system with the partially observable system in which, instead of observing the full system state, customers observe queue length only at arrival to each queue.

We study a queueing system with returns where at each service completion epoch, the decision maker can choose to reduce the probability of return for the departing customer at a cost that is convex increasing in the amount of reduction in the return probability. We characterize the structure of optimal long-run average and bias-optimal transient control policies for associated fluid control problems. Our results provide insights on the design of post-service intervention programs.

We study the problem when a firm sets prices for products based on past transaction data. We do not impose a model on the distribution of the customers' valuations and only assumes purchase choices satisfy incentive-compatible constraints. The valuation of each past customer can be encoded as a polyhedral set, and our approach maximizes the worst-case revenue. We show optimal prices in this setting can be approximated by solving a compact mixed-integer linear program.

We study a bundle pricing policy, Component Pricing with Bundle Size Discount (CPBSD). It sells bundles at the sum of component prices minus a discount depending on the bundle size. It subsumes many mechanisms including Component Pricing and Bundle Size Pricing. We show that CPBSD attains the optimal profit asymptotically among all pricing policies under a weak condition. We formulate MILP for the optimal CPBSD. Comprehensive numerical experiments demonstrate the good performance of CPBSD.

We consider a firm that sells some primary and ancillary products (services) to heterogeneous customers. The challenge is to determine the prices for all the products and services simultaneously, to optimize profits to the firm. We consider random utility model for customers' choice problem, and show that the choice model can be reformulated into a perturbed utility model (PUM) over the convex hull of the feasible solutions. Furthermore, we demonstrate how we can obtain a good approximation.

A driver of service value, relevant in practice but understudied in the literature, is the experience of the customer while the service is performed. The design of the service experience is nontrivial given that it needs to tie in pricing and service delivery, and requires carefully designed incentives for the service workers to deliver the experience promised to customers. In this paper we use a novel framework that helps firms understand the essential tradeoffs underlying these choices.

This paper investigates how service evaluations, often collected as star ratings and comments, are biased and how this bias can be mitigated through the ordering of questions. Our findings suggest that writing comments first provides the time and space for the participants to reflect on their entire experience and allows the subsequent star ratings to capture a more holistic assessment of server quality.

Service providers are challenged by the demand to deliver near-customized services without noteworthy wait times, known as the “customization-responsiveness (CR) squeeze”. The approaches so far to manage the CR squeeze are mostly conceptual, or focused on a single dimension. Using choice-based offer selection and queuing theory, we develop a mathematical model to address the CR squeeze from a formalized perspective, and derive managerial insights and recommendations for specific applications.

We study the assortment planning problem for a single product category when a retailer faces multi-item purchasing, so-called “n-pack” consumers as introduced by Fox et al (2017). We obtain interesting properties of the product demand functions and establish that the optimal assortment is a popular set. We evaluate our model on a real-life data set and find that the demand proportions predicted by our model can be made extremely close to the actual proportion of sales.

Determining the assortment and inventory levels based on their shelf life is essential for retailers to maximize profits while avoiding food waste. Assortment and inventory decisions are interrelated by the limited shelf space. A joint approach is needed that defines the assortment size and the maximum possible inventory levels while considering product ages. We develop the first multi-period approach to integrate product shelf life and product outdating.

Motivated by the product ranking in online retail, we introduce and study the "sequential submodular maximization problem". Given an ordered list of products, a user inspects first $k$ products in the list for a $k$ drawn from a given distribution, and decides whether to purchase an item based on a choice model. The goal is to find an ordering maximizing the probability of purchase. We design near-optimal approximation algorithms for this problem, with or without group fairness constraints.

We examine grocery retailer's selling strategies of cosmetically imperfect produce. We consider three strategies: (1) Discarding of the imperfect produce; (2) Differentiating perfect and imperfect produce and selling at different prices; (3) Bunching strategy where perfect and imperfect produce are sold together. We identify when each strategy is optimal under quality uncertainty and varying degrees of consumer valuations and price sensitivity.

Growing consumer demand for sustainable food and transparent supply chains, makes direct-to-consumer sales an attractive alternative for farmers, next to the retail market. We study how a dual-channel structure can improve an agricultural food supply chain’s profitability and sustainable transformation, considering the effect of random crop-yield, customer appreciation of additional producer information, and esthetical requirements of retailers.

We perform a game-theoretic analysis of two approaches to finding adoptive parents for children in foster care. We develop a new search-and-matching model and provide analytical results that suggest several advantages of having children's caseworkers drive the search process rather than prospective parents. Numerical case studies show that caseworker-driven search can result in both reduced search efforts and better matches for children.

The food delivery market migrates to platforms that allow optimizing courier routing by sharing couriers among many restaurants. We address the question: how courier sharing contribute to the reduction of delivery costs? We consider a spatial queuing model in which couriers are servers. We show that in several scenarios dedicated courier system achieves higher profit than a shared courier system. This result can be attributed to the imbalance in the courtier allocation that sharing creates.

Many platforms let guests reserve assets ahead of the rental start time. We find that this feature also allows hosts to decide when to create a listing of their asset, especially given most platforms use a commission contract along with dynamic prices. We use a game-theoretic framework to model such platforms. We theoretically find conditions where the hosts withhold asset availability information and find empirical support for the same. We find the optimal contract for such platforms.

We propose, analyze and solve three decision optimization models for online retailers to make personalized upgraded substitution and co-purchase recommendations. Specifically, we explore: (i) pure expected revenue maximization, (ii) maximization of a weighted average of the expected revenue and the expected consumer surplus, and (iii) maximization of the expected revenue with a constraint on the minimum expected size of the shopping basket.