Online discussion platforms facilitate remote discussions between users. This paper explores the impact of consensus on engagement and proposes algorithms to optimize rankings. Consensus is identified as a crucial engagement driver, and our proposed algorithm outperformed current approaches in an experiment. Promoting debate over echo chambers, consensus is essential for user engagement and platform design.

Retailers use our framework to leverage category sales data for individual product demand prediction. Integrating category-product information, we exploit risk pooling through transfer learning. Our approach combines data from different levels, treating top-level sales as regularization. It outperforms JD.com benchmarks by over 9%, highlighting the value of transfer learning in demand prediction for cost savings in low-margin e-retail.

Human donor milk is vital for preterm infants, but its macronutrient content varies, necessitating pooling. To address resource limitations in milk banks, we propose a data-driven framework using machine learning and optimization. Collaborating with a milk bank, we collect data, fine-tune models, and simulate operational scenarios. Our approach improves macronutrient target achievement by 31-76% and reduces recipe creation time by 67% compared to baselines.

In repeated games, strategies are often evaluated by their ability to guarantee the performance of the single best action that is selected in hindsight. Yet, the efficacy of the single best action as a benchmark is limited, as static actions may perform poorly in common dynamic settings. We propose the notion of dynamic benchmark (DB) consistency and we characterize the possible empirical joint distributions of play that may emerge when all players are relying on DB consistent strategies.

TBD

Split liver transplantation (SLT) can potentially save two lives using one liver. To facilitate increased SLT usage, we formulate a multi-queue fluid model, incorporating size matching specifics, dynamic health conditions, transplant type, and fairness. We find the optimal organ allocation policy, and evaluate its performance versus other common allocations.

We study the deceased-donor liver allocation policies in the United States. In the transplant community, broader organ sharing is believed to mitigate geographic inequity in organ access, and recent policies are moving in that direction in principle. The key message to policymakers is that they should move away from the `one-size-fits-all' approach and focus on matching supply and demand to develop organ allocation policies that score well in terms of efficiency and geographic equity.

Problem: We enhance a hospital's abdominal surgery efficiency using patient-specific information. Our framework addresses heterogeneous surgical volume effects and generates patient-specific data. Regression models, causal forest, and optimization reveal significant effects and reduce surgery duration by 3-18%. This improves efficiency by matching patients to surgeons based on specific volume effects.

The capacity market is a marketplace for trading generation capacity, and is viewed by its proponents as a mechanism to ensure power system reliability. Based on practice at NYISO, we propose optimization models for capacity markets and analyze outcomes. We find that with capacity markets, more generators are profitable, especially the ones with a lower net cost of new entry. Also, it is possible for a generator to earn more by exercising market power. We conduct case studies on NYISO dataset.

The seller wants to sell an item to n i.i.d. buyers and only the support [a,b] are known; a/b quantifies relative support information (RSI). The seller either minimizes worst-case regret or maximizes worst-case approximation ratio. We show that i) with low RSI, second-price auctions (SPA) is optimal; ii) with high RSI, SPA is not optimal, and we introduce a new mechanism, the "pooling auction" (POOL), which is optimal; iii) with moderate RSI, a combination of SPA and POOL is optimal.

We study resale platforms, a growing type of online marketplaces in developing countries. These platforms allow their users to earn profits by selling products to their contacts, who do not typically shop online. We analyze data from a major resale platform in India and leverage our empirical findings to develop a model of the platform. We provide insights into key design aspects of the platform, such as the structure that determines resellers’ margins and the product ranking algorithm.

The sharing economy business model has recently received much attention to determine whether it is economically viable and environmentally friendly. The trade-off is between a decrease in production volume and an increase in product usage. We investigate both economic and environmental impacts of the sharing economy business model for comparison with the traditional models of pure sales and servicizing. Sharing economy can yield a win-win outcome as to profitability and environmental impact.

Our paper investigates the effectiveness of IoT-based nudging on energy savings. The results of our study highlight the value of using IoT technology to influence energy consumption behavior and reduce energy usage.

Drug-drug interactions (DDI) are common medical errors due to the negative interaction of medicines taken by a patient. up to 96% instances, doctors ignore the even the mandatory DDI alerts, leading to adverse reactions and poor health outcomes. To solve this issue, we collaborated with a large health-tech firm in India, which conducted a field experiment to identify the impact of real-time information nudges on DDIs. We find that nudges not only reduce DDIs by 8.7% but also promotes learning.

Due to social distancing requirements during the Covid-19 pandemic, the capacity of elevators in high-rise buildings has been reduced by 50-70%. This reduction in capacity has led to queue build-ups, making it difficult to maintain social distancing in the lobby and increasing the risk of disease transmission.

The objective of this work is to minimize elevator wait times in a multi-clinic facility by optimizing the clinic scheduling.

Should firms differentiate quality to soften competition, or equalize it by sourcing? We study sourcing by inefficient firm 2 from efficient firm 1. We show firm 2 prefers sourcing, if it well improves quality. Otherwise no sourcing even for free. Firm 1 averts sourcing if it has large quality leadership under no-sourcing and low horizontal differentiation. While sourcing pools resource to firm 1, it may yield lower quality than no-sourcing for consumers if firm 1 has small efficiency advantage.

The growing complexity of consulting, new product development, and information technology projects has led firms to increasingly adopt the strategy of collaborative value cocreation with their vendors. Research in this field overlooks value cocreation that often involves a client firm engaging with more than one vendor. We bridge this gap by studying a cocreation environment involving a client, a primary vendor, and a potential secondary vendor.

Firms rely on suppliers for innovation, often through procurement or innovation contests. Supplier ownership and control impact contest outcomes. We analyze the effects on buyer's profits and product innovation, identifying when they align or diverge. Our findings inform optimal supplier base structure and explain recent industry developments. Highly profitable structures may hinder innovation, prompting firms to prioritize long-term innovativeness over short-term profits.

Optimal stopping is the problem of determining when to stop a stochastic system in order to maximize reward. We propose a methodology for optimal stopping based on randomized linear policies, which choose to stop with a probability that is determined by a weighted sum of basis functions. We develop a practical heuristic for solving our randomized policy problem, and numerically show that our approach can substantially outperform state-of-the-art methods.

​​​​A multinational risk averse newsvendor produces goods at home (domestically) and sells both overseas and at home, over multiple periods. ​​We consider ​risks due to uncertain exchange rate as well as uncertain demand and investigate the effectiveness of ​(a) general financial hedging contracts​ and (b) operational hedging, which is to allow production both domestically and overseas. ​W​e evaluate both types of hedging and describe the situations that favor each type of hedging.

Firms can adjust operations based on financial asset price impact on product demand. We develop a model integrating financial risk hedging into pricing decisions. Hedging generally lowers optimal price and service level. The impact of asset price on demand determines the effect on service level. Our model reduces risk without significantly reducing operational profit. Including operational payoff functions reveals when hedging reduces optimal operational levels.

Consider an omni-channel retail system that consists of an online sales website and a physical store. This study examines the assortment planning problem in the physical store under decentralized and centralized decision structures, while also allowing for product returns. We use combinatorial optimization and game theory techniques to find the optimal assortments. Moreover, we propose economies/diseconomies of scope (EOS/DEOS) contracts that are capable of fully coordinating the retail system.

Problem: Separate product ranking and inventory ordering decisions in e-commerce lead to inefficiencies. We propose a joint approach to optimize them. Methodology/Results: Our algorithms improve profitability by coordinating inventory and ranking decisions. They provide practical solutions for a large number of products. Managerial Implications: Coordinated decisions boost profits in e-commerce.

TBD

We present a collaboration with the OCP Group, one of the world's largest producers of phosphate, in support of a green initiative to reduce OCP's carbon emissions significantly. We study the problem of decarbonizing OCP’s electricity supply by installing a mixture of solar panels and batteries to minimize its investment cost plus the cost of satisfying its remaining demand via the Moroccan national grid. This forms the basis for a one billion USD investment in renewable energy generation.

Problem: Evaluating an ED delay announcement system in Hong Kong's healthcare. Methodology: Studying 1.3M patient visits, we estimate patient sensitivity to announced waiting times (WT) and factors influencing it. Results show potential improvements in reducing WT and patients leaving without being seen. Implications: Increase patient awareness, reduce WT update window, focus on older population and Kowloon district for promotion.

Online agri-platforms, like Karnataka's United Market Platform (UMP), aim to improve smallholder farmers' welfare. This study designs a two-stage auction on the UMP, increasing farmers' revenue with over $6 million in commodity trades. Results show a 3.6% price increase (55%-94% profit gain), benefiting 10,000+ farmers. Lessons highlight innovative price discovery in resource-constrained environments, emphasizing operational and behavioral factors for success.

We present a framework for optimizing the selection of which content to promote on digital distribution channels, consisting of: (1) A model of the decision-problem, (2) An off-policy identification and evaluation method, (3) the design of ranking policies, and (4) a causal machine learning procedure. We partner with an international newspaper and show that our optimal policy improves the newspapers outcome by 18 percent. Our work contributes by supporting the curation of digital channels.

Tree-based models have interpretability but are not able to capture complex relationships while other ML models often lack interpretability. We propose neural-informed decision trees (NIDTs) to combine predictive power with interpretability. We evaluate NIDTs on over 20 UCI datasets and show they outperform multiple ML benchmarks. We demonstrate interpretability by extracting an explainable warfarin prescription policy and show how they can be used on a pricing problem.

Platforms run many A/B tests, but testing all combinations is impractical. DeDL, our debiased deep learning framework, estimates causal effects and identifies optimal treatments. It combines deep learning with double machine learning, yielding consistent estimators. Results on a video-sharing platform show accurate estimation and efficient iteration. DeDL enables platforms to iterate operations effectively.

Physicians spend more than 5 hours a day on EHR and 1 hour after work, causing burnout, attrition, and appointment delays. This paper examines the impact of workflow decisions on EHR time. Using 150,000 appointments from 74 family medicine physicians, we find pre-appointment EHR reduces workload and after-work hours, while post-appointment EHR decreases after-work hours but increases total EHR time. Our findings help healthcare administrators design EHR workflows to reduce physician burnout.

We study how task switching in emergency departments (ED) impacts physician efficiency, quality of care, and patient routing. We find task switching hurts physician productivity, while its influence on quality is insignificant. ED physicians are switch-averse when routing patients. Leveraging the heterogeneity among task switches, we propose an implementable data-driven queue management method to partition patients into two queues. The simulation shows our method effectively improves efficiency.

This study investigates the impact of telehealth expansion on rural-urban healthcare access disparities during COVID-19. The findings reveal an increased gap between rural and urban areas, with telehealth contributing to a 3.9% rise in the disparity. Urban patients adopt telehealth more, while rural patients rely on in-person visits. The research highlights the need to address barriers and promote equitable access to remote care, informing policymakers, healthcare providers, and researchers.

In traditional assortment optimization, profits may be improved by making some products unavailable as part of the lost demand flows to higher margin products. The refined assortment optimization problem takes a subtler approach to improving profits without necessarily making some unavailable. This is done by judiciously making some products less appealing thereby steering demand to more profitable products. We develop a theoretical framework and provide tight revenue bounds.

We study efficient real-time data collection approaches for an online retailer that dynamically personalizes the assortment offerings based on customers’ attributes to learn their preferences and maximize revenue. We study the structure of efficient exploration in this setting, prove a performance lower bound, and propose efficient learning policies. We illustrate the superior performance of our policies over existing benchmark using a dataset from a large Chilean retailer.

We investigate assortment problem under distributionally robust and robust satisficing formulations. We generalize the optimality of revenue-ordered set for both distributionally robust assortment and robust assortment satisficing problems. We further discuss the choice model based on worst-case distribution and derive managerial insights by theoretical analysis. Moreover, we discuss the two robust formulations of assortment problem with cardinality constraint and develop efficient algorithms.

Crowdsourced delivery platforms offer workers complete flexibility in scheduling their own hours. However, since workers are treated as independent contractors, they do not receive minimum wage protection. Here we examine the integration of driver compensation guarantees in a platform's matching decisions. We design dynamic matching policies that guarantee a particular level of utilization or wage for active workers, while maintaining the inherent work hour flexibility of the sharing economy.

TBD

TBD

In team settings, it is critical to mitigate moral hazard, and induce reporting of any shirking. However, agents can free-ride, which exacerbates moral hazard; and deep-rooted loyalty considerations prevent peer reporting. We study peer reporting in a team of one principal and two agents. One of the agents may be loyal: doesn’t shirk and doesn’t report shirking. We show that even a small probability of such a loyal agent unravels reporting.

The flow of humanitarian aid as a response to the Russian Federation’s full-scale invasion of Ukraine was unprecedented. In this paper, we analyze the pressing issues of how to regulate the delivery of humanitarian aid to the final beneficiaries, preventing its loss and misuse due to corrupt behavior in the delivery process.

Girl Scouts' Cookie Program aims to develop girls' skills, but retaining their engagement is crucial. We analyze troop-based factors influencing engagement retention.

Small troop size, balanced sales performance, high adult-to-girl ratio, scout level heterogeneity, and troop-based booth sales positively impact retention.

NPOs can enhance engagement by focusing on team composition, relationship-building, and understanding the factors driving retention in volunteer and fundraising activities.

Research suggests that the regulatory inspection outcomes are influenced by factors unrelated to non-conformance. One of the reasons is shift in the U.S. political geography every two years which exposes firms to differential policy risks. We use political alignment to empirically uncover the impact of time-varying policy risks on quality performance (recalls) of food operations via regulatory inspection outcomes. We identify firm-level strategies to mitigate it and make policy recommendations.

In a contest, participants exert effort to win the a prize. The cost of their effort is irreversible. Most of the research on contests has treated a contest as an isolated event. However, in many real-life situations, we choose among several contests that are going on in parallel. We combine theoretical and empirical question to address the following questions: How does an individual choose among contests and how does the existence of multiple alternative contests affect her behavior?

New apparel brands have emerged that offer sustainable fashion at affordable prices. Designing an effective CSR disclosure strategy for affordable apparel presents new challenges, as the literature has shown that consumers often equate low prices with low quality. We use an experiment in an online purchase context to examine the mechanisms behind consumers' valuations of such CSR disclosures and how price-related factors (retail price, historical price paid) moderate these indirect effects.

We study a new business practice of “crowdsourcing buying” wherein a retailer first seeks input from customers on the desirability of a product and then bases their purchasing decision on their votes. We collaborate with an apparel rental platform and analyze a dataset comprised of the platform’s products and users. We find that the initiation of the service improves both conversion rate and the number of turns. We also investigate several potential mechanisms that drive this improvement.

We show that WCs can improve their resilience to digital competition by making strategic retail location decisions. We find that ecommerce intensity in a geographic market negatively impacts Costco foot traffic in the same market. However, the location matters. Specifically, locating WCs in a “low agglomeration” area, distant from competing WCs and grocery stores, not only contributes to increased foot traffic but also has the potential to offset the negative impact from ecommerce competition.

Loot boxes are a primary source of revenue in video game industry. Loot boxes randomly drop items of differing value. To design a loot box, sellers choose the purchase price and drop rates. We show that, in general, the loot box design problem is NP-hard. By restricting the form of player utilities, we can solve the problem exactly in polynomial time when the number of items is fixed. Under different restrictions, we solve the problem approximately in polynomial time with fixed precision.

The degree of modularity embodied by a system overall architecture has clearly received quite some attention in the product/service literature but there has been much less theory established to specifically guide the incentive design for the development of a modular system within an organization. We seek to understand how the product architecture and the sources of uncertainty that affect the value interdependences of such a system, influence the optimal design of individual and team incentives.

In innovation contests, organizers seek solutions from solvers individually or as teams. High-novelty, nondecomposable problems benefit organizers with team submissions despite lower solver efforts. For low-novelty, nondecomposable problems, teams may not benefit organizers. Decomposable problems have specific conditions for organizer benefits. Contests favor team collaboration over in-house innovation. Solvers benefit from teams when novelty is moderate.

We present ways to improve weather-based agricultural insurance for smallholder farmers in the developing world, where post-season inspection of crop outcomes is not feasible. We seek policies that better align insurance payments with adverse weather-related yield reductions to help stabilize the farmer’s income over time. To illustrate the ideas and results, we utilize parameters for crops that are commonly grown in India.

Health on Loan: The Effect of Local Credit Availability on Hospital (Re)Admissions

Some life-improving technologies for the poor are unaffordable to them, as their limited liquidity puts the purchase costs out of reach. Thus, business models have emerged where the consumers pay a fraction of the price upfront to acquire the technology and make a series of payments for continued access. Using the optimal contracting approach, we investigate payment mechanisms that balance flexibility, discipline, and ownership incentives. Several implementable features emerge from our analysis.

The current business practice and existing academic literature estimate customers’ lifetime value (CLV) for the company, considering customers’ frequency, recency, and monetary value (RFM). In this paper, we uncover the limitations of this approach and propose the Operational Customer Lifetime Value (OCLV) as a more comprehensive way to think of the long-term value of a customer for the company.

In this paper, we study a retailer's returnless refund (RR) policy, under which the retailer may issue a refund without asking the buyer to return the product. We develop parsimonious models to formulate the problem. We show that an early commitment to RR improves the retailer's profit, but only when the net salvage value is small or moderate. We also consider customers who may abuse this policy and show that surprisingly, their existence may increase profit.

Solar power integration into the grid is important for combating climate change, but variability in solar irradiance poses a challenge. We introduce CrossViViT, a deep learning model that utilizes satellite data to accurately forecast Global Horizontal Irradiance (GHI) for the next day. Our model also provides prediction intervals for each time-step, indicating the level of forecasting uncertainty. CrossViViT performs well in solar irradiance forecasting, even in unobserved solar stations.

We present a collaboration with the OCP Group, one of the world's largest producers of phosphate, in support of a green initiative to reduce OCP's carbon emissions significantly. We study the problem of decarbonizing OCP’s electricity supply by installing a mixture of solar panels and batteries to minimize its investment cost plus the cost of satisfying its remaining demand via the Moroccan national grid. This forms the basis for a one billion USD investment in renewable energy generation.

In rural areas of Sub-Saharan Africa, people’s main source of clean drinking water comes from handpumps. However, across SSA, one in four handpumps are broken at any given moment. In this project, we study the maintenance program of three NGOs. We develop a MDP model to study the optimal schedule of mechanics visiting water points under different information availability and logistics structure. The goal is to minimize water point downtime.

Deep learning systems use huge training data sets to meet desired performances, but over/under-collecting training data can incur unnecessary costs and workflow delays. We propose and then solve an optimal data collection problem incorporating performance targets, collection costs, a time horizon, and penalties. Experiments on six deep learning tasks show that we reduce the risks of failing to meet performance targets by over 2x compared to existing estimation-based heuristics.

Learning optimal policies from historical data enables personalization in many applications. Adaptive data collection is becoming more common for allowing to improve inferential efficiency and optimize operational performance, but adaptivity complicates policy learning ex post. Our work complements the literature by learning policies with adaptively collected data. We propose an algorithm with proven finite-sample regret bound, which is minimax optimal and meets our established lower bound.

We study the performance implications of quality and quantity of data in contextual decision-making. We focus on the Newsvendor loss and consider a data-driven model in which outcomes observed in similar contexts have similar distributions. We characterize exactly the worst-case regret of a classical class of kernel policies. Our exact analysis unveils new structural insights on the learning behavior of these policies that cannot be observed through state-of-the-art general purpose bounds.

Current trends suggest telemedicine will continue to play a key role in post-pandemic care delivery. Some empirical studies, however, observed that adopting telemedicine can trigger more demand for in-person visits and overcrowd the clinic. We develop a queueing game model to assess the impact of telemedicine in equilibrium. Analyzing this model allows us to characterize the optimal resource allocation for outpatient clinics and the conditions under which introducing telemedicine is beneficial.

We describe a novel approach to automatically generating data-driven simulation models from event log data by combining process mining, queue mining, and machine learning techniques. The resulting model can be used for mapping and improving the process. We describe a healthcare application of this technique where the focus is on estimating direct and indirect effects of congestion. We also discuss how to overcome a challenge posed by very scarce event log.

Problem: Impact of telehealth on patient behaviors (no-shows, unpunctuality) unclear. Analysis of 280,067 appointments shows telehealth reduces no-shows by 4.0% and late arrivals by 10.2%. Adherence improves for follow-up patients (convenience) and new patients (timely access). Improved adherence enhances throughput, efficiency, and access to care. Telehealth implementation justified for increased revenue.

We consider an assortment optimization problem where the customer may purchase multiple products and possibly more than one unit of each product. We adopt the customer consumption model based on the multiple-discrete-choice (MDC) model. We present an algorithmic framework that delivers near-optimal algorithms for different variations of the assortment problem.

Essentially every choice involves an information collection or search phase prior to a decision-making phase. To study how these phases interact, we embed the Exponomial Choice model (Alptekinoglu and Semple 2016) in a classical model of sequential search with perfect recall (Weitzman 1979). We derive the search path and final choice probabilities in closed form and develop all the analytical tools to enable joint optimization of assortment and prices efficiently.

Online discussion platforms facilitate remote discussions between users. This paper explores the impact of consensus on engagement and proposes algorithms to optimize rankings. Consensus is identified as a crucial engagement driver, and our proposed algorithm outperformed current approaches in an experiment. Promoting debate over echo chambers, consensus is essential for user engagement and platform design.

We consider the firm maximizes the total expected revenue over a finite time horizon by optimizing the assortment/pricing of each time period.

Customers make choices under MNL with popularity effect, which also considers the historical sales.

The optimal prices can be solved by concave programming.

The heuristic algorithms we propose for assortment optimization have a 1/T performance ratio in the general case, and the ratio improves to 1/ln(T) when the product's utility is constant over time.

TBD

TBD

Managed charging (MH) programs allow utilities to centrally manage electric vehicle (EV) drivers' home charging to reduce cost, avoid new and aggravated peaks and blackouts, and ensure grid stability. In this paper, we present a program design model and a load management model for jointly designing and executing various MH programs.

Do energy price surcharges encourage energy efficiency investments without harming domestic manufacturing?

We analyze a supply chain model and find that energy price surcharges can have varying impacts on domestic sourcing. Supplemental policies like surcharge reduction and EE investment subsidies can help mitigate potential harm.

Policymakers should consider these findings when designing energy policies to promote energy efficiency while minimizing adverse effects on domestic manufacturing.

This study explores multi-agent data-driven decision-making in a supply chain with vertical competition. We find that the sample average approximation and kernel optimization approaches can tacitly collude, reduce double marginalization, improve profits for all participants while protecting consumer welfare. Besides, non-coordinating contracts can even reduce more double marginalization than coordinating ones in the data-driven setting, especially when the feature distribution exhibits skewness.

We propose a novel bimodal inventory optimization (BIO) model and algorithm to position inventory across a retail chain to meet time-varying omnichannel demand. While prior Robust optimization (RO) models emphasize the downside, i.e., worst-case adversarial demand, BIO also considers the upside to remain resilient like RO while also reaping the rewards of potential positive outcomes. Experiments project a profitability gain of 15% for BIO over RO on real-life data from a major retail chain.

The automotive sales has focused on selling an experience or a lifestyle via features. Features are optional add-ons that increase product attractiveness (from stereo audio to cruise control) and basis for purchasing decision. While previous studies conducted car-level analysis on sales, the effect of features has remained unexplored. We examine features' collective impact on sales time along new and used-car segments, propose a two-stage estimation framework, and suggest policy recommendations.

We analyze food purchasing patterns among low-income individuals in India using scanner data. Heavy consumption of packaged junk food is observed. A subsidy program is implemented, resulting in decreased snack purchases and increased purchases of complementary foods. Working parents show the strongest effects. No negative impact on nutrient purchases is found. Customized subsidy programs show a tradeoff between nutrient richness and customer appeal.

We study how payment delays affect project delays. We develop theories linking the two, and empirically test our hypotheses using the QuickPay reform, which expedited payments to certain projects, as an exogenous shock. Surprisingly, faster payments led to greater project delays. We identify contractor liquidity constraints and aggressive bidding as causes for this increase. We also reveal a spillover effect: some projects are accelerated although their payment timing is unaffected by QuickPay.

Valuation of target firms in mergers and acquisitions (M&As) has far-reaching implications for shareholder wealth. This study investigates the effect of a target’s operations capability on the valuation premiums above its market value. We find that target firms’ operations capability has a positive effect on their M&A valuation premiums. However, this effect becomes weaker with increasing market overlap between the target and acquiring firms.

We study dual sourcing under non-stationary demand. The actual demand distribution is not observable, yet demand observations reveal partial information about it. We propose a novel policy that combines a base order on the slow source with flexible orders from the fast and slow source. We formulate the problem as a partially observable Markov decision process and prove the optimal policy structure. Our results show how partial demand information and flexible slow source orders may reduce costs.

We partner with a major online retailer to optimize congestion-aware task assignment in robotic warehousing. We develop an original integer optimization formulation using a time-space network representation of fulfillment center operations. To solve it, we develop a machine-learning-guided large-scale neighborhood search algorithm to iteratively construct and re-optimize small subproblems. We demonstrate the benefits of our model and algorithm as compared to baseline algorithms and heuristics.

We study the structural properties of the optimal ordering policy for perishable products by considering demand forecasts in the inventory model. The optimal ordering policy is a state-dependent base-stock policy that depends on the inventory level, current and previous forecasts, and previous demands. We also propose a linear heuristic that integrates demand forecasts in the ordering policy. Despite the simplicity of the heuristic, its performance is comparable to that of the optimal policy.

We study tactical decision-making for a last-mile e-retail system in which same-day delivery (SDD) and next-day delivery (NDD) orders are fulfilled by the same fleet. We build a continuous approximation model of the system. Unlike prior work using similar methods, we explicitly consider customers' price sensitivity and the relationship between consecutive days in the system. Using this model, we analyze a customer satisfaction objective, then maximize revenue via a differentiated pricing scheme.

Ultrafast grocery deliveries are struggling to become economically sustainable. One way to do so is to optimize the committed delivery time. We show that the optimal delivery radius might be different for different locations. Our results highlight while retailers commit to an identical delivery time, this policy harms their economic sustainability. Our study provides an extensive understanding of the influential factors on profitability and identifies the bottlenecks to improve operations.

E-commerce focuses on improving shipping experience. Pick-up stations boost sales by 3.9%, driven by logistic flexibility, not shipping speed. Consumer choice model shows value in time (76.2%) and choice (23.8%) flexibility. Fewer pick-up stations achieve sales lift. A new strategy without pick-up stations improves sales by 8.4%. Counterfactual logistic strategies increase consumer welfare by 2.0%-10.0%.

Counter-trafficking efforts are often conducted with a local scope. However, sex traffickers commonly move individuals geographically. Thus, there is a need for organizations in different geographical regions to collaborate in counter-trafficking efforts. This research leverages a large dataset of commercial sex ads and techniques from the areas of machine learning and network science to identify prominent geographical circuits and individuals operating on these circuits.

Community corrections provide alternatives for incarcerations, which can reduce jail overcrowding and recidivism rate, particularly for individuals with substance use disorder. We study the placement decisions for community corrections and relevant capacity planning via an MDP model and prove structural properties for policy insights. To address the complex dependence between optimal placement and system congestion, we leverage a two-timescale approach to develop algorithmic solutions.

This paper presents a data-driven approach to mitigate industrial plant air pollution on nearby cities by linking operational decisions with weather conditions. Our method combines predictive and prescriptive machine learning models to forecast short-term wind patterns and recommend production adjustments. Implemented at a chemical plant in Morocco, our algorithm improves weather forecasts by 40-50% and offers valuable trade-offs, reducing emissions by 33% and costs by 63%.

While the cost-saving benefit of generic drugs is obvious, the treatment effectiveness remains unclear. We examine the effect of generic drug usage on health outcomes and address the potential endogeneity concern using instrumental variables and a difference-in-differences framework. We find that generic drug usage leads to higher healthcare service utilization and worse clinical outcomes. Moreover, Our findings highlight the effectiveness heterogeneity of generics from different manufacturers.

Problem: We examine telemedicine's impact on antibiotic prescription errors using patient-provider encounter data. Two-stage least squares regressions show lower overall errors with telemedicine. Effects vary by provider's patient volume and relationship. Reduction in errors mainly stems from type I errors, without compromising patient health outcomes. Telemedicine reduces drug waste and antibiotic resistance, benefiting decision-makers and patients.

The use of telemedicine has grown rapidly. To understand patient behaviors in telemedicine and in-person visits, we studied service incompletion. Physician availability affects in-person visits but not no-shows. Using a multivariate probit model, we found that intra-day delay increases telemedicine service incompletion by 7.40% but has no significant effect on in-person visits. Differentiating incompletions from no-shows is crucial for optimal patient sequencing decisions.

In this work, we study the product replacement problem. This problem arises in the fast fashion industry, where the company's headquarters frequently send new items to stores, expecting store managers to remove current items to make room for the new items and display them. Our study addresses a general case, including misaligned incentives among decision-makers and cross-item effects. We present optimal strategies for the headquarters, but also investigate different store managers' behaviors.

We study the impact of wait time information distortion on customer join decision and the service provider’s revenue in virtual queues. We show that the optimal revenue is first increasing and then decreasing in the distortion level. A zero distortion is optimal only when the service capacity is sufficient and the service value is high. Customer joining rate could be either increasing or decreasing in the distortion level.

TBD

Limited resources in the judicial system can lead to costly delays and even failure to deliver justice. Using the Supreme Court of India as an exemplar for such resource-constrained settings, we apply ideas from service operations to study delay. Court dynamics constitute a case-management queue which is known to be intractable. Hence, we employ data-driven simulations and find that even small interventions can improve the system performance dramatically.

We develop asymptotically optimal policies to track queue lengths under different cost structures in a setting with inaccurate arrival and departure sensor data. We propose an idleness detection policy and explore the value of queue inspections. Our model is motivated by queue tracking implemented at a large airport.

TBD

Efficiently managing inventories requires an accurate estimation of stockout costs. This estimation is complicated by challenges in determining how to compensate consumers monetarily to ensure they will maintain the same level of utility they would have obtained had stockouts not occurred. This paper presents an analysis of these compensation costs, as applied to the design of optimal stockout rates by an online retailer marketing to consumers aesthetically substandard fruits and vegetables.

In this study, we evaluate the effectiveness of AI-based systems in reducing the food waste generated in commercial kitchens. Using a multi-year field experiment in 500+ commercial kitchens, we estimate the waste reductions due to the use of (1) an automated transaction-level scaling system, and (2) an automated weighing and computer vision based waste-classification system. We also explore methods that use the waste data to identify the root causes behind the sub-optimal levels of food waste.

Relay logistics optimize long-haul shipments through a pit-stop network and separate drivers. This model benefits drivers and the environment, while reducing costs and lead times. An efficient algorithm solves the Relay Pickup-and-Delivery Problem, generating multiple arcs for optimal routing. Real-world experiments confirm its superiority over benchmarks, resulting in fewer miles traveled and win-win outcomes for sustainability.

Freight-sharing platforms, where shippers can offer excess transportation capacity at a discounted cost, enable the effective sharing of shipping capacity. This occasional offer of excess transportation capacity introduces new opportunities in inventory management. We study inventory decisions for a company that interacts with a freight-sharing platform. We derive optimal ordering policy characteristics and propose replenishment heuristics.

In this paper, we formulate an optimal information gathering strategy (IGS) to identify which uncertainties in the supply chain network drive our decisions. Existing approaches consider uncertainties, but do not consider the benefit of resolving them. Based on stylized, numerical, and case study results, we show a significant value of optimal IGS. As a non-monotone non-submodular minimization problem, we solve the problem with an algorithm which achieves a constant approximation guarantee.

Leading thinkers in supply chain management have proposed the long-term effects of supply chain disruptions on market share as customers shift their purchases to competitors. Motivated by this insight, we empirically analyze the lasting effects of supply chain disruptions on firms' market shares. Focusing on pharmaceutical supply chain disruptions, we find that products do not fully recover from the market share loss even after they recover from supply chain disruptions.

This paper investigates the impact of advance commitment on supply chain network design under demand uncertainty. The study develops a robust cooperative model and solves it using a column-and-constraint generation algorithm, finding that this approach significantly reduces conservatism and improves performance compared to a non-cooperative model. The research highlights the potential benefits of contractual commitment for companies seeking to enhance their supply chain operations.

This study examines the relationship between the digital divide and online retail sales, encompassing internet infrastructure and socioeconomic inequalities. We demonstrate how retailers can better bridge the divide by leveraging omnichannel and online fulfillment options. Our research emphasizes the significance of incorporating the digital divide into ecommerce fulfillment strategies. We make a contribution to the literature on diversity, equity, and inclusion in operations management.

We propose a data-driven platform that provides traceability to the first mile of agricultural supply chains by coordinating the transactions of farmers and intermediaries. We model unique aspects of the supply chain, including pre-existing informal relationships between farmers and intermediaries, and we develop algorithms to solve real-world instances. We test the results on data from the palm oil supply chain and show the platform’s potential to reduce costs and increase farmers’ welfare.

Interference poses challenges in panel experiments. Aggregating units into clusters is common, but optimal aggregation level is unclear. We propose a randomized design for grid-based units. Our design features randomized spatial clustering and balanced temporal randomization. Theoretical performance, inferential techniques, and simulations validate its superiority.

Challenge: Estimating long-term treatment effects in early-stage experiments is costly. Methodology: We propose a surrogate model using short-term data and historical observations. Results: Verified on WeChat, our method effectively estimates long-term treatment effects. Implications: Our approach reduces experiment duration and provides efficient empirical estimation of long-term effects.

Problem: Content platforms lack effective promotion policies utilizing the diffusion effect. Methodology: We propose a diffusion model, formulate the optimization problem, and introduce D-OLS estimators. Results: We prove submodularity and achieve a 1-1/e-approximation solution. D-OLS estimators are consistent and efficient. Our model improves adoption by 22.48% compared to existing policies. Implications: Our diffusion model enhances content promotion for online platforms.

Innovation projects typically integrate several distinct components which can be completed either iteratively or sequentially. In our experimental study we examine the behavioral underpinnings of iterative vs. sequential workflow in two innovation tasks. Our results show that the iterative workflow significantly outperforms the sequential workflow in a design task, and vice versa in a search task. This cautions against the uniform adoption of iterative workflow methods such as Lean and Agile.

The high patient volumes and acuity levels placed extraordinary stress on the nursing workforce. Increased nursing workload is linked to nurse burnout and patient safety concerns. In this work, we take an empirical approach to understanding the effect of historical workload on nurses’ perceived workload. Quantifying this temporal effect of nursing workload allows us to design patient-to-nurse assignment policies that achieve a more balanced workload and create a fairer and safer working environment.

Can past helping behaviors improve shift scheduling? We study organizational citizenship behavior (OCB) and its impact on shift performance. Past helping predicts patient length of stay better than team familiarity. Small shift composition changes reduce length of stay significantly. Scheduling based on past helping holds promise beyond ICU nursing.

Workers in developing economies may be forced to work excessive overtime, which not only causes severe mental and physical issues to the workers but also results in significant brand damage to the buyers if exposed in public. In this paper, we develop a game-theoretic model of a dyadic supply chain and analyze the buyer’s strategies to combat such excessive overtime issues of the supplier, including auditing the supplier’s practice and conducting supplier development activities.

Problem: Studying disruption impact on centralized and decentralized supply chains. Methodology/Results: Analyzing three-tier supply chain with disruption and network incompleteness. Centralized supply chain maximizes profit. Managerial Implications: Complete network favors centralized supply chain for higher profit and reliability. Incomplete network amplifies disruption's negative impact, reducing profit by up to 20%/30%.

In this paper, we examine the value of vertical integration as a disruption risk mitigation strategy. Vertical integration increases firms' control over their supply chain and could reduce the risk of production being disrupted due to material or component shortages. Using a stylized model of a manufacturer sourcing a critical component from a multi-tier supply chain, we show that vertical integration is most valuable for moderate probability, severe disruptions that are positively correlated.

Companies operating under the gig-worker model are offering full-time jobs to ensure better performance resulting from increased work availability. Using panel data from an online grocery platform offering flexible workhours, we explore if voluntary availability affects performance. We find that voluntary work positively affects productivity and service quality, controlling for worker and task-specific characteristics, and order batching, task complexity and discretion moderate the main effect.

Communication and customer interaction design have been used as elements to improve customer satisfaction and purchasing behavior, but little is known regarding their use as levers to improve operational efficiency. We show the relevance of communication channels as levers of operational performance and provide evidence that these channels can be used to manage customers’ expectations and improve operational performance.

The impact of formal incentives on team productivity is not fully understood and only a handful of field experiments have documented either a small or a null impact. In this study, we empirically explore their effect on individuals and teams. We calibrate a theoretical model to show that the null effect observed in teams is not due to formal incentives losing their effectiveness but because weak formal incentives display larger social incentives than teams with strong formal incentives.

TBD

TBD

We consider the dynamic scheduling problem of a classical single-server multi-class queueing system where the system manager does not have full knowledge about the cost/reward of the customers. One of the key results is that the Whittle index policy is optimal for a two-class queue if the system manager knows the distribution of the reward of one class and dynamically learns the distribution parameters of the other class.

Sepsis is a life-threatening response to infection that leads to organ failure, tissue damage, and oftentimes, death. Our work leverages historical health data in order to learn treatment strategies for sepsis that result in improved patient outcomes under limited resources. We learn an interpretable, Markov Decision Process (MDP) model of the system, formulate a robust value iteration algorithm, and solve the problem of limited resource allocation for optimal sepsis treatment.

Problem: ED overcrowding disrupts public safety net. Impact of increasing ED entry fee is unknown.

Methodology/Results: In Hong Kong, increasing the ED fee reduced traffic by 6.3%, targeting less-urgent visits. Frequent visitors decreased their visits, reducing patient abandonment.

Managerial implications: Financial access hurdles alleviate healthcare congestion without discouraging urgent visits. Managing external demand complements internal process improvements in ED management.

In many service systems, service times of customers can be correlated with waiting times. Scheduling under such dependency is challenging as a Markovian state description requires keeping track of all customers' waiting history. We propose an approximate dynamic programming algorithm for multi-class scheduling with wait-dependent service times. Our algorithm can generate policies with simple structures and achieve strong performance which we illustrate in a healthcare setting using real data.

We study stochastic online resource allocation: a decision maker needs to allocate limited resources to i.i.d. sequential requests generated from an unknown distribution in order to maximize reward. Online resource allocation has been studied extensively in the past, but prior results crucially rely on the assumption that the total number of requests (the horizon) is known to the decision maker in advance. In this work, we develop online algorithms that are robust to horizon uncertainty.

We consider an online advance reservation problem under customer heterogeneity. The decision maker decides on admitting a customer based on the customer's class and the current system information. Our problem model captures applications in medical appointment and hotel room booking, where a resources are to be scheduled for customers' uses in advance. We derive a novel rejection price based algorithm, and we quantify the theoretical guarantee of our algorithm in terms of its competitive ratio.

TBD

We introduce the problem of finding a student-optimal stable matching under adaptive priorities, i.e., when priorities depend on the assignment of other agents. We show that the problem is NP-hard, provide math-programming formulations for the problem, and introduce several heuristics to preprocess the instances and solve them. Finally, using both synthetic and real data from Chile, we show that clearinghouses can significantly improve students' welfare when considering dynamic priorities.

TBD

This research optimizes driver shifts and itineraries for paratransit operators, considering uncertainties like cancellations and no-shows. The SIPPAR model, using a shareability network representation and a two-stage stochastic optimization approach, reduces operating costs and improves robustness. The algorithm outperforms benchmarks in real-world instances, providing faster computational times and higher-quality solutions.

By analyzing economic incentives, we show that reducing the number of stop signs at crossroads (e.g., from four to three) leads to a self-enforcing equilibrium with significant benefits. This simpler mechanism saves fuel gas, reduces time and infrastructure costs, lowers carbon emissions, and decreases police expenses. The new approach generates substantial economic gains compared to the traditional practice of erecting one stop sign in each direction.

Competition and market composition impact corporate adoption of green technologies. Cooperative markets with both green and non-green goods see increased green investment with reduced production costs. In competitive markets without green products, cost reductions facilitate green product introduction. Policy-making can leverage these findings for effective green production cost reductions.

We consider market and technological equilibria in Cournot competition with linear and isoelastic demand between firms heterogeneous in operational and environmental efficiency. We examine possibilities and limitations of incentivizing “green” technology choice with environmental/”carbon” taxes. The resultant equilibria and the impact of taxation qualitatively differ with demand function.

We examine whether and how renewable energy (RE) adoption can increase or decrease energy efficiency (EE) improvement. Using site-level data, we estimate the impact of changes in RE usage and in the acquisition approach on the EE. We find that using RE to meet 10% more of a site's energy demand led to an additional 2.0% improvement in EE. While purchasing RE credits or entering into power purchase agreements led to gains in EE, installing on-site RE generators had no effect.

Clean energy technologies require large amounts of critical metals and minerals, the demand for which is skyrocketing as the global energy sector rapidly shifts towards renewable energy. The scarcity of critical materials may prevent governments from reaching their ambitious clean energy targets and jeopardize the profitability of the renewable energy sector. Inspired by this challenge, we investigate mechanisms that can address the impact of scarcity on the renewable energy transition momentum.

The ability to produce electricity when renewables allow and store it for later demand is crucial for emissions reduction. Electric Vehicles (EVs) could provide distributed energy storage to the electric grid through optimal charging and discharging, however, highly complex, non-linear driver behavior must be accounted for. In this work we collaborate with an American EV manufacturer to combine machine learning with optimization to model driver behavior to size the capacity of EV energy storage.

Quantifying the operating cost of a battery has been considered a key challenge for economic profitability in the battery literature/industry. One key modeling difficulty is that the cost is realized through different types of performance degradation. This work takes an analytical approach to compare the optimal dynamic policy under different degradation models (capacity-degradation-only, efficiency-degradation-only, and correlated degradation).

Organizational support improves worker retention. In K-12 education, teacher crowdfunding platforms bridge resource gaps. We analyze teachers' actions on DonorsChoose and employment data. Crowdfunding reduces attrition rates by 1.7 pp, benefiting projects that personalize classrooms and teaching pedagogy. Identity-relevant resources decrease school and system exit by 1.8-2.4 pp. Findings highlight the role of accessible resources in enhancing worker retention.

AVs revolutionize transportation, but cost barriers persist. Open platforms with human drivers aid deployment. Incentive alignment in the supply chain is vital. We analyze the game involving platforms, AV suppliers, and contractors. Underutilization risk affects profit. Contracting solutions align the chain, ensuring half of optimal profit. Misalignment hampers AV adoption. Contracts mitigate efficiency loss.

Algorithms control pricing and match customers and workers in the gig economy. However, algorithms face several critiques: they lack transparency, can be biased, and can be inefficient. We empirically analyze these issues and show that algorithms lose efficiency from two sources: competition between platforms and misaligned worker incentives. We model workers' strategic responses to variation in pricing and estimate counterfactuals on the effects of minimum wage and transparent pricing policies.

Typically, fairness properties of algorithmic decisions are analyzed as if the machine predictions were implemented directly. However, many machine predictions are instead deployed to assist a human decision-maker who retains the ultimate decision authority. In this article, we therefore consider how properties of machine predictions affect the resulting human decisions through both a formal model and lab experiment.

The use of algorithms may be preferred over human judgment; however, algorithms are prone to bias. In this study, we designed a behavioral lab experiment environment to imitate the processes instated by the companies when asking their programmers to develop an ML algorithm for any given task. We investigate the causes of programmers developing biased ML algorithms and study how these causes can be mitigated through policies to encourage the development of fair and less-biased ML algorithms.

TBD

We use a cluster randomized field experiment to study the effect of performance monitoring on the efficiency of public procurement in Chile. In collaboration with the Chilean Public Procurement Office, we provided monthly reports on the purchasing performance of individual procurement officers and services. After 5 months of treatment exposure, we find that the reports generated significant reductions in overspending in interventions that point towards extrinsic motivation of buyers.

Problem: This paper examines the impact of auction timing on bidder behavior and welfare. Methodology/Results: A structural auction model is developed and applied to real data. Complementarities in bids are found to affect contract combinations. An optimization algorithm for auction schedules is proposed. Managerial Implications: Optimizing auction schedules can reduce procurement costs by over 8%.