Knowledge that Transforms

Problem definition: Motivated by ride-hailing platforms such as Uber, Lyft and Didi, we study the problem of matching riders with self-interested drivers over a spatial network. We focus on the performance impact of two operational platform controls—demand-side admission control and supply-side repositioning control—considering the interplay with two practically important challenges: (i) spatial demand imbalances prevail for extended periods of time; and (ii) self-interested drivers strategically decide whether to join the network, and, if so, whether to reposition when not serving riders. Methodology/results: We develop and analyze the steady-state behavior of a novel game-theoretic fluid model of a two-location, four-route loss network. First, we fully characterize and compare the steady-state system equilibria under three control regimes, from minimal control to centralized control. Second, we provide insights on how and why platform control impacts equilibrium performance, notably with new findings on the role of admission control: the platform may find it optimal to strategically reject demand at the low-demand location even if drivers are in excess supply, to induce repositioning to the high-demand location. We provide necessary and sufficient conditions for this policy. Third, we derive upper bounds on the platform’s and drivers’ benefits caused by increased platform control; these are more significant under moderate capacity and significant cross-location demand imbalance. Managerial implications: Our results contribute important guidelines on the optimal operations of ride-hailing networks. Our model can also inform the design of driver compensation structures that support more centralized network control. Supplemental Material: The e-companion and Supplemental Material are available at https://doi.org/10.1287/msom.2023.1221 .

Problem definition: Suppliers are increasingly involved in innovation activities that contribute to a firm’s product quality and introduce risks to firms’ quality control, leading to quality failures and recalls. This quality trade-off suggests the possibility of a nonlinear relationship between supplier innovation and product recalls, which is the focus of this research. Recall literature focuses on firms’ internal drivers of recalls, whereas anecdotal evidence increasingly points to the role of external drivers, such as suppliers. We contribute to the literature by examining supplier innovation as an external driver leading to recalls via quality and risk spillovers. Methodology/results: We collect and assemble a unique panel data set of consumer product recalls from firms and their supply bases (i.e., first tier suppliers). We estimate econometric models to examine the nonlinear relationship between supply base innovation, measured by research and development (R&D) intensity of the supply bases, and the likelihood of product recalls. We find a quadratic (i.e., U-shaped) relationship between the probability of recalls and supply base R&D intensity. We also find that this nonlinear relationship is critically related to three specific sources of risk: radicalness of supplier innovation, technological distance between firms and their suppliers, and complexity of supply base. Managerial implications: Our findings suggest that firms should be mindful of the quality trade-offs in encouraging supplier innovation to reduce product recalls. Further, to minimize recall risks, firms should better evaluate and manage the risks associated with external supplier knowledge that is novel and different and closely work with global suppliers to reduce coordination challenges in knowledge transfer and integration. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.1213 .

Problem definition: Patient-level data from 72 magnetic resonance imaging (MRI) hospitals in Ontario, Canada from 2013 to 2017 show that over 60% of patients exceeded their wait time targets. We conduct a data-driven analysis to quantify the reduction in the patient fraction exceeding (FET) target for MRI services through geographic virtual resource-sharing while limiting incremental driving time. We present a data-driven method to solve the geographic pooling problem of partitioning 72 hospitals with heterogeneous patients with different wait time targets located in a two-dimensional region into a set of clusters. Methodology/results: We propose an “augmented-priority rule,” which is a sequencing rule that balances the patient’s initial priority class and the number of days until her wait time target. We then use neural networks to predict patient arrival and service times. We combine this predicted information and the sequencing rule to implement “advance scheduling,” which informs the patient of her treatment day and location when requesting an MRI scan. We then optimize the number of geographic resource pools among the 72 hospitals using genetic algorithms. Our resource-pooling model lowers the FET from 66% to 36% while constraining the average incremental travel time below three hours. In addition, our model shows that only 10 additional scanners are needed to achieve 10% FET, whereas 50 additional scanners would be needed without resource sharing. Over 70% of the hospitals are not worse off financially. Each individual hospital, measured over at least two weeks, achieves a higher machine utilization and a lower FET. Managerial implications: Our paper provides a practical, data-driven geographical resource-sharing model that hospitals can readily implement. Our method achieves a near-optimal solution with low computational complexity. Using smart data-driven scheduling, a little extra capacity placed at the right location is all we need to achieve the desired FET under geographic resource-sharing. Funding: This paper is supported by the following grant: Canadian Institutes of Health Research (CIHR) [Grant CIHR-950-231935]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.1225 .

Problem definition: In this work, we examine how ridesharing platforms affect changes in short-term vehicle purchasing. On the one hand, if the introduction of such platforms motivates owners to use the idle capacity of their existing vehicles to accrue rents, vehicle sales might fall. On the other hand, if the platform induces would-be drivers to purchase new vehicles in order to participate, vehicle sales might rise. Academic/practical relevance: Whereas operations management researchers have begun to broach this subject analytically, this work provides empirical evidence of the impact of ridesharing platforms on new vehicle ownership. Further, we assess heterogeneity in the effect across vehicle type and location. Methodology: We examine this tension using a unique data set of new vehicle registrations in China. In doing so, we exploit the variation in the timing of Uber entry using a difference-in-difference approach. Results: Findings suggest Uber entry is associated with a significant short-term increase in private new vehicle ownership, indicating that consumers actively change their stock of held resources to capture excess rents offered by the platform. These effects exclusively manifest among vehicle brands that qualify for the platform. Further, inasmuch as sales of vehicles with smaller displacement increase more than large-displacement vehicles, results indicate that the effect of Uber entry varies considerably across vehicle types. Finally, results indicate that the effects are stronger in locations where established public transportation options are weaker. Managerial implications: Results provide initial evidence that manufacturers can benefit from the emergence of the sharing economy, especially manufacturers whose products align with the needs of platform participants. For policy makers, our findings further undercut claims made by platforms that the individuals working on them are exploiting already existing resources, suggesting some form of nascent professionalism on the part of platform workers. Funding: The research was supported by the Key Program of National Natural Science of China [Grant 71832010]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2022.1183 .

Problem definition: Mitigating the COVID-19 pandemic poses a series of unprecedented challenges, including predicting new cases and deaths, understanding true prevalence beyond what tests are able to detect, and allocating different vaccines across various regions. In this paper, we describe our efforts to tackle these issues and explore the impact on combating the pandemic in terms of case and death prediction, true prevalence, and fair vaccine distribution. Methodology/results: We present the methods we developed for predicting cases and deaths using a novel machine-learning-based aggregation method to create a single prediction that we call MIT-Cassandra. We further incorporate COVID-19 case prediction to determine true prevalence and incorporate this prevalence into an optimization model for efficiently and fairly managing the operations of vaccine allocation. We study the trade-offs of vaccine allocation between different regions and age groups, as well as first- and second-dose distribution of different vaccines. This also allows us to provide insights into how prevalence and exposure of the disease in different parts of the population can affect the distribution of different vaccine doses in a fair way. Managerial implications: MIT-Cassandra is currently being used by the Centers for Disease Control and Prevention and is consistently among the best-performing methods in terms of accuracy, often ranking at the top. In addition, our work has been helping decision makers by predicting how cases and true prevalence of COVID-19 will progress over the next few months in different regions and utilizing the knowledge for vaccine distribution under various operational constraints. Finally, and very importantly, our work has specifically been used as part of a collaboration with the Massachusetts Institute of Technology's (MIT’s) Quest for Intelligence and as part of MIT’s process to reopen the institute. Funding: Financial support from MIT Quest for Intelligence is gratefully acknowledged. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2022.1160 .

Problem definition: How to use social media to predict style color and jeans fit sales for a retailer. Academic/practical relevance: Neither retail practice nor the academic literature provides a method for using social media to predict style color and jeans fit sales for a retailer. We present and validate a systematic approach for doing that. Methodology: Demand forecasting in the fashion industry is challenging due to short product lifetimes, long manufacturing lead times, and constant innovation of fashion products. We investigate the value of social media information for color trends and jeans fit forecasting. We partner with three multinational retailers, two apparel and one footwear, and combine their proprietary data sets with web-crawled publicly available data on Twitter and the Google Search Volume Index. We implement a variety of machine learning models to develop forecasts that can be used in setting the initial shipment quantity for an item, arguably the most important decision for fashion retailers. Results: Our findings show that fine-grained social media information has significant predictive power in forecasting color and fit demands months in advance of the sales season, and therefore greatly helps in making the initial shipment quantity decision. The predictive power of including social media features, measured by the improvement of the out-of-sample mean absolute deviation over current practice ranges from 24% to 57%. Managerial implications: To our knowledge, this study is the first to explore and demonstrate the value of social media information in fashion demand forecasting in a way that is practical and operable for fashion retailers. With consistent results across all three retailers, we demonstrate the robustness of our findings over market and geographic heterogeneity, and different forecast horizons. Moreover, we discuss potential mechanisms that might be driving this significant predictive power. Our results suggest that changes in fashion demand are driven more by “bottom-up” changes in consumer preferences than by “top-down” influence from the fashion industry. Funding: This work was supported by Wharton School Fishman-Davidson Center for Service and Operations Management, the Wharton School Baker Retailing Center, and the Wharton School Risk Management Center Russell Ackoff Doctoral Student Fellowship. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.1193 .

Problem definition: In brick-and-mortar fashion retail stores, inventory stockouts are frequent. When a specific size of a fashion product is out of stock, the unmet demand might not be completely lost because of spillovers to adjacent sizes of the same style or to other styles. Little research has been done to study consumer response to stockouts of fashion products because researchers had limited access to proprietary data of fashion retailers and because it is challenging to estimate stockout-based demand spillover patterns using existing approaches due to the enormous number of stockkeeping units (SKUs) and frequent stockouts in fashion retail stores. To fill this void in the literature, we empirically estimate the stockout-based demand spillover effect in a fashion retail setting. Methodology/results: We obtain a large-scale data set from a fashion retail chain selling world-renowned sportswear brands. The retail stores in the sample are dedicated to products of a single brand. Using around 1.5 million granular and real-time sales and inventory records of 217 stores, 503 men’s footwear products, and 4,024 SKUs over a two-year period, we develop a difference-in-differences framework to estimate the stockout-based cross-size demand spillover effect. We demonstrate the validity of this framework by conducting a pretrend test and a placebo test. We find that roughly 51.7% of the unmet demand of an out-of-stock SKU spills over to adjacent sizes of the same style when they are in stock: 25.1% to the adjacent-larger size and 26.6% to the adjacent-smaller size. The cross-size demand spillover effect is larger in regular stores than in flagship stores, larger for casual sports shoes than for specialized sports shoes, and larger for low-price products than for high-price products. Adapting an existing attribute-based demand model to our setting, we estimate that roughly 20.2% of the unmet demand of an out-of-stock SKU spills over to different styles when they are in stock. Taken together, these estimations suggest that about 28.1% of the unmet demand of an out-of-stock SKU becomes lost sales. We further find that when stockouts are widespread among SKUs, stockout-based demand spillovers are significantly reduced, resulting in much increased lost sales. Managerial implications: First, we empirically quantify the stockout-based cross-size demand spillover effect and its impact on lost sales in a brick-and-mortar fashion retail setting. Second, our simulation analysis shows that incorporating the cross-size demand spillover effect into the sportswear retail chain’s proactive transshipment decision can substantially reduce its transshipment cost and improve its profitability. Funding: S. Li and S. Huang were supported by the National Natural Science Foundation of China [Grant 72188101] and the Center for Data Centric Management in the Department of Industrial Engineering at Tsinghua University. Supplemental Material: The online appendices are available at https://doi.org/10.1287/msom.2022.1135 .

Problem definition: We present a data-driven study of the secondary ticket market. In particular, we are primarily concerned with accurately estimating price sensitivity for listed tickets. In this setting, there are many issues including endogeneity, heterogeneity in price sensitivity for different tickets, binary outcomes, and nonlinear interactions between ticket features. Our secondary goal is to highlight how this estimation can be integrated into a prescriptive trading strategy for buying and selling tickets in an active marketplace. Academic/practical relevance: We present a novel method for demand estimation with heterogeneous treatment effect in the presence of confounding. In practice, we embed this method within an optimization framework for ticket reselling, providing the ticket reselling platform with a new framework for pricing tickets on its platform. Methodology: We introduce a general double/orthogonalized machine learning method for classification problems. This method allows us to isolate the causal effects of price on the outcome by removing the conditional effects of the ticket and market features. Furthermore, we introduce a novel loss function that can be easily incorporated into powerful, off-the-shelf machine learning algorithms, including gradient boosted trees. We show how, in the presence of hidden confounding variables, instrumental variables can be incorporated. Results: Using a wide range of synthetic data sets, we show this approach beats state-of-the-art machine learning and causal inference approaches for estimating treatment effects in the classification setting. Furthermore, using National Basketball Association ticket listings from the 2014–2015 season, we show that probit models with instrumental variables, previously used for price estimation of tickets in the resale market, are significantly less accurate and potentially misspecified relative to our proposed approach. Through pricing simulations, we show our proposed method can achieve an 11% return on investment by buying and selling tickets, whereas existing techniques are not profitable. Managerial implications: The knowledge of how to price tickets on its platform offers a range of potential opportunities for our collaborator, both in terms of understanding sellers on their platform and in developing new products to offer them. History: This paper has been accepted as part of the 2019 Manufacturing & Service Operations Management Practice-Based Research Competition. Funding: This work was supported by the National Science Foundation [Grant CMMI-1563343]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/msom.2021.1065 .

Problem definition: Firms heavily invest in big data technologies to collect consumer data and infer consumer preferences for price discrimination. However, consumers can use technological devices to manipulate their data and fool firms to obtain better deals. We examine how a firm invests in collecting consumer data and makes pricing decisions and whether it should disclose its scope of data collection to consumers who can manipulate their data. Methodology/results: We develop a game-theoretic model to consider a market in which a firm caters to consumers with heterogeneous preferences for a product. The firm collects consumer data to identify their types and issue an individualized price, whereas consumers can incur a cost to manipulate data and mimic the other type. We find that when the firm does not disclose its scope of data collection to consumers, it collects more consumer data. When the firm discloses its scope of data collection, it reduces data collection even when collecting more data is costless. The optimal scope of data collection increases when it is more costly for consumers to manipulate data but decreases when consumer demand becomes more heterogeneous. Moreover, a lower cost for consumers to manipulate data can be detrimental to both the firm and consumers. Lastly, disclosure of data collection scope increases firm profit, consumer surplus, and social welfare. Managerial implications: Our findings suggest that a firm should adjust its scope of data collection and prices based on whether the firm discloses the data collection scope, consumers’ manipulation cost, and demand heterogeneity. Public policies should require firms to disclose their data collection scope to increase consumer surplus and social welfare. Even without such a mandatory disclosure policy, firms should voluntarily disclose their data collection scope to increase profit. Moreover, public educational programs that train consumers to manipulate their data or raise their awareness of manipulation tools can ultimately hurt consumers and firms. Funding: X. Li acknowledges financial support from the Hong Kong Research Grants Council [Grant 21500920]. K. J. Li acknowledges financial support from the Weimer Faculty Fellowship. Supplemental Material: The Online Appendix is available at https://doi.org/10.1287/msom.2022.1153 .

Problem definition: In a Markovian queueing system with strategic customers, a reward is gained from completing service, and a loss is incurred while waiting to be served. The common assumption in the queueing literature is that such loss is a function of the customer’s waiting time. This paper takes a different and novel approach in that it models the customer’s loss incurred because of negative network effects while waiting with others, which increases as the exposure to others increases. Methodology: Waiting time is complemented by two innovative measures that capture negative effects on a tagged customer joining an M/M/c queue: the total number of customers the tagged customer meets and person-time exposure to these customers while waiting to be served. Threshold joining strategies inducing M/M/c/n–type queues are studied in this context. Results: The distributions of exposure size and exposure time of a customer joining the queue at a given position are analytically derived. Equilibria under conditions of no reneging are identified as threshold strategies. If the customer’s loss function is concave (such as an exponential model for the chance of infection during a pandemic), there is an equilibrium threshold strategy under which customers do not renege from the queue, even if reneging is allowed. The price of anarchy caused by lack of coordination among the individuals acting is identified. Unlike the equilibrium threshold built under the restrictive assumption that all potential customers have the same utility function, a novel safe threshold concept is introduced, a queue size at which a customer who joins the facility and stays until completing service has positive expected utility regardless of the actions of the other customers. Managerial implications: The implications of negative network effects caused by congestion in a queueing system are of interest to queue managers and, in particular, affect the optimal size of the waiting area. Safe and equilibrium thresholds are contrasted with the socially optimal threshold set by a regulator, and the safe threshold is suggested as a managerial tool to design the waiting room size. Funding: This work was supported by the Israel Science Foundation [Grants ISF 1898/21 and ISF 852/22]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.1223 .