Robert Louis Bray
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Robert Louis Bray has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
Ahmet Colak
College of Business, Clemson University, Clemson, South Carolina 29634
3 shared publicationsDecio Coviello
HEC Montréal, Montréal, Québec H3T 2A7, Canada
2 shared publicationsHaim Mendelson
Graduate School of Business, Stanford University, Stanford, California 94305
2 shared publicationsIoannis Stamatopoulos
2 shared publications- Ioannis Stamatopoulos
University of Texas at Austin Red McCombs School of Business, Information, Risk, and Operations Management, Austin, Texas, 78705, United States
2 shared publications Andrea Ichino
University of Bologna
1 shared publication
Latest Publications
- 1 June 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Service Science
- Vol. 18 (2) , 96-119
The Impact of Market Growth on Delivery Speed: Evidence from JD.com
This paper examines how rapid market expansion in e-commerce platforms can generate disparities in customer experience, focusing on fulfillment speed as a key operational outcome. Using granular order-level data from JD.com, we document a robust negative relationship between customer tenure and delivery time; newer customers systematically receive slower deliveries than legacy customers. We show that this lifetime effect is not explained by differences in customer preferences, order composition, or SKU availability but instead reflects frictions in the supply chain network. In particular, newer customers are more likely to reside in recently entered or logistically peripheral areas, where fulfillment infrastructure is less developed. Even after controlling for routes and urban context, the delivery gap persists, driven primarily by longer last-mile segments. We further show that regions with more centralized hub-and-spoke network structures exhibit larger disparities in delivery performance, whereas more decentralized configurations yield more uniform outcomes across cohorts. Together, these findings highlight the operational challenges of matching fulfillment capacity to demand in a spatially expanding market and underscore the need to align logistics strategy with patterns of customer growth.
- 1 June 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Service Science
- Vol. 18 (2) , 96-119
The Impact of Market Growth on Delivery Speed: Evidence from JD.com
This paper examines how rapid market expansion in e-commerce platforms can generate disparities in customer experience, focusing on fulfillment speed as a key operational outcome. Using granular order-level data from JD.com, we document a robust negative relationship between customer tenure and delivery time; newer customers systematically receive slower deliveries than legacy customers. We show that this lifetime effect is not explained by differences in customer preferences, order composition, or SKU availability but instead reflects frictions in the supply chain network. In particular, newer customers are more likely to reside in recently entered or logistically peripheral areas, where fulfillment infrastructure is less developed. Even after controlling for routes and urban context, the delivery gap persists, driven primarily by longer last-mile segments. We further show that regions with more centralized hub-and-spoke network structures exhibit larger disparities in delivery performance, whereas more decentralized configurations yield more uniform outcomes across cohorts. Together, these findings highlight the operational challenges of matching fulfillment capacity to demand in a spatially expanding market and underscore the need to align logistics strategy with patterns of customer growth.
- Open Access
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Training Neural Networks Embedded in Dynamic Discrete Choice Models
We develop the first general-purpose estimator for infinite-horizon dynamic discrete choice models whose estimation problem, after pre-computation, is unencumbered by large systems of linear equations-either imposed as constraints, or embedded in the objective function. Our unnested fixed point (UFXP) and optimal unnested fixed point (OUFXP) estimators exploit a dual representation of Bellman's equation to separate the utility parameters from the dynamic programming fixed point. We establish the consistency and asymptotic normality of UFXP and OUFXP, as well as the efficiency of the latter. Our estimators enable researchers to model utility functions non-parametrically via flexible neural-network approximations.
- 20 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTraining Neural Networks Embedded in Dynamic Discrete Choice Models
We develop the first general-purpose estimator for infinite-horizon dynamic discrete choice models whose estimation problem, after pre-computation, is unencumbered by large systems of linear equations-either imposed as constraints, or embedded in the objective function. Our unnested fixed point (UFXP) and optimal unnested fixed point (OUFXP) estimators exploit a dual representation of Bellman's equation to separate the utility parameters from the dynamic programming fixed point. We establish the consistency and asymptotic normality of UFXP and OUFXP, as well as the efficiency of the latter. Our estimators enable researchers to model utility functions non-parametrically via flexible neural-network approximations.
- Top Cited Papers
- 1 July 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 73 (4) , 2188-2203
Logarithmic Regret in Multisecretary and Online Linear Programs with Continuous Valuations
In 2019, Alessandro Arlotto and Itai Gurvich proved that the expected regret in the multisecretary problem is uniformly bounded as the number of jobs to fill goes to infinity. However, they modeled the secretary valuation distribution as discrete and known, and they concluded their article by explaining that they did not know whether their bound would hold with continuous or unknown valuation distributions. In “Logarithmic Regret in Multisecretary and Online Linear Programs with Continuous Valuations,” Robert L. Bray provides an answer, furnishing lower and upper bounds that demonstrate that the regret in multisecretary problem grows like the logarithm of the number of positions to fill when the distribution of secretary valuations continuous or unknown. He then develops the argument to establish analogous bounds for the more general online linear program, defined by Xiaocheng Li and Yinyu Ye. Bray tightens Li and Ye’s upper regret bound and develops a corresponding lower regret bound. To create these bounds, Bray develops several new convergence results for the online linear program’s dual variables.
- 1 July 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 73 (4) , 2188-2203
Top Cited PapersLogarithmic Regret in Multisecretary and Online Linear Programs with Continuous Valuations
In 2019, Alessandro Arlotto and Itai Gurvich proved that the expected regret in the multisecretary problem is uniformly bounded as the number of jobs to fill goes to infinity. However, they modeled the secretary valuation distribution as discrete and known, and they concluded their article by explaining that they did not know whether their bound would hold with continuous or unknown valuation distributions. In “Logarithmic Regret in Multisecretary and Online Linear Programs with Continuous Valuations,” Robert L. Bray provides an answer, furnishing lower and upper bounds that demonstrate that the regret in multisecretary problem grows like the logarithm of the number of positions to fill when the distribution of secretary valuations continuous or unknown. He then develops the argument to establish analogous bounds for the more general online linear program, defined by Xiaocheng Li and Yinyu Ye. Bray tightens Li and Ye’s upper regret bound and develops a corresponding lower regret bound. To create these bounds, Bray develops several new convergence results for the online linear program’s dual variables.
- 5 March 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Applied Analytics
A Tutorial on Teaching Data Analytics with Generative AI
This tutorial discusses how large language models (LLMs) can transform data analytics instruction. It details innovative ways to blend AI into classes, from turning problem sets into AI tutoring sessions to crafting AI-driven group activities.
- 5 March 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Applied Analytics
A Tutorial on Teaching Data Analytics with Generative AI
This tutorial discusses how large language models (LLMs) can transform data analytics instruction. It details innovative ways to blend AI into classes, from turning problem sets into AI tutoring sessions to crafting AI-driven group activities.
- Open Access
- 14 August 2024
- Published by Elsevier in SSRN Electronic Journal
Observational Price Variation in Scanner Data Does Not Reproduce Experimental Price Elasticities
We compare experimental elasticity estimates from 82 “test” stores to observational estimates from 34 “control” stores of a Midwestern grocery retailer. The experiment generated 389,890 prices and lasted 35 weeks. During the experiment, the average experimental elasticity is -0.34, whereas the average observational elasticity is -1.97. The gap is even wider in the difference-in-differences, and replicates at the category and the product levels. We cannot reconcile this gap by controlling for promotions, conducting an event study around each price change, focusing on base-price changes, accounting for longer-term price effects, disaggregating estimates by price level and price-change magnitude, or instrumenting with chain price, lagged price, raw input price, or wholesale price.
- 14 August 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessObservational Price Variation in Scanner Data Does Not Reproduce Experimental Price Elasticities
We compare experimental elasticity estimates from 82 “test” stores to observational estimates from 34 “control” stores of a Midwestern grocery retailer. The experiment generated 389,890 prices and lasted 35 weeks. During the experiment, the average experimental elasticity is -0.34, whereas the average observational elasticity is -1.97. The gap is even wider in the difference-in-differences, and replicates at the category and the product levels. We cannot reconcile this gap by controlling for promotions, conducting an event study around each price change, focusing on base-price changes, accounting for longer-term price effects, disaggregating estimates by price level and price-change magnitude, or instrumenting with chain price, lagged price, raw input price, or wholesale price.
- 1 May 2023
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Manufacturing & Service Operations Management
- Vol. 25 (3) , 812-826
Operational Transparency: Showing When Work Gets Done
Problem definition: Do the benefits of operational transparency depend on when the work is done? Academic/practical relevance: This work connects the operations management literature on operational transparency with the psychology literature on the peak-end effect. Methodology: This study examines how customers respond to operational transparency with parcel delivery data from the Cainiao Network, the logistics arm of Alibaba. The sample comprises 4.68 million deliveries. Each delivery has between 4 and 10 track-package activities, which customers can check in real time, and a delivery service score, which customers leave after receiving the package. Instrumental-variable regressions quantify the causal effect of track-package-activity times on delivery scores. Results: The regressions suggest that customers punish early idleness less than late idleness, leaving higher delivery service scores when track-package activities cluster toward the end of the shipping horizon. For example, if a shipment takes 100 hours, then delaying the time of the average action from hour 20 to hour 80 increases the expected delivery score by approximately the same amount as expediting the arrival time from hour 100 to hour 73. Managerial implications: Memory limitations make customers especially sensitive to how service operations end. History: This paper has been accepted as part of the 2018 MSOM Data Driven Research Challenge.
- 1 May 2023
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Manufacturing & Service Operations Management
- Vol. 25 (3) , 812-826
Operational Transparency: Showing When Work Gets Done
Problem definition: Do the benefits of operational transparency depend on when the work is done? Academic/practical relevance: This work connects the operations management literature on operational transparency with the psychology literature on the peak-end effect. Methodology: This study examines how customers respond to operational transparency with parcel delivery data from the Cainiao Network, the logistics arm of Alibaba. The sample comprises 4.68 million deliveries. Each delivery has between 4 and 10 track-package activities, which customers can check in real time, and a delivery service score, which customers leave after receiving the package. Instrumental-variable regressions quantify the causal effect of track-package-activity times on delivery scores. Results: The regressions suggest that customers punish early idleness less than late idleness, leaving higher delivery service scores when track-package activities cluster toward the end of the shipping horizon. For example, if a shipment takes 100 hours, then delaying the time of the average action from hour 20 to hour 80 increases the expected delivery score by approximately the same amount as expediting the arrival time from hour 100 to hour 73. Managerial implications: Memory limitations make customers especially sensitive to how service operations end. History: This paper has been accepted as part of the 2018 MSOM Data Driven Research Challenge.
- Open Access
- 10 October 2022
- Published by Elsevier in SSRN Electronic Journal
A Disquieting Lack of Evidence for Disintermediation in a Home-Cleaning Platform
We study a sample of data from an online platform that matches home cleaners with people who want their homes cleaned. The dataset has a key feature: it reports with high frequency the geographic distances between the cleaners and their appointed residences during both the cleaners' working hours and off hours---viz., anytime the app is open on their phones. For example, we observe 57 and 46 distance snapshots a day for the mean and median cleaner, respectively. We use these distance measurements to test whether the cleaners disintermediate the work, returning to the residences to perform some undisclosed cleanings for which they will not have to pay the platform middleman its cut. We find no evidence of such disintermediation---in fact, we find strong evidence to the contrary. Specifically, we reject the null hypothesis that at least one in 10,000 (cleaner, residence, date) triples without an official cleaning had an unofficial cleaning.
- 10 October 2022
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Disquieting Lack of Evidence for Disintermediation in a Home-Cleaning Platform
We study a sample of data from an online platform that matches home cleaners with people who want their homes cleaned. The dataset has a key feature: it reports with high frequency the geographic distances between the cleaners and their appointed residences during both the cleaners' working hours and off hours---viz., anytime the app is open on their phones. For example, we observe 57 and 46 distance snapshots a day for the mean and median cleaner, respectively. We use these distance measurements to test whether the cleaners disintermediate the work, returning to the residences to perform some undisclosed cleanings for which they will not have to pay the platform middleman its cut. We find no evidence of such disintermediation---in fact, we find strong evidence to the contrary. Specifically, we reject the null hypothesis that at least one in 10,000 (cleaner, residence, date) triples without an official cleaning had an unofficial cleaning.
- 1 March 2022
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 70 (2) , 748-765
Menu Costs and the Bullwhip Effect: Supply Chain Implications of Dynamic Pricing
Suppose that technology reduces price-adjustment costs (e.g., the costs of printing and changing price tags), and as a result prices at grocery stores change more dynamically. Will this change mean less stability or more stability for grocery supply chains? In other words, will more dynamic pricing downstream mean more last-minute purchases, more overtime work, and more empty space in trucks and warehouses? Or will it mean more regular and more standardized orders, smoother schedules, and less waste? To answer this question, we fit a pricing and ordering model to data from a large Chinese supermarket chain (daily prices, sales, inventories, and shipments from products from seven categories at 78 stores for 3.5 years) and then simulate a counterfactual grocery chain in which the estimated price-adjustment costs are set to zero. We find that the removal of price-adjustment costs stabilizes the supply chain, reducing both its shipment volatility, its sales volatility, and its bullwhip (the difference between the shipment and sales volatility).
- 1 March 2022
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 70 (2) , 748-765
Menu Costs and the Bullwhip Effect: Supply Chain Implications of Dynamic Pricing
Suppose that technology reduces price-adjustment costs (e.g., the costs of printing and changing price tags), and as a result prices at grocery stores change more dynamically. Will this change mean less stability or more stability for grocery supply chains? In other words, will more dynamic pricing downstream mean more last-minute purchases, more overtime work, and more empty space in trucks and warehouses? Or will it mean more regular and more standardized orders, smoother schedules, and less waste? To answer this question, we fit a pricing and ordering model to data from a large Chinese supermarket chain (daily prices, sales, inventories, and shipments from products from seven categories at 78 stores for 3.5 years) and then simulate a counterfactual grocery chain in which the estimated price-adjustment costs are set to zero. We find that the removal of price-adjustment costs stabilizes the supply chain, reducing both its shipment volatility, its sales volatility, and its bullwhip (the difference between the shipment and sales volatility).
- 1 October 2019
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Management Science
- Vol. 65 (10) , 4598-4606
Markov Decision Processes with Exogenous Variables
I present two algorithms for solving dynamic programs with exogenous variables: endogenous value iteration and endogenous policy iteration. These algorithms are always at least as fast as relative value iteration and relative policy iteration, and they are faster when the endogenous variables converge to their stationary distributions sooner than the exogenous variables. This paper was accepted by Yinyu Ye, optimization.
- 1 October 2019
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Management Science
- Vol. 65 (10) , 4598-4606
Markov Decision Processes with Exogenous Variables
I present two algorithms for solving dynamic programs with exogenous variables: endogenous value iteration and endogenous policy iteration. These algorithms are always at least as fast as relative value iteration and relative policy iteration, and they are faster when the endogenous variables converge to their stationary distributions sooner than the exogenous variables. This paper was accepted by Yinyu Ye, optimization.