Knowledge that Transforms

Live-streaming e-commerce has gained tremendous success as a new form of business model over the past few years. Nevertheless, there has been scarce research examining the effect of exogenous stimulus-driven factors (i.e., social cues such as promotion coupons) on the profit of the live-streaming platform. We model the continuous evolvement of the aggregate viewer involvement level with geometric Brownian motion and investigate the optimal timing and depth of promotion insertion policy. Our findings reveal that the optimal promotion insertion policy for the live-streaming room is a threshold policy that consists of a start-to-promote threshold and an involvement target when the trend of the aggregate involvement level is not large and the promotion cost is not high. Under some scenarios, the promotion planning process is constrained by various business rules, such as the promotion depth being limited by a discrete set (e.g., integral multiples of D ). Although this problem involves more complex constraints, we analytically show that a variation of the abovementioned threshold policy is the optimal strategy among all feasible policies, i.e., when the aggregate viewer involvement drops below a certain level ( P ), the live-streaming host promotes with the depth of the least integral multiple of D to raise the viewer involvement level above P . Finally, to examine the effectiveness of our proposed policy, we compare it with prevalent industry practices. The result shows that the proposed promotion insertion policy outperforms prevailing ones significantly in generating profit for the live-streaming room.

In software markets, the sheer number of available applications makes it rather challenging for any given new one to stand out and be noticed by consumers. Moreover, a push towards privacy by regulators and consumers is making it harder to target consumers. As such, firms have to rely on more non-targeted go-to-market strategies. We explore two popular strategies through which developers can catalyze adoption by helping consumers directly or indirectly learn the value of their products— seeding (free full-feature product giveaways to a subset of the consumer base) and time-limited freemium ( TLF ). Seeding, as a business strategy, existed for a long time. On the other hand, the feasibility to offer market-wide TLF became mainstream more recently, with the advent of the Internet and a plethora of digital tools. Thus, a natural question emerges—if TLF represents nowadays a feasible and easily implementable strategy for software applications, has seeding approach been rendered irrelevant in these markets? In this study, we provide managerial recommendations on when each of these strategies with a free full-feature-consumption component is optimal, based on social and self-learning dynamics, consumer priors, adoption costs, and individual product value depreciation. To that end, under a multi-period parsimonious unifying framework, we show that S becomes dominated as free trials enter the picture. We identify two specific market factors that, when present, can induce seeding to be optimal when consumers initially underestimate true product value—(i) user adoption costs and/or (ii) individual depreciation of value by usage. Moreover, we show that these two factors have a moderating effect on the impact of word-of-mouth (WOM) effects on the optimality of seeding. In the absence of these factors, stronger WOM effects alone cannot give seeding an edge against the other business strategies. However, once either depreciation or adoption costs are accounted for, strong WOM effects increase the relevance of seeding (enlarging its optimality region in the parameter space). Our results remain qualitatively consistent under a battery of robustness checks.

Automotive manufacturers have been under intense pressure to improve the fuel efficiency, or miles per gallon (MPG), of their internal combustion engine (ICE) vehicles. Research that explores hidden costs of continuing to push the limits of ICE MPG is sparse. We analyze the relationship between the MPG of 643 unique model-engines produced and sold by 18 manufacturers over an 18-year period and 19,785 vehicle safety defects reported to the National Highway Traffic Safety Administration (NHTSA). Each additional MPG improvement is associated with a 17.5 percent increase in safety defects. Results are confirmed when using fuel prices as an instrument for MPG. In post-hoc analyses, we find that a 2011 federal government fuel efficiency policy change explains an increase in the adoption of MPG-enhancing engine technologies, which subsequently led to an increase in safety defects. Moreover, we find that an increase in MPG is associated with an increase in safety recalls, and that safety defects mediate the relationship between MPG and safety recalls. To address selection bias concerns, we confirm our results using a separate quality reporting channel, demonstrating that increases in MPG are associated with a reduction in JD Power Quality & Reliability ratings. We also examine a key alternative explanation, that increasing MPG causes vehicles to be driven more miles, inherently leading to more defects. No evidence is found for this alternative; in fact we find the opposite to be true: an increase in MPG is associated with lower mileage at the time of the defect, indicating that increasing MPG leads to both more and faster defects. For manufacturers, our results identify an unexpected downside to striving to meet aggressive, federally mandated fuel efficiency standards: less safe vehicles. For NHTSA, our findings identify a possible contradiction in their two key objectives: sustainability and safety. Our study leads to policy recommendations, which we were fortunate to present to NHTSA’s Associate Administrator for Vehicle Safety Research and members of his staff.

Recent years have witnessed the widespread use of data to recognize repeat and new consumers to offer them different prices, that is, behavior-based pricing (BBP). While extensive research has explored the market impacts of BBP, most studies overlook the congestion effects arising from increased demand. This study addresses this gap by investigating the impact of BBP in congestion-prone systems, such as food delivery, with a particular focus on how delay commitment strategies influence both firms and consumers. Employing a two-period game-theoretic duopoly model integrated with a queueing system, where service qualities can be dynamically adjusted in each period, our analysis shows that BBP consistently intensifies firm competition, leaving firms worse off while simultaneously benefiting consumers. To enable BBP as a viable equilibrium, we propose a delay commitment strategy. Our findings show that while BBP intensifies price competition, it further alleviates service competition when combined with delay commitment. Notably, when capacity costs are high, the moderating effect on service competition becomes dominant, thereby offsetting the adverse impact of BBP on firms and ultimately leading to higher profits. From a long-term perspective, we further confirm that adopting delay commitment constitutes a Pareto-dominant equilibrium for both firms, irrespective of whether BBP is implemented. We further show that BBP continues to mitigate service competition even under partial coverage and may help increase market coverage. Moreover, when firms differ in their capacity costs, our results show that BBP can enhance social welfare by alleviating the negative externalities associated with congestion, termed the load-balancing effect. In summary, our research highlights the potential drawbacks of BBP for firms in service systems, but suggests that adopting a delay commitment strategy can revitalize BBP.

Decentralized applications (DApps)—digital applications operating on blockchain platforms—are transforming diverse industrial sectors. As blockchain platforms feature protocol-based governance, the mechanisms driving DApp adoption differ fundamentally from those in traditional digital environments. Drawing on observational learning theory, we conceptualize individual DApp adoption as decision interdependence driven by the visible actions of others. Building on platform governance theory, we identify three core governance-driven characteristics—decentralization, trustlessness, and token incentives—as key contextual moderators of this interdependence. We empirically investigate these dynamics on Ethereum, the world’s largest DApp platform, leveraging an unprecedented dataset covering its entire history through July 2025. Our findings confirm that decision interdependence is a significant driver of individual DApp adoption decisions, and that this effect is further amplified by blockchain’s unique governance characteristics: the effect of decision interdependence strengthens with higher decentralization and higher trustlessness, and is more pronounced for DApps offering token incentives. Extensive robustness checks, including DID and DDD analyses, alternative measures of blockchain characteristics, and accounting for copycat competition, reinforce these findings. By uncovering how blockchain’s distinctive governance mechanisms reshape decision interdependence, this study advances technology adoption theory beyond socially embedded settings and highlights how platform design fundamentally conditions user decision-making. Platform designers and DApp developers can leverage these insights to strategically calibrate blockchain governance features to amplify peer-driven adoption, advancing the operations management literature on blockchain-enabled digital platforms.

Emerging technologies like artificial intelligence (AI) can influence the way people relate and coordinate to improve performance. Our research explores how human–AI interactions affect relational coordination and operational performance. Competing views suggest that AI may augment or substitute for human capabilities. To examine conflicting views, we build on affordance theory, which suggests that the interactions between people and technology provide (i.e., afford) multiple potential actions, which could stimulate or diminish human interactions. Given the importance of human interactions during times of increasing technology use, we focus on relational coordination with its explicit emphasis on human relationships built on mutual respect and shared goals in fostering coordination. We introduce and find evidence of the concepts of convergent and divergent affordances to explain how AI affordances can result in similar or different outcomes. We perform multiple progressive, behavioral experiments in a virtual factory setting where participants work in teams with three functional roles, including operations, demand, and quality. We randomly assign individuals to one of three treatment groups that differ in the number of teammates that receive recommendations from an analytical AI based on machine learning. Through quantitative, qualitative, replication, recording, and multiphase methods, we find that human–AI interactions can stimulate relational coordination between individuals and improve team performance. We propose a theoretical framework, the ACT cycle, that explains how people interact with AI and with one another based on discussion, trust, and use of AI. Through recording and observing teams, we find evidence of various paths of interaction in that framework. We also find that discussion of AI can stimulate relational coordination, particularly in early phases of team interactions. The results provide theoretical and practical insights into how AI can stimulate relational benefits of human–technology interactions.

Personalized treatment planning requires various patient-level considerations including personal risk factors and contraindications. However, existing algorithms for facilitating treatment planning frequently fail to account for uncertainties in their recommendations arising from the frequent updating of risk-scoring tools. We propose an algorithmic framework called Data-driven Augmentation of Treatment Planning via Clinical Role Model Generation and Selection (DreAMS). DreAMS integrates risk-scoring tools and data-driven optimization to augment treatment planning by identifying clinical role models, i.e., low-risk patients whose physiological measurements and medications can inform treatment planning for high-risk patients. The problem of optimally generating clinical role models amidst uncertainty in frequently updated risk-scoring tools can be tractably reformulated by leveraging two data sources: (i) a patient-specific database ensuring actionability and (ii) historical data from risk-scoring tools to mitigate risks of erroneously recommending high-risk role models. We develop greedy and active-learning algorithms to solve this problem and derive complexity bounds. We present a case study using multiple datasets containing patients at risk for atherosclerotic cardiovascular disease (ASCVD). DreAMS effectively augments treatment planning for high-risk patients despite frequent updating of ASCVD risk-scoring tools, selecting role models whose predicted ASCVD risk falls within acceptable levels in over 60% of high-risk patients and outperforming benchmarks by over 20%.

Telehealth services became popular due to the COVID-19 pandemic, yet their operational impacts on healthcare organizations are not well-understood. Patient behaviors can vary significantly between telehealth and in-person appointments, introducing new challenges and opportunities for healthcare delivery. We examine two key behaviors contributing to nonadherence to medical appointments: No-shows and unpunctuality. Analyzing 412,415 telehealth and in-person appointments across a major US medical system from 2020 to 2022, we find that telehealth appointments reduce no-shows by 3.0 percentage points (23.1%) and late-arrivals by 11.4 percentage points (35.6%), indicating significant improvements in appointment adherence. We also find that telehealth is particularly effective in improving adherence to follow-up appointments but may be less suitable for initial consultations with new patients. In addition, telehealth improves adherence most among demographic groups with historically lower in-person attendance—women, racial minorities, Medicaid patients, and younger adults—underscoring its potential to reduce disparities in access. Our analysis suggests that while telehealth may increase patient revisit rates and create extra work for providers, the gains from reduced no-shows, particularly for follow-ups, lead to a net boost in provider productivity. Finally, we explore the best strategies to integrate telehealth into a provider’s daily scheduling template, showing that scheduling telehealth appointments before in-person visits enhances operational efficiency compared to the opposite sequence. Policymakers should recognize telehealth’s capacity to improve appointment adherence, reduce disparities, and enhance productivity, and support its adoption through appropriate regulations. Healthcare organizations should strategically deploy telehealth to address the root causes of patient nonadherence. By offering telehealth appointments to patients facing barriers to in-person care, they can simultaneously optimize both access and productivity.

In this paper, we study contextual stochastic optimization (CSO), where decisions are made under uncertainty and the distribution of random parameters can be partially inferred from covariates observed prior to decision-making. In many practical settings, these distributions also depend on the decisions themselves, a phenomenon known as the decision-dependent effect . Most existing studies address this issue by imposing structural assumptions on the relationship between decisions and the underlying distributions. However, such assumptions may lead to model misspecification when the true relationship deviates from the assumed form. A prominent alternative is the weighted sample average approximation (wSAA) method proposed by Bertsimas and Kallus (2019), which adapts sample weights based on their similarity to the current decision–context pair. Nevertheless, because these weights are typically computed using complex machine learning models and depend on the decision variables in decision-dependent settings, solving the resulting optimization problem becomes computationally challenging. To overcome this challenge, we extend the wSAA framework from the loss function to its gradient, leading to the notion of the contextual gradient . We show that the contextual gradient serves as a meaningful indicator of optimality and leverage this property to develop the contextual gradient descent (CGD) algorithm. Our analysis establishes that CGD converges to a neighborhood of the global optimum when the loss function exhibits sufficient strong convexity. Moreover, the derived bounds reveal a key insight: the strength of convexity in the loss function can compensate for the uncertainty introduced by decision-dependent effects. Extensive numerical experiments on both synthetic and real-world datasets demonstrate that CGD consistently outperforms existing methods for contextual optimization under decision-dependent uncertainty.

We investigate the fleet repositioning problem aimed at dynamically optimizing vehicle distributions to maximize long-run average social welfare in a vehicle-sharing system. We model the problem as a Markov decision process under the ex ante committed decision scheme, characterizing the balanced myopic policy as optimal for the average reward setting. This policy efficiently aligns vehicle supply with trip demand and mitigates the curse of dimensionality, enhancing computational efficiency significantly. Our analysis demonstrates that although the balanced myopic policy operates with less information, potentially leading to performance losses, the maximum performance gap relative to the ex post decision scheme asymptotically converges to zero as the system size increases. This finding underscores the asymptotic optimality of the balanced myopic policy, particularly in large systems, making it a robust and effective solution for fleet repositioning. Moreover, we extend our investigation to settings with seasonal demand, confirming that a generalized balanced myopic policy remains optimal. Through comprehensive numerical experiments and a counterfactual case study of a real-world vehicle-sharing system, we quantify the operational value of our approach. This study not only validates the balanced myopic policy against more information-intensive solutions but also illuminates effective heuristic design strategies for improving the efficiency of fleet repositioning in vehicle sharing systems.