Knowledge that Transforms

Problem definition: This paper considers the operations management problems under a newly proposed choice model referred to as a focal multinomial logit (FMNL) model. It generalizes the famous multinomial logit (MNL) model and various well-studied consideration-set choice models and can effectively capture irrational choice behaviors such as the context effect, halo effect, and choice overload, as well as the effect of focality. Methodology/results: We focus on the threshold focal set and various focal parameter settings, including the constant, cardinality, and linear threshold FMNL models, as well as a broader model that satisfies certain regularity conditions and subsumes the above models. We analyze the computational complexity and propose polynomial-time exact or approximation algorithms for assortment optimization problems under different focal parameters. We then characterize the optimal strategy for the joint price and assortment optimization problem. Our investigation into the statistical properties of maximum-likelihood estimators addresses identifiability, consistency, and convergence rates, as well as their implications on operations decisions. We also present a convex mixed-integer nonlinear programming reformulation method that achieves a global optimal estimator for model calibration. Managerial implications: Through extensive numerical experiments on synthetic and real data sets, we demonstrate the efficiency of the proposed algorithms, highlight the issues of model misspecification, and reveal revenue improvement under the family of FMNL models. Our analyses suggest that retailers should consider the impact of focality to potentially improve demand estimation accuracy and operations performance.

Problem definition: This work investigates the effect of technological monitoring on salesforce efficacy. While prior research has investigated the creation of incentive packages to resolve the principal-agent problem, the efficacy of increasingly inexpensive and widely available technological monitoring has received less attention. We argue that the effect of such programs on salesperson performance is ambiguous. On the one hand, such programs may increase performance by increasing transparency between the principal and agent. On the other hand, such programs may create crowd out effects that erode performance. Methodology/results: We investigate these questions using a unique Chinese B2B platform specializing in products for infants. To identify the effect of monitoring on performance, we exploit the rollout of a de novo cell phone GPS monitoring program using a difference-in-difference approach. Results indicate that implementing a monitoring system significantly increases salesperson performance. However, results also suggest considerable heterogeneity across salespersons, with underperformers experiencing significant improvements but negative effects accruing to high performers. Empirical extensions suggest that the beneficial effect for underperformers is primarily due to effort (as opposed to upskilling), while the deleterious effect for high performers results from the disruption of established routines, leading to suboptimal effort allocation across their assigned tasks and store portfolio. Managerial implications: These findings shed light on the benefits and drawbacks of monitoring programs and the need for managers to be cognizant of both the reasons for salesforce underperformance (i.e., salesperson capability versus effort) and whom they are electing to monitor (high versus low performers). Further, these findings provide cautionary evidence against the wholesale implementation of monitoring, as it may disrupt well-established and successful routines. Theoretically, this work integrates the hitherto disparate concepts of compensation design and the effects of technology-enabled monitoring. Finally, by stratifying salespersons based on their past performance, we illustrate boundary conditions between transparency and crowd out effects and provide insights into the mechanisms behind performance changes.

Problem definition: Multistage stochastic programs involving mixed-integer state variables and continuous local variables (MSILPs) present a challenging class of optimization problems with limited techniques available to obtain high-quality solutions efficiently. These problems arise in many practical applications, including disaster relief logistics planning for natural disasters such as hurricanes, which is increasingly important due to its significant societal impacts. Methodology/results: We introduce an aggregation framework to address MSILPs that impose additional structure on the integer state variables by leveraging information from the underlying stochastic process, which is modeled as a Markov chain (MC). We demonstrate that the aggregated MSILP can be solved exactly via a branch-and-cut algorithm integrated with a variant of the stochastic dual dynamic programming method. To improve tractability, we use this approach to obtain dual bounds. To obtain high-quality decision policies with significantly reduced computational effort, we apply two-stage linear decision rule (2SLDR) approximations, including a new MC-based variant that we propose. Managerial implications: We test the proposed methodologies on an MSILP arising from hurricane disaster relief logistics planning. Our empirical evaluation compares the effectiveness of the various proposed approaches and analyzes the tradeoffs between policy flexibility, solution quality, and computational effort. Specifically, the 2SLDR approximation yields provable high-quality solutions for our test instances, supported by the proposed bounding procedure. We also extract valuable managerial insights from the solution behaviors exhibited by the underlying decision policies: (i) adaptive planning is most valuable in constrained systems; (ii) effective policies incorporate trend indicators to anticipate future needs; and (iii) adaptability can capture the majority of performance gains over static approaches.

Problem Definition: This study examines a seller’s joint information disclosure and pricing decisions when launching a new product with an easy-to-communicate objective attribute and a hard-to-describe subjective attribute. We analyze how consumer search cost, inventory level, and channel structure shape the seller’s optimal information strategy. Results: For brick-and-mortar retailing, the seller should conceal (disclose) attribute information when the search cost is low (high). Interestingly, when the inventory level is high, the seller conceals information to retain the “diamond-in-the-rough” consumers, aiming to match the abundant inventory with high demand. Conversely, when inventory is low, the seller discloses information to dissuade some consumers with low objective valuations from searching, thereby reducing demand and alleviating consumers’ concerns about product availability. Thus, the seller strategically uses attribute information to balance search and stock. This result remains robust under extensions such as partial disclosure, atomic consumers, and efficient allocation. Moreover, intensified demand-side competition (e.g., a larger consumer base) can unexpectedly reduce the seller’s revenue because of price markdown pressures to ease consumers’ concerns about competition. In an omnichannel setting, where an online channel supplements physical retail, a high inventory level may reverse the optimal information policy. Managerial Implications: Our findings offer insights into how demand-side factors (e.g., search cost) and supply-side factors (e.g., inventory level and channel structure) shape optimal disclosure. We also reveal risks associated with demand-side competition. Funding: X. Fu acknowledges financial support from the University of New South Wales [UNSW Business School Dean’s Research Fellowship]. P. Gao’s research is supported by the National Natural Science Foundation of China [Grants 72522026, 72201234, 72192805], the Collaborative Research Funding Hong Kong [Grant C6032-21G], the Guangdong Provincial Key Laboratory of Mathematical Foundations for Artificial Intelligence [Grant 2023B1212010001], and industry collaborators including Meituan and Fengyi Technology. Y.-J. Chen acknowledges financial support from the Hong Kong Research Grants Council [Grants 16501722, 16212821, and C6020-21GF].

Problem definition: We propose a feature-based consideration set choice model (FCM) motivated by customers’ purchasing behavior in online retail platforms. In our model, customers form a consideration set by including products with the highest utilities computed based on a subset of product features visible on the search results page. The customers then make a purchase decision from the consideration set by accounting for all features available on the product pages. The FCM incorporates heterogeneity in customer preferences across both the consideration and purchase stages, and it allows customers to re-evaluate products’ utility in the second stage. Methodology/results: We develop an efficient maximum likelihood estimation procedure for estimating the model parameters from customers’ click and purchase data. We show that the assortment optimization problem under FCM is NP-hard by drawing a connection to the assortment problem under the latent-class multinomial logit (LC-MNL) model. Moreover, we establish that the problem remains NP-hard even under a variant of our model termed the deterministic feature-based consideration set choice model (D-FCM), which imposes a deterministic structure in the consideration stage. On the positive side, the D-FCM admits a tractable mixed integer linear programming (MILP) formulation for solving the assortment problem and facilitates the study of the more complex joint assortment and pricing problem, for which we provide a polynomial-time solution (in the number of products) for both homogeneous and heterogeneous settings. Managerial implications: Through numerical experiments on real and synthetic data, we demonstrate that our proposed model outperforms the MNL and LC-MNL benchmarks on out-of-sample prediction accuracy and decision performance. Finally, we introduce a novel operational problem termed feature selection, which identifies the subset of features to display during the consideration formation stage to maximize expected revenue. We establish the NP-hardness of this problem under both model variants and propose a tractable MILP formulation for solving it.

Problem definition: How should (apparel) retailers manage product sizes? For example, if most customers wearing a given shoe size, such as 9.5, are willing to accept a half-size up or down, is it necessary for a retailer to carry that size at all? Additionally, although identical products in different sizes are treated as distinct stock-keeping units in inventory management, they are often aggregated for assortment and strategic planning. However, there is no theoretical justification for this approach. In this paper, we address the fundamental questions about size management that have remained largely unexplored in the operations literature. Methodology/results: We propose a choice model where each customer forms a consideration set based on the in-stock availability of products of her best-fit size and adjacent sizes. Using a real-world data set from a large footwear retailer, we show that nearly 25% of the unmet demand caused by stockouts spills over to adjacent sizes. We further solve the assortment and inventory optimization problems under the proposed choice model. Our findings demonstrate that the optimal assortment remains unchanged, regardless of the likelihood that customers might purchase adjacent sizes. We utilize this finding and further show that inventory policies that ignore size substitution can be (asymptotically) optimal when the demand rate is high or the selling horizon is long. We also propose a mixed-integer program to determine inventory levels that account for size substitution and achieve higher profits in low-demand settings. Managerial implications: We show that the prevalent size-aggregation approach employed in apparel retail operations is sensible in high-demand settings, such as e-commerce. In contrast, when the expected demand over the selling horizon is low, size substitution can be relevant and should be considered in stocking decisions.

Problem definition: The transition from fossil-fuel generators to renewable energy requires significant growth of flexible resources to manage weather-dependent output variations. Key among these are large-scale storage assets. Although storage is mostly carbon neutral in its direct operations, its arbitrage activities influence the scheduled quantities of other producers, thereby affecting market-level carbon emissions indirectly. This raises questions about the extent of these indirect emissions and how to limit them effectively. Methodology/results: We develop a model to analyze the emission impact of profit-maximizing large-scale storage agents in a competitive electricity market. We derive a tight condition for the worst-case rate of added emissions from a storage transaction. Accordingly, we characterize the minimum sufficient carbon levy to keep emissions below a desired threshold. We support our theoretical findings with numerical studies based on the Dutch electricity market. Our results show that both emissions and the corresponding carbon levy depend on the round-trip efficiency of the storage asset and the characteristics of technologies in the energy mix (e.g., marginal costs, emissions, and capacities). The findings remain robust under uncertainty in demand and renewables. Managerial implications: The framework developed in this work enables market participants and regulators to assess, interpret, and potentially control the unintended carbon impacts of storage asset operations. Several findings are nontrivial and carry important implications for regulation. For instance, we show that counterintuitively, storage assets with higher round-trip efficiency can increase system emissions more—and require higher carbon levies to curb them—than less efficient ones. Additionally, although a carbon levy reduces the worst-case emission rates of a storage transaction, we identify scenarios where the emission impact of a storage agent may rise with higher levies. Notably, the indirect emissions of storage agents are also sensitive to whether solar or wind is the dominant renewable.

Problem definition: Vaccination campaigns often face significant operational challenges such as limited stockpiles, vaccine delivery delays, and constrained administration capacity. In such contexts, fractional-dose vaccines have been described in the medical literature as a possible strategy because their efficacy reduction is typically not commensurate with the level of fractionation, allowing greater population coverage. We seek to determine the optimal use and potential benefits of a fractionated vaccine dose with lower and more uncertain efficacy, given the specific supply constraints faced by a country. Methodology/results: We employ a susceptible-infected-recovered (SIR) epidemic model integrating vaccination with full and fractional doses over time. We embed it within a deterministic optimal control model aimed at identifying vaccination policies that minimize total infections during an epidemic, given operational constraints restricting the stockpile, delivery rate, and administration of vaccines. Using a statistical approach described in the clinical literature for estimating the uncertainty around fractional-dose efficacy, we conduct two application case studies grounded in real-world scenarios. Our theoretical analysis provides an intuitive characterization of the optimal vaccination policy that, depending on the epidemic and operational parameters, may utilize a combination of full- and fractional-dose vaccines, either simultaneously or sequentially. We also examine simpler policies that employ a single vaccine dosage throughout the epidemic. We conclude that, although these single-dose policies can often be almost as effective as the optimal policy in averting infections, they are not as robust to the uncertainty affecting fractional-dose vaccine efficacy. Managerial implications: Fractional-dose vaccines, used either alone or in conjunction with full-dose vaccines, present an opportunity to significantly reduce infections during an epidemic in resource-constrained settings. The proportion of fractional-dose vaccines relative to full-dose vaccines in a campaign should generally increase with the maximum vaccine administration rate and decrease with the total antigen stockpile available. History: This paper was selected as part of the 1RR initiative between the M&SOM journal and the MSOM Society. This particular paper was part of the 2024 MSOM Service Operations SIG Conference.

Problem definition: We examine how market and economic factors influence the occurrence of forced labor in supply chains and how buying companies can develop optimal contracts to prevent forced labor in the presence of information asymmetry between the buyer and the agent. Methodology/results: We develop a game-theoretic model of a labor supply chain comprising a socially aware buyer and a profit-maximizing labor agent. Our equilibrium analysis shows that the audit cost affects the extent to which the buyer can extract surplus from the agent. In the asymmetric information case, we design a menu of contracts and show that the difference in the agent’s earnings dictates how an unconstrained optimal contract can be adjusted to be incentive compatible. Our result suggests that it is optimal for the buyer to leave a surplus to agents with high recruitment capability (measured in terms of the labor pool size) regardless of the audit cost. As we extend our analysis to the multiagent case, we develop a “sequential” menu of contracts that ensures no coercion and maximum buyer profit. We apply our model to a data set of labor agents for recruiting foreign agricultural workers in the United States. Managerial implications: To implement incentive-compatible contracts that deter coercive labor outcomes, the buyer may need to allocate informational rents to the agent by foregoing a portion of its surplus. We find that information asymmetry regarding the agent’s true recruitment capability necessitates that the buyer offers a “menu of contracts” to prevent the risk of forced labor. This menu benefits exactly one type of agent, depending on the earnings differential between the available contracts, but always leaves a financial surplus for agents with strong recruitment capabilities. When multiple potential agents are available, the buyer can employ a reverse auction mechanism to select one agent. Though the core insights from contracting with a single agent continue to apply, their impact diminishes as the number of potential agents grows.

Problem definition: We consider the joint optimization of ordering and upgrading decisions in a dynamic multiproduct system over a finite horizon of T periods. In each period, multiple types of demand arrive stochastically and can be satisfied either with supply of the same type or by upgrading to a higher-quality product. The goal is to find an optimal joint replenishment and allocation policy that maximizes total expected profit, both when the firm knows the demand distributions a priori and when the firm must learn them over time. Methodology/results: We first characterize the structure of the clairvoyant optimal joint ordering and allocation policy. Building on this structure, we propose a new online learning algorithm, termed stochastic subgradient descent with perturbed subgradient (SGD-PG for short), and show that it achieves cumulative regret growing on the order of the square root of T, which matches the known lower bound for any online learning method. We further show that SGD-PG can be extended to a nested censored demand setting. In the course of the algorithmic design, we propose a linear programming (LP)-based approach to compute the subgradient and prove that it produces the same output as the perturbed subgradient method. The LP-based method also allows us to extend the results to general upgrading structures. We demonstrate the efficacy of the proposed algorithms in numerical experiments. Managerial implications: This work provides practitioners with the optimal policy for inventory replenishment and allocation in a multiproduct system with upgrading. When the demand distribution is unknown, we propose an easy-to-implement and provably good algorithm for demand learning. In addition, our numerical results quantify the value of optimal upgrading and identify the conditions under which upgrading is most beneficial.