Knowledge that Transforms

Problem definition: We consider a revenue management problem that arises from the selling of high-speed train tickets in China. Compared with traditional network revenue management problems, the new feature of our problem is the assign-to-seat restriction. That is, each request, if accepted, must be assigned instantly to a single seat throughout the whole journey, and later adjustment is not allowed. When making decisions, the seller needs to track not only the total seat capacity available, but also the status of each seat. Methodology/results: We build a modified network revenue management model for this problem. First, we study a static problem in which all requests are given. Although the problem is NP-hard in general, we identify conditions for solvability in polynomial time and propose efficient approximation algorithms for general cases. We then introduce a bid-price control policy based on a novel maximal sequence principle. This policy accommodates nonlinearity in bid prices and, as a result, yields a more accurate approximation of the value function than a traditional bid-price control policy does. Finally, we combine a dynamic view of the maximal sequence with the static solution of a primal problem to propose a “re-solving a dynamic primal” policy that can achieve uniformly bounded revenue loss under mild assumptions. Numerical experiments using both synthetic and real data document the advantage of our proposed policies on resource-allocation efficiency. Managerial implications: The results of this study reveal connections between our problem and traditional network revenue management problems. Particularly, we demonstrate that by adaptively using our proposed methods, the impact of the assign-to-seat restriction becomes limited both in theory and practice.

Problem definition: Probabilistic selling (PS) is a business model whereby, in addition to selling transparent products, a firm sells an opaque good, which is unknown to buyers until after purchase. We examine how PS affects retailer-manufacturer interactions in markets for physical goods and how upstream competition impacts channel members’ incentives to facilitate PS. Methodology/results: Using a Hotelling-based model of a multiproduct retailer, we find that a retailer maximizes its profit by assigning equal probability to each product even when the products have different wholesale prices. We also find that PS mitigates the inefficiencies caused by the double-marginalization problem. Although the potential benefit from PS is greater for a decentralized channel than for a centralized one, the market conditions for which PS arises are narrower for a decentralized channel. Furthermore, PS shifts channel power toward the manufacturer. However, it is possible for a win-win-win outcome to arise in which the manufacturer, retailer, and consumers benefit from PS. As expected, upstream competition shifts channel profit toward the retailer. However, competition also has surprising effects: It shrinks, rather than expands, the viability of PS and makes it possible for each manufacturer to benefit from its rival’s cost reduction. Managerial implications: A manufacturer should induce the retailer to offer an opaque good if its production costs are sufficiently low and the products are sufficiently close substitutes. It is optimal for the manufacturer to increase (decrease) its wholesale prices in response to the retailer’s ability to offer opaque goods if product differentiation is low (modest). Setting a wholesale price below cost sometimes maximizes a manufacturer’s profit. Furthermore, a retailer can achieve a strategic advantage by using products from multiple manufacturers to construct opaque goods. However, the retailer’s ability to leverage this advantage is curtailed because its use of equal-probability assignments relaxes competition between manufacturers.

Problem definition: We study a procurement problem, where the supplier holds superior cost information and can learn to improve efficiency over time. Despite its prevalence, the supply chain literature provides limited guidance on how to manage learning suppliers with evolving private information. Methodology/results: We use mechanism design. We show that supplier learning has both efficiency and agency effects, it can induce countervailing incentives, and the agency effect can overwhelm the efficiency effect. As a result, (i) supplier learning can hurt profits, (ii) information asymmetry can improve efficiency, (iii) production distortion can go upward, and (iv) ignoring the agency effect of learning can mislead contract design and inflict severe losses. Managerial implications: Our results suggest that previous studies may have overlooked the downside of learning and overestimated the harm of information asymmetry. Moreover, our results help explain when and why firms should overproduce output and disclose private information voluntarily. By highlighting the strategic role of supplier learning, this study sharpens our understanding of supply chain management.

Problem definition: We study the dynamic fulfillment problem in e-commerce, in which incoming (multi-item) customer orders must be immediately dispatched to (a combination of) fulfillment centers that have the required inventory. Methodology/results: A prevailing approach to this problem, pioneered by Jasin and Sinha in 2015 , has been to write a “deterministic” linear program that dictates, for each item in an incoming multi-item order from a particular region, how frequently it should be dispatched to each fulfillment center (FC). However, dispatching items in a way that satisfies these frequency constraints, without splitting the order across too many FCs, is challenging. Jasin and Sinha in 2015 identified this as a correlated rounding problem and proposed an intricate rounding scheme that they proved was suboptimal by a factor of at most [Formula: see text] on a q-item order. This paper provides, to our knowledge, the first substantially improved scheme for this correlated rounding problem, which is suboptimal by a factor of at most [Formula: see text]. We provide another scheme for sparse networks, which is suboptimal by a factor of at most d if each item is stored in at most d FCs. We show both of these guarantees to be tight in terms of the dependence on q or d. Our schemes are simple and fast, based on an intuitive idea; items wait for FCs to “open” at random times but observe them on “dilated” time scales. This also implies a new randomized rounding method for the classical Set Cover problem, which could be of general interest. Managerial implications: We numerically test our new rounding schemes under the same realistic setups as Jasin and Sinha and find that they improve runtimes, shorten code, and robustly improve performance. Our code is made publicly available online. History: This paper was selected for Fast Track in the M&SOM Journal from the 2022 MSOM Supply Chain Management SIG Conference.

Problem definition: We consider a periodic-review dual-sourcing inventory system with a regular source (lower unit cost but longer lead time) and an expedited source (shorter lead time but higher unit cost) under carried-over supply and backlogged demand. Unlike existing literature, we assume that the firm does not have access to the demand distribution a priori and relies solely on past demand realizations. Even with complete information on the demand distribution, it is well known in the literature that the optimal inventory replenishment policy is complex and state dependent. Therefore, we focus our attention on a class of popular, easy-to-implement, and near-optimal heuristic policies called the dual-index policy. Methodology/results: The performance measure is the regret, defined as the cost difference of any feasible learning algorithm against the full-information optimal dual-index policy. We develop a nonparametric online learning algorithm that admits a regret upper bound of [Formula: see text], which matches the regret lower bound for any feasible learning algorithms up to a logarithmic factor. Our algorithm integrates stochastic bandits and sample average approximation techniques in an innovative way. As part of our regret analysis, we explicitly prove that the underlying Markov chain is ergodic and converges to its steady state exponentially fast via coupling arguments, which could be of independent interest. Managerial implications: Our work provides practitioners with an easy-to-implement, robust, and provably good online decision support system for managing a dual-sourcing inventory system.

Problem definition: Large cities around the globe are facing an alarming growth in traffic congestion and greenhouse gas emissions, to which a significant contributor in recent years are on-demand cabs operated by ride-hailing platforms. Newly emerged pooled transportation options like shuttle services are cheaper and greener alternatives. However, those alternatives are still new to many customers and policy makers. The design of their promotion policies demands careful investigation. This paper studies how we can reduce the number of on-demand cabs on the road and, therefore, their GHG emissions by promoting pooled transportation such as shuttle services. Methodology/Results: In this work, we use detailed usage data and build a structural model to study customer preferences of price and service features when choosing between private cabs and a scheduled shuttle service. Using the estimated model, we identify and evaluate the efficacy of improving service features like reducing the walking distance to shuttle stops on customers’ choices of transport and, therefore, the number of ride-hailing vehicles on the road. We find that a 20% decrease in walking distance can achieve 40% of the benefits of commonly adopted congestion surcharge policies. It can also reduce up to 4.8 thousand tonnes of GHG emissions, which is worth over a million dollars per year. In addition, we demonstrate the implementability of walking distance reduction policies by adding stops on existing shuttle routes. Managerial implications: Reducing the number of ride-hailing vehicles on the road has become an important goal in many cities’ green transport policy design. For example, cities like New York have implemented congestion surcharge policies targeting ride-hailing vehicles in recent years. Our findings suggest that, by changing operations levers such as service features of pooled transport, cities can achieve a substantial amount of benefits from reducing congestion compared with congestion surcharge policies with essentially zero cost, leading to much more efficient green transport policies.

Problem definition: Quick response is a classic operations strategy that allows a retailer to place a rapid replenishment order during the selling season using information learned from early sales. The benefits of quick response are widely studied in the literature under the condition that the manufacturer’s wholesale prices are exogenously given. Motivated by the practice of emerging small and medium-sized enterprise (SME) fashion brands, this paper revisits the value of quick response for a retailer when a manufacturer can strategically set its wholesale prices. Methodology/results: We develop a game-theoretic model consisting of one manufacturer and one retailer. In contrast to the traditional quick response setting, the manufacturer can dynamically adjust wholesale prices for both regular and replenishment orders. First, we investigate whether and when quick response still benefits the retailer. We find that, under low or significantly high demand uncertainties, the firms share a common preferred ordering strategy, and quick response benefits the retailer as well as the supply chain. But, under moderately high demand uncertainty, the retailer’s favored ordering strategy conflicts with the manufacturer’s interest; as a result, the manufacturer would set wholesale prices to counter the retailer’s ordering strategy, which makes quick response detrimental to the retailer. Second, we search for mechanisms that can resolve this conflict and restore the beneficial effect of quick response. We show that letting the manufacturer commit to wholesale prices up front is ineffective in fixing the problem. However, if the retailer can propose a take-it-or-leave-it wholesale price for the replenishment order (possibly with the replenishment quantity) once the regular wholesale price is set, then quick response leads to a win–win outcome for both firms. Managerial implications: The findings caution retailers with weak power (e.g., SMEs) when adopting quick response, especially when facing moderately high demand uncertainties. The retailer, although weak, should be aware of the retailer’s natural ability to propose replenishment terms because, otherwise, the retailer can always forgo quick response; this opens up an opportunity to design more favorable arrangements.

Problem definition: U.S. nonprofits declare three types of expenses in their IRS 990 forms: program spending to meet beneficiaries’ needs; fundraising spending to raise donations; and administration spending to build and maintain capacity. Charity watchdogs, however, expect nonprofits to prioritize program spending over other categories. We study when such expectations may lead to the “starvation cycle” or underspending on administration and fundraising. Methodology/results: We characterize optimal budget allocations to program, fundraising, and administration spending categories using a two-period model, which also includes the nonprofit’s capacity, return on program spending (the net value of program spending to beneficiaries), and beneficiaries’ uncertain future needs. We find that the nonprofit’s capacity plays a significant role in the optimal allocation. The nonprofit should (a) at high capacity, spend only the necessary amount on administration to maintain its current capacity; (b) at moderate capacity, maintain its current capacity while limiting program spending in favor of fundraising; and (c) at low capacity, increase administration spending to expand its future capacity. When we compare the optimal allocations prescribed by our model to the actual spending levels reported by a foodbank network, we find that the foodbank underspends on administration and fundraising, suggesting the forces that lead to the starvation cycle may be in play. Another possibility is that the nonprofit’s own estimate of its return on program spending is higher than our estimate—At higher estimates of return on program, the gap between our prescribed solutions versus actual spending levels decreases. Managerial implications: Our paper introduces an important discussion on nonprofits’ starvation cycle and finds conditions that justify prioritizing administration and fundraising expenses. It also highlights that watchdogs should consider nonprofits’ return on program spending in addition to their capacity and future needs when evaluating them.

Problem definition: Designing effective operational strategies requires a good understanding of customer behavior. The classic economic theory of customer choice has long been the paradigm in the operations literature. However, the rise of online marketplaces such as e-commerce has triggered considerable efforts in academia and industry to develop alternative models that not only provide a good approximation of customer behavior but also are easily scalable for large-scale implementations. In this paper, we consider a multiproduct dynamic pricing problem with limited inventories under the so-called cascade click model, which is one of the most popular click models used in practice and has been intensively studied in the computer science literature. Methodology/results: We present some fundamental results. First, we derive a sufficiently general characterization of the optimal pricing policy and show that it has a different structure than the optimal policy under the standard pricing model. Second, we show that the optimal expected total revenue under the cascade click model can be upper bounded by the objective value of an approximate deterministic pricing problem. Third, we show that two policies that are known to have strong performance guarantees in the standard revenue management setting can be properly adapted (in a nontrivial way) to the setting with cascade click model while retaining their strong performance. Finally, we also briefly discuss the joint ranking and pricing problem and provide an iterative heuristic to calculate an approximate ranking. Managerial implications: Taking into account customers’ click-and-search behavior leads to different structures of the optimal pricing policy, and some common insights under the standard pricing models may no longer hold. Moreover, our simulation studies show that pricing under a (misspecified) classic choice model that is oblivious to customers click-and-search behavior can severely impact profitability.

Problem definition: Last-mile delivery is a critical component of logistics networks, accounting for approximately 30%–35% of costs. As delivery volumes have increased, truck route times have become unsustainably long. To address this issue, many logistics companies, including FedEx and UPS, have resorted to using a “driver aide” to assist with deliveries. The aide can assist the driver in two ways. As a “jumper,” the aide works with the driver in preparing and delivering packages, thus reducing the service time at a given stop. As a “helper,” the aide can independently work at a location delivering packages, and the driver can leave to deliver packages at other locations and then return. Given a set of delivery locations, travel times, service times, jumper’s savings, and helper’s service times, the goal is to determine both the delivery route and the most effective way to use the aide (e.g., sometimes as a jumper and sometimes as a helper) to minimize the total routing time. Methodology/results: We model this problem as an integer program with an exponential number of variables and an exponential number of constraints and propose a branch-cut-and-price approach for solving it. Our computational experiments are based on simulated instances built on real-world data provided by an industrial partner and a data set released by Amazon. The instances based on the Amazon data set show that this novel operation can lead to, on average, a 35.8% reduction in routing time and 22.0% in cost savings. More importantly, our results characterize the conditions under which this novel operation mode can lead to significant savings in terms of both the routing time and cost. Managerial implications: Our computational results show that the driver aide with both jumper and helper modes is most effective when there are denser service regions and when the truck’s speed is higher (≥10 miles per hour). Coupled with an economic analysis, we come up with rules of thumb (that have close to 100% accuracy) to predict whether to use the aide and in which mode. Empirically, we find that the service delivery routes with greater than 50% of the time devoted to delivery (as opposed to driving) are the ones that provide the greatest benefit. These routes are characterized by a high density of delivery locations.