Bounding Optimal Expected Revenues for Assortment Optimization under Mixtures of Multinomial Logits

We consider assortment problems under a mixture of multinomial logit models. There is a fixed revenue associated with each product. There are multiple customer types. Customers of different types choose according to different multinomial logit models whose parameters depend on the type of the customer. The goal is to find a set of products to offer so as to maximize the expected revenue obtained over all customer types. This assortment problem under the multinomial logit model with multiple customer types is NP‐complete. Although there are heuristics to find good assortments, it is difficult to verify the optimality gap of the heuristics. In this study, motivated by the difficulty of finding optimal solutions and verifying the optimality gap of heuristics, we develop an approach to construct an upper bound on the optimal expected revenue. Our approach can quickly provide upper bounds and these upper bounds can be quite tight. In our computational experiments, over a large set of randomly generated problem instances, the upper bounds provided by our approach deviate from the optimal expected revenues by 0.15% on average and by less than one percent in the worst case. By using our upper bounds, we are able to verify the optimality gaps of a greedy heuristic accurately, even when optimal solutions are not available.     

Integrated retail decisions with multiple selling periods and customer segments: Optimization and insights

Integrating retail decisions on such aspects as assortment, pricing, and inventory greatly improves profitability. We examine a multi-period selling horizon where a retailer jointly optimizes assortment planning, pricing, and inventory decisions for a product line of substitutable products, in a market with multiple customer segments. Focusing on fast-moving retail products, the problem is modeled as a mixed-integer nonlinear program where demand is driven by exogenous consumer reservation prices and endogenous assortment and pricing decisions. A mixed-integer linear reformulation is developed, which enables an exact solution to large problem instances (with up to a hundred products) in manageable times. Empirical evidence is provided in support of a classical deterministic maximum-surplus consumer choice model. Computational results and managerial insights are discussed. We find that the optimal assortment and pricing decisions do not exhibit a simple, intuitive structure that could be analytically characterized, which reflects the usefulness of optimization approaches to numerically identify attractive trade-offs for the decision-maker. We also observe that suboptimal inventory policies significantly decrease profitability, which highlights the importance of integrated decision-making. Finally, we find that the seasonality of consumer preferences and supply costs present an opportunity for boosting the profit via higher inventory levels and wider assortments.     

Product Assortment and Price Competition under Multinomial Logit Demand

The role of assortment planning and pricing in shaping sales and profits of retailers is well documented and studied in monopolistic settings. However, such a role remains relatively unexplored in competitive environments. In this study, we study equilibrium behavior of competing retailers in two settings: (i) when prices are exogenously fixed, and retailers compete in assortments only; and (ii) when retailers compete jointly in assortment and prices. For this, we model consumer choice using a multinomial Logit, and assume that each retailer selects products from a predefined set, and faces a display constraint. We show that when the sets of products available to retailers do not overlap, there always exists one equilibrium that Pareto‐dominates all others, and that such an outcome can be reached through an iterative process of best responses. A direct corollary of our results is that competition leads a firm to offer a broader set of products compared to when it is operating as a monopolist, and to broader offerings in the market compared to a centralized planner. When some products are available to all retailers, that is, assortments might overlap, we show that display constraints drive equilibrium existence properties.     

On the Benefits of Assortment‐Based Cooperation Among Independent Producers

Motivated by the challenges small‐ to medium‐size companies face in export‐oriented industries, we consider a competitive market for a set of substitutable products. Depending on the assortment of the firms and the substitution behavior of the customer, either a product is sold to the customer or the sale is lost. We consider the cooperation of independent producers that offer a combined set of products to their customers. Producers use discounted price contracts to manage the exchange of products among themselves. We propose an analytical model that enables us to determine the characteristics of firms and their products that would facilitate a beneficial cooperation. We conclude that a cooperation between symmetric single‐product firms is always beneficial, whereas threshold‐type criteria should be satisfied so that assortment‐based cooperation is beneficial for asymmetric firms. We also show that commonality in product assortments of cooperating firms has adverse effects on the benefit from cooperation. For the most general problem setting, we propose a method to determine the set of firms that should cooperate and set the parameters of the contract among the members of cooperation in such a way that each member of the cooperation is better off. We use a numerical study to draw insights on the conditions for which our cooperation scheme is beneficial in the most general problem setting.     

Tractable Consideration Set Structures for Assortment Optimization and Network Revenue Management

Discrete‐choice models are widely used to model consumer purchase behavior in assortment optimization and revenue management. In many applications, each customer segment is associated with a consideration set that represents the set of products that customers in this segment consider for purchase. The firm has to make a decision on what assortment to offer at each point in time without the ability to identify the customer's segment. A linear program called the Choice‐based Deterministic Linear Program (CDLP) has been proposed to determine these offer sets. Unfortunately, its size grows exponentially in the number of products and it is NP‐hard to solve when the consideration sets of the segments overlap. The Segment‐based Deterministic Concave Program with some additional consistency equalities (SDCP+) is an approximation of CDLP that provides an upper bound on CDLP's optimal objective value. SDCP+ can be solved in a fraction of the time required to solve CDLP and often achieves the same optimal objective value. This raises the question under what conditions can one guarantee equivalence of CDLP and SDCP+. In this study, we obtain a structural result to this end, namely that if the segment consideration sets overlap with a certain tree structure or if they are fully nested, CDLP can be equivalently replaced with SDCP+. We give a number of examples from the literature where this tree structure arises naturally in modeling customer behavior.     

The Role of Slotting Fees in the Coordination of Assortment Decisions

Large numbers of new products introduced annually by manufacturers may strain the relationship between retailers and manufacturers regarding assortments carried by retailers. For example, many retailers in the grocery industry will agree to broaden their assortments only if the manufacturer agrees to pay slotting fees for the new products. We investigate the role played by slotting fees in coordinating the assortment decisions in a supply chain. To do so, we study a single‐retailer, single‐manufacturer supply chain, where the retailer decides what assortment to offer to end customers. Double marginalization results in a discrepancy between the retailer's optimal assortment and the assortment that maximizes total supply chain profits. We consider a payment scheme that is analogous to slotting fees used in the grocery industry: the manufacturer pays the retailer a per‐product fee for every product offered by the retailer in excess of a certain target level. We show that, if the wholesale price is below some threshold level, this payment scheme induces the retailer to offer the supply‐chain‐optimal assortment and makes both parties better off.     

Revenue Management with End‐of‐Period Discounts in the Presence of Customer Learning

Consider a firm that sells identical products over a series of selling periods (e.g., weekly all‐inclusive vacations at the same resort). To stimulate demand and enhance revenue, in some periods, the firm may choose to offer a part of its available inventory at a discount. As customers learn to expect such discounts, a fraction may wait rather than purchase at a regular price. A problem the firm faces is how to incorporate this waiting and learning into its revenue management decisions. To address this problem we summarize two types of learning behaviors and propose a general model that allows for both stochastic consumer demand and stochastic waiting. For the case with two customer classes, we develop a novel solution approach to the resulting dynamic program. We then examine two simplified models, where either the demand or the waiting behavior are deterministic, and present the solution in a closed form. We extend the model to incorporate three customer classes and discuss the effects of overselling the capacity and bumping customers. Through numerical simulations we study the value of offering end‐of‐period deals optimally and analyze how this value changes under different consumer behavior and demand scenarios.     

消费者导向类型与电商全渠道决策研究

从现实出发,基于不同消费者导向类型研究了电商的全渠道决策问题。首先建立电商单渠道销售的基础模型,并得出最优产品组合、价格和交付时间。在此基础上,给出了消费者为产品导向型和渠道导向型情况下电商采取全渠道决策的条件,得出双渠道产品组合、价格及交付时间的最优决策;并进行了数值验证。研究发现,同一渠道提供的产品其最优价格相同,最优交付时间只和消费者耐心程度及交付成本函数有关。若消费者为产品导向型,仅当双渠道运营成本之差较小且消费者耐心程度较低时,电商进入线下渠道才有利可图,且在线下渠道提供最受欢迎的产品,线上渠道提供剩余产品。若消费者为渠道导向型,电商进入线下渠道必然有利可图,且在线下渠道提供最受欢迎的产品,线上渠道提供所有产品。     

Optimal pricing and production decisions in the presence of symmetrical and asymmetrical substitution

Firms may produce a variety of generally similar products or may practice “scientific pricing” or revenue management where the firm will offer similar or somewhat differentiated products in multiple market segments at different prices. Whenever generally similar products are available, the demand for the products is linked through the ability of the customer to substitute one product for another. One widely known type of demand substitution is referred to as inventory-driven substitution where a customer will substitute for a product that is out of stock by buying a similar product. A second type of substitution occurs as a response to price-differences when a customer substitutes a less expensive product for a similar higher priced product.     

On the Use of Buy Up as a Model of Customer Choice in Revenue Management

We consider settings in which a revenue manager controls bookings over a sequence of flights. The revenue manager uses a buy‐up model to select booking limits and updates estimates of the model parameters as data are accumulated. The buy‐up model we consider is based upon a simple model of customer choice, wherein each low‐fare customer who is not able to purchase a low‐fare ticket will, with a fixed probability, “buy up” to the high fare, independent of everything else. We analyze the evolution of the parameter estimates (e.g., the buy‐up probability) and chosen booking limits in situations where the buy‐up model is misspecified, that is, in situations where there is no setting of its parameters for which its objective function gives an accurate representation of expected revenue as a function of the booking limit. The analysis is motivated by the common situation in which a revenue manager does not know precisely how customers behave but nevertheless uses a parametric model to make decisions. Under some assumptions, we prove that the booking limits and parameter estimates converge and we compare the actual expected revenue at the limiting values with that associated with the booking limits that would be chosen if the revenue manager knew the actual behavior of customers. The analysis shows that the buy‐up model often works reasonably well even when it is misspecified, and also reveals the importance of understanding how parameter estimates of misspecified models vary as functions of decisions.     

Revenue Management vs. Newsvendor Decisions: Does Behavioral Response Mirror Normative Equivalence?

We study and compare decision‐making behavior under the newsvendor and the two‐class revenue management models, in an experimental setting. We observe that, under both problems, decision makers deviate significantly from normative benchmarks. Furthermore, revenue management decisions are consistently higher compared to the newsvendor order quantities. In the face of increasing demand variability, revenue managers increase allocations; this behavior is consistent with normative patterns when the ratio of the selling prices of the two customer segments is less than 1/2, but is its exact opposite when this ratio is greater than 1/2. Newsvendors' behavior with respect to changing demand variability, on the other hand, is consistent with normative trends. We also observe that losses due to leftovers weigh more in newsvendor decisions compared to the revenue management model; we argue that overage cost is more salient in the newsvendor problem because it is perceived as a direct loss, and propose this as the driver of the differences in behavior observed under the two problems.     

Randomization Approaches for Network Revenue Management with Customer Choice Behavior

In this study, we present new approximation methods for the network revenue management problem with customer choice behavior. Our methods are sampling‐based and so can handle fairly general customer choice models. The starting point for our methods is a dynamic program that allows randomization. An attractive feature of this dynamic program is that the size of its action space is linear in the number of itineraries, as opposed to exponential. It turns out that this dynamic program has a structure that is similar to the dynamic program for the network revenue management problem under the so called independent demand setting. Our approximation methods exploit this similarity and build on ideas developed for the independent demand setting. We present two approximation methods. The first one is based on relaxing the flight leg capacity constraints using Lagrange multipliers, whereas the second method involves solving a perfect hindsight relaxation problem. We show that both methods yield upper bounds on the optimal expected total revenue. Computational experiments demonstrate the tractability of our methods and indicate that they can generate tighter upper bounds and higher expected revenues when compared with the standard deterministic linear program that appears in the literature.     

A Single‐Period Assortment Optimization Model

This paper studies a single‐period assortment optimization problem with unequal cost parameters. The consumer choice process is characterized by a Multinomial Logit (MNL) model. When the store traffic is a continuous random variable, we explicitly derive the structure of the optimal assortment. Our approach is to use a comprehensive measure–profit rate to evaluate the profitability of each variant and then determine which product should be offered. We demonstrate that the optimal assortment contains the few items that have the highest profit rate. When the store traffic is discrete, the optimal solution is difficult to obtain. We propose a “profit rate” heuristic, which is inspired by the result for the case of continuous store traffic. In a special case with equal cost parameters and normal demand distribution, the profit rate heuristic is indeed optimal. Using randomly generated data, we test the effectiveness of the heuristic and find that the average percentage error is less than 0.1% and that the hit rate is above 90%. Our research provides managerial insights on assortment planning and accentuates the importance of measuring the profitability of each product when the demand is random and cannibalization among different products exists.     

Quality‐Speed Competition in Customer‐Intensive Services with Boundedly Rational Customers

We consider a system in which two competing servers provide customer‐intensive services and the service reward is affected by the length of service time. The customers are boundedly rational and choose their service providers according to a logit model. We demonstrate that the service provider revenue function is unimodal in the service rate, its decision variable, and show that the service rate competition has a unique and stable equilibrium. We then study the price decision under three scenarios with the price determined by a revenue‐maximizing firm, a welfare‐maximizing social planner, or two servers in competition. We find that the socially optimal price, subject to the requirement that the customer actual utility must be non‐negative, is always lower than the competition equilibrium price which, in turn, is lower than the revenue‐maximizing monopoly price. However, if the customer actual utility is allowed to be negative in social optimization, the socially optimal price can be higher than the other two prices in a large market.     

Increasing the Revenue of Self‐Storage Warehouses by Facility Design

Self‐storage is a booming industry. Both private customers and companies can rent temporary space from such facilities. The design of self‐storage warehouses differs from other facility designs in its focus on revenue maximization. A major question is how to design self‐storage facilities to fit market segments and accommodate volatile demand to maximize revenue. Customers that cannot be accommodated with a space size of their choice can be either rejected or upscaled to a larger space. Based on data of 54 warehouses in America, Europe, and Asia, we propose a new facility design approach with models for three different cases: an overflow customer rejection model and two models with customer upscale possibilities, one with reservation and another without reservation. We solve the models for several real warehouse cases, and our results show that the existing self‐storage warehouses can be redesigned to generate larger revenues for all cases. Finally, we show that the upscaling policy without reservation generally outperforms the upscaling policy with reservation.     

Improving Profits by Bundling Vertically Differentiated Products

We consider a firm managing a category of vertically differentiated goods, that is, products which differ with respect to an attribute for which all consumers prefer more to less. The goods can be sold individually, in which case they are referred to as components, or in bundles. The firm chooses the assortment of components and bundles and their selling prices to maximize profit. We show that each bundling strategy (pure components, pure bundling or mixed bundling) can be optimal and obtain closed‐form expressions for the optimal selling prices. We provide insights on the structure of the optimal assortment and prices. In particular, we show that, when consumers benefit from consuming the components jointly, the products in the optimal assortment form nested sets. When consumers do not benefit from the joint consumption of components, the bundles should be offered at a positive discount. We find that bundling vertically differentiated products can significantly improve profits, even if consumers do not benefit from consuming the components jointly. The value of bundling comes from increased sales: a firm, which understands that its customers may buy multiple types of components, offers bundles of components, incentivizing customers to buy more.     

A New Dynamic Programming Decomposition Method for the Network Revenue Management Problem with Customer Choice Behavior

In this paper, we propose a new dynamic programming decomposition method for the network revenue management problem with customer choice behavior. The fundamental idea behind our dynamic programming decomposition method is to allocate the revenue associated with an itinerary among the different flight legs and to solve a single‐leg revenue management problem for each flight leg in the airline network. The novel aspect of our approach is that it chooses the revenue allocations by solving an auxiliary optimization problem that takes the probabilistic nature of the customer choices into consideration. We compare our approach with two standard benchmark methods. The first benchmark method uses a deterministic linear programming formulation. The second benchmark method is a dynamic programming decomposition idea that is similar to our approach, but it chooses the revenue allocations in an ad hoc manner. We establish that our approach provides an upper bound on the optimal total expected revenue, and this upper bound is tighter than the ones obtained by the two benchmark methods. Computational experiments indicate that our approach provides significant improvements over the performances of the benchmark methods.     

Viability of Auction‐Based Revenue Management in Sequential Markets

The Internet is providing an opportunity to revenue management practitioners to exploit the potential of auctions as a new price distribution channel. We develop a stochastic model for a high‐level abstraction of a revenue management system (RMS) that allows us to understand the potential of incorporating auctions in revenue management in the presence of forecast errors associated with key parameters. Our abstraction is for an environment where two market segments book in sequence and revenue management approaches consider auctions in none, one, or both segments. Key insights from our robust results are (i) limited auctions are best employed closest to the final sale date, (ii) counterbalancing forecast errors associated with overall traffic intensity and the proportion of customer arrivals in a segment is more important if an auction is adopted in that segment, and (iii) it is critically important not to err on the side of overestimating market willingness to pay.     

Assortment Planning for Vertically Differentiated Products

We consider the problem of a retailer managing a category of vertically differentiated products. The retailer has to pay a fixed cost for each product included in the assortment and a variable cost per product sold. Quality levels, fixed, and variable costs are exogenously determined. Customers differ in their valuation of quality and choose the product (if any) that maximizes their utility. First, we consider a setting in which the selling prices are also fixed. We find that the optimal set of products to offer depends on the distribution of customer valuations and might include dominated products, that is, products which are less attractive than at least one other product, on every possible dimension. We develop an efficient algorithm to identify an optimal assortment. Second, we consider a setting in which the retailer also determines the selling prices. We show that in this case the optimal assortment does not include any dominated product and does not vary with the distribution of customer valuations when there is no fixed cost. We develop several efficient algorithms to identify an optimal assortment and optimally price the products. We also test the applicability of our methods with realistic data for two product categories.     

Optimal Dynamic Upgrading in Revenue Management

We consider a revenue management problem involving a two compartment aircraft flying a single leg, with no cancellations or over‐booking. We apply the practice of transforming a choice revenue management model into an independent demand model. Within this assumed independent model, there are two sets of demands, business and economy, each with multiple fare class products. A business passenger can only be accepted into business. An economy passenger can be accepted into economy or upgraded into business. We define a two‐dimensional dynamic program (DP) and show that the value function is sub‐modular and concave in seat availability in the two compartments. Thus the bid prices are non‐decreasing with respect to these state variables. We use this result to propose an exact algorithm to solve the DP. Our numerical investigation suggests that in contrast to standard backward induction, our method could be included in production revenue management systems. Further, when the economy compartment is capacity constrained, we observe a substantial monetary benefit from optimal dynamic upgrading compared to the static upgrading procedures currently used in practice.