Airline Revenue Management Systems Research Articles

Demand change detection in airline revenue management

Demand shocks—unobservable, sudden changes in customer behavior—are a common source of forecast error in airline revenue management systems. The COVID-19 pandemic has been one example of a highly impactful macro-level shock that significantly affected demand patterns and required manual intervention from airline analysts. Smaller, micro-level shocks also frequently occur due to special events or changes in competition. Despite their importance, shock detection methods employed by airlines today are often quite rudimentary in practice. In this paper, we develop a science-based shock detection framework based on statistical hypothesis testing which enables fast detection of demand shocks. Under simplifying assumptions, we show how the properties of the shock detector can be expressed in analytical closed form and demonstrate that this expression is remarkably accurate even in more complex environments. Simulations are used to show how the shock detector can successfully be used to identify positive and negative shocks in both demand volume and willingness-to-pay. Finally, we discuss how the shock detector could be integrated into an airline revenue management system to allow for practical use by airline analysts.

Premium cabin capacity sharing strategies: airline RM insights

This paper explores revenue management strategies for premium cabin capacity sharing, that is, accepting additional bookings in economy fare classes when some premium cabin seats are expected to remain empty. Two heuristics are presented and tested to evaluate the impacts of premium cabin capacity sharing on airline revenues and loads. These RM strategies are tested using the Passenger Origin Destination Simulation (PODS), which replicates the interactions between airline revenue management systems and passenger choice of airline itineraries, fare types and service offerings. Based on the simulations, the methodologies proposed can generate total revenue gains of up to 1.1%. However, two caveats are identified: first, losses in the revenue captured from premium fare classes are likely to be experienced due to displacement by economy fare class passengers. Second, premium cabin capacity sharing should only be implemented in the final stages of the booking process; otherwise, the sharing heuristics can result in revenue losses for the airline.

EVALUATION OF UNCONSTRAINING METHODS IN AIRLINES’ REVENUE MANAGEMENT SYSTEMS

Airline revenue management systems are used to calculate booking limits on each fare class to maximize expected revenue for all future flight departures. Their performance depends critically on the forecasting module that uses historical data to project future quantities of demand. Those data are censored or constrained by the imposed booking limits and do not represent true demand since rejected requests are not recorded. Eight unconstraining methods that transform the censored data into more accurate estimates of actual historical demand ranging from naive methods such as discarding all censored observation, to complex, such as Expectation Maximization Algorithm and Projection Detruncation Algorithm, are analyzed and their accuracy is compared. Those methods are evaluated and tested on simulated data sets generated by ICE V2.0 software: first, the data sets that represent true demand were produced, then the aircraft capacity was reduced and EMSRb booking limits for every booking class were calculated. These limits constrained the original demand data at various points of the booking process and the corresponding censored data sets were obtained. The unconstrained methods were applied to the censored observations and the resulting unconstrained data were compared to the actual demand data and their performance was evaluated.

Reinforcement learning applied to airline revenue management

Reinforcement learning (RL) is an area of machine learning concerned with how agents take actions to optimize a given long-term reward by interacting with the environment they are placed in. Some well-known recent applications include self-driving cars and computers playing games with super-human performance. One of the main advantages of this approach is that there is no need to explicitly model the nature of the interactions with the environment. In this work, we present a new airline Revenue Management System (RMS) based on RL, which does not require a demand forecaster. The optimization module remains but works in a different way. It is theoretically proven that RL converges to the optimal solution; however, in practice, the system may require a significant amount of data (a booking history with millions of daily departures) to learn the optimal policies. To overcome these difficulties, we present a novel model that integrates domain knowledge with a deep neural network trained on GPUs. The results are very encouraging in different scenarios and open the door for a new generation of RMSs that could automatically learn by directly interacting with customers.

Can demand forecast accuracy be linked to airline revenue?

Since accurate demand forecasts are a key input to any airline revenue management system, it is reasonable to assume that an improvement in demand forecast accuracy would lead to increased revenues. However, this relationship has often been called into question. Past work has not conclusively proven that more accurate demand forecasts lead to higher revenue, causing researchers and practitioners to debate whether the concept of demand forecast accuracy itself is “myth or reality.” In this paper, we demonstrate that it is possible to consistently link demand forecast accuracy to airline revenue. After discussing why traditional demand forecast error metrics have struggled to demonstrate this relationship, we evaluate a novel conditional demand forecast error metric which compares demand forecasts to historical bookings conditional on the set of fare classes that were open at the time of booking. We prove under some mild assumptions that minimizing conditional demand forecast error will maximize revenue under any fare structure and customer choice behavior. These theoretical findings are supported by simulations in both a simple, single-leg model and in a complex multiple-airline network in the Passenger Origin–Destination Simulator. We find that price elasticity parameter bias of ± 10% can reduce revenues by up to about 1%, while price elasticity parameter bias of ± 20% can reduce revenues by up to 4%. We close by discussing the implications of the findings for revenue management practitioners.

Network Revenue Management with Cancellations and No‐Shows

Customer cancellations and no‐shows have an important impact on the performance of an airline's revenue management system. In this study, we study airline network revenue management problems with customer cancellations and no‐shows, considering two types of demand models: independent‐demand and choice‐based‐demand models. In independent‐demand models, the booking request rates for different products are independent of the assortment of products offered by the airline (exogenous), whereas in choice‐based‐demand models they depend on the airline's assortment (endogenous). The stochastic problems of optimizing the booking policies for the two demand models are not tractable because of the high‐dimensional state space. Therefore, for each model, we use the corresponding deterministic, continuous‐time, and continuous‐state model, known as the fluid model. A booking policy based on the fluid solution can be implemented for the original stochastic, discrete model. This fluid policy is shown to be asymptotically optimal when the arrival rates become high and the seat capacities become large. In the independent‐demand setting, an optimal fluid solution can be computed by solving a convex optimization problem for which efficient algorithms exist, whereas for some choice‐based‐demand models, the fluid control problems are known to be intractable. We also develop mixed fluid policies for the independent‐demand setting, taking advantage of the real‐time booking state, to improve the performance over the fluid policy. In a factorial experiment of 20 small‐sized single‐leg problems, the revenue loss under one of the mixed fluid policies is demonstrated to be less than 1% of the optimal value for all problems.

Yield optimization for airlines from ticket resell

During the reservation period of some flights, the demand significantly fluctuates due to changes in the business environment. Some demand variations can hardly be anticipated by airline revenue management systems. Therefore, it often happens that the sales policy applied at a given time in a flight turns out to be suboptimal a posteriori. A mechanism of ticket buy-back can then be an interesting tool for the airlines aiming at boosting revenue from these flights. This paper addresses the main stages of a buy-back process triggered by an airline. The process includes the revenue management-based solutions to support the selection of the tickets and the computation of the proposed prices to get them back. The main contribution of this paper is a new mathematical model which optimizes airline-expected revenue from buy-back according to the probability of passenger acceptance. This model can be applied for many different compensation schemes that could be put in place by an airline to spur some of their passengers to sell back air tickets to the airline. Three of them are further analyzed. We simulate buy-back campaigns for four flights with data drawn from real operations and compare additional revenue due to buy-back, according to the selected compensation scheme. Results emphasize our intuition that business benefits can be expected from a well-automated mechanism of ticket buy-back and resell in the airline industry. Depending on the flight and demand characteristics, up to over $$10\%$$ additional revenue can be expected to be added on top of the revenue obtained from a standard revenue management system on a single flight.

Optimization models in RM systems: Optimality versus revenue gains

Optimization models in airline revenue management (RM) systems have evolved from single flight leg to network revenue maximization to marginal revenue optimization for less restricted fare structures. This article reviews the most common optimization approaches that have been widely implemented in airline RM systems, with a focus on how the mismatch between model assumptions and reality can affect achievable revenue performance. Simulation findings from the Passenger Origin-Destination Simulator are used to illustrate how robustness and revenue gains, as opposed to theoretical optimality, have driven the widespread adoption of practical optimization models in RM systems.

Revenue impacts of demand unconstraining and accounting for dependency

Accurate forecasts are crucial to an airline revenue management system. Historical demand data for two fare classes used for forecasting in an airline are often censored and correlated. In this article, we propose a methodology that takes into account both censorship and correlation of demands. We then study its impact on the expected revenue and compare it with three standard methodologies available in the literature by using extensive simulation. Our results show that the opportunity cost of neglecting demand censorship is upto 1 per cent whereas that for neglecting the dependency of demands can be of the order of 2 per cent. Consideration of both truncated demand and dependency between fare classes can lead to a significant (of the order of 2.5 per cent) revenue increase.

Solving the hierarchical data selection problem arising in airline revenue management systems

In this paper, we describe an important optimisation problem arising in airline revenue management systems. The problem is to select the maximum number of average fare data while keeping the selected data in a non increasing hierarchical order. We first formulate the problem mathematically using 0?1 binary integer programming, and then further derive a stronger formulation using clique cuts. Moreover, an extension of the problem is studied where the relative importance of each data point can be derived from passenger count information. We develop an efficient dynamic programming-based algorithm to solve the problem optimally. The preliminary computational results using real airline data show that our approach can solve the problem efficiently and save significantly much information that are previously discarded.

Special Issue Papers: Revenue management performance in a low-fare airline environment: Insights from the Passenger Origin—Destination Simulator

The growth of low-fare airlines has led researchers and network airlines alike to question the validity of existing pricing and revenue management strategies in the changing competitive environment. Given the less restricted fare structures offered by low-fare new entrant carriers, the revenue benefits of existing revenue management systems are likely to be affected. The Passenger Origin—Destination Simulator (PODS) is used to evaluate the changes in airline revenue management system performance in a hypothetical network served by incumbent network carriers and a low-fare new entrant airline. Simulation results show that, while incumbent network carrier revenues are indisputably reduced by low-fare entry, the incumbents and low-fare new entrant alike realise substantial benefit from the use of revenue management. Furthermore, the incremental revenue gains from the use of network revenue management (relative to leg-based revenue management) are very robust to changes in the competitive environment.

Preserving Your Revenue-management System as a Trade Secret

Abstract A hotel's or airline's revenue-management system is ( 1 ) quite valuable and ( 2 ) a trade secret. Trade secrets involve special proprietary information and processes, including publicly available information that is assembled in a special manner by the trade secret's owner. Establishing that a process is a trade secret involves demonstrating that it is, in fact, a secret, that information about the process is restricted to employees and licensess, and that the owner makes a specific effort to shield the information. Hotels' revenue-management systems are particularly at risk for either general disclosure or corporate espionage because of the relatively rapid turnover of employees and the fact that those employees often go to work for competitors. Thus, a wise hotelier will take the following steps to protect the company's revenue-management system: ( 1 ) maintain the information's physical security by restricting access, ( 2 ) place restrictive legends on the information stating that it is in fact confidential, ( 3 ) state in employee handbooks that the revenue-management information is a trade secret and will be shared only in confidence, ( 4 ) require revenue-management employees to sign confidentiality agreements, ( 5 ) train employees and remind them regularly of the confidentiality of the revenue-management information, and ( 6 ) ask exiting employees to reaffirm their agreement not to share the confidential trade secrets. The former employer may also consider sending an advisory letter to a departing employee's new employer stating that the revenue-management information is confidential, but that letter-writing step involves other issues that require legal counsel.

Revenue Management Without Forecasting or Optimization: An Adaptive Algorithm for Determining Airline Seat Protection Levels

We investigate a simple adaptive approach to optimizing seat protection levels in airline revenue management systems. The approach uses only historical observations of the relative frequencies of certain seat-filling events to guide direct adjustments of the seat protection levels in accordance with the optimality conditions of Brumelle and McGill (1993). Stochastic approximation theory is used to prove the convergence of this adaptive algorithm to the optimal protection levels. In a simulation study, we compare the revenue performance of this adaptive approach to a more traditional method that combines a censored forecasting method with a common seat allocation heuristic (EMSR-b).

Revenue Management Without Forecasting or Optimization

We investigate a simple adaptive approach to optimizing seat protection levels in airline revenue management systems. The approach uses only historical observations of the relative frequencies of c...