Laying The Foundation For The Future Of Airline Technology
Since the late 1950s, Operations Research (OR) has helped the worldwide airline industry continually transform itself to effectively compete in the marketplace and match the services airlines provide with increasingly complex consumer demands. In the process, OR has saved airlines millions of dollars, improved their operational efficiency and equipped them to generate additional incremental revenues.
In 2013 alone, the world’s airlines carried more than 3 billion passengers and directly employed more than 8.7 million people. In fact, if the aviation industry were a country, it would rank 21st in the world in terms of gross domestic product (GDP), generating US$606 billion in GDP last year. By 2026, the industry is forecast to contribute US$1 trillion to the world GDP.
While these facts are impressive and the economic impact clearly significant, the airline industry worldwide has continually struggled to transform itself over the years to remain relevant and reliable. In doing so, operations have become more complex in response to an increasingly competitive marketplace and a rise in customer expectations.
The introduction of the “jet age” in the 1950s moved the airline industry from a novelty for a few elite passengers to a viable mode of transportation for the general public. The transition ignited an explosion in the number and utilization of scarce and expensive resources, including aircraft, crew and airport facilities, over the decades and accelerated the need for airlines to effectively and efficiently plan and manage these dynamic assets in a continually fluctuating environment.
To handle this level of complexity, airlines began developing and using tools and techniques to model this volatile environment and subsequently, the impact of changes and resource utilization on various scenarios. Some carriers established their own operations research (OR) groups, which have saved millions of dollars over the years, to lead this initiative.
In 1961, a handful of OR professionals from five airlines formed the foundation of The Airline Group of the International Federation of Operational Research Societies (AGIFORS). Dedicated to the advancement and application of OR within the airline industry, to date, AGIFORS has more than 2,800 members representing over 500 airlines, aircraft manufacturers and aviation-related companies and universities.
That same year, another landmark achievement was realized when the OR department at American Airlines, from which Sabre® was launched, introduced the first automated real-time reservations system and subsequently began hosted reservations services for carriers on a massive scale, laying the groundwork for the future of technology in the airline industry. After airline deregulation in 1979 under the administration of President Jimmy Carter, advanced decision-support applications were developed and deployed for travel suppliers, such as airlines, hotels, rental car agencies, and rail and tour operators, to help them remain competitive and boost their bottom lines.
The Rise Of OR
OR first attracted large-scale recognition in the mid-1980s when a small group of American Airlines’ OR professionals, now part of Sabre, developed the world’s first yield-management (or “revenue management,” as it is called today) system at the request of Robert L. Crandall, the airline’s former chairman and chief executive officer. Discount carrier PEOPLExpress threatened the very existence of American Airlines with its deeply discounted tickets, which undercut major carrier fares by 50 percent to 70 percent. The airline needed a strategic and tactical weapon to counter this competitive threat.
Left with no choice but to match the low fares, Crandall relied on yield management to control the availability of these deeply discounted fares. In the absence of yield-management controls, PEOPLExpress sold all seats at deeply discounted prices, which were not enough to support the airline’s cost structure over the long run. By September 1986, PEOPLExpress was gone. It failed for two reasons: rapid uncontrolled expansion of the fleet and the absence of yield-management controls.
“We were a vibrant, profitable company from 1981 to 1985, and then we tipped right over into losing US$50 million a month,” said Donald Burr, chairman and CEO of PEOPLExpress, in 1986. “We were still the same company. What changed was American’s ability to do widespread yield management in every one of our markets. We had been profitable from the day we started until American came at us with Ultimate Super Saver fares. That was the end of our run because they were able to underprice us at will and surreptitiously. There was nothing left to defend us.”
In the following years, OR professionals have used techniques such as mathematical modeling, stochastic processes, deterministic and stochastic large-scale network-optimization models, data analytics, algorithms, statistical analysis, machine learning and decision sciences to solve complex challenges in all aspects of consumer travel. As a result, significant progress has been made in the areas of flight scheduling, pricing, revenue management, crew planning, airline operations, flight planning, staff planning and cargo.
OR In Action
A common trait of OR-led projects is the application of scientific methods to solve complex problems involving a large number of decision variables. The process begins by quantifying the business objectives that need to be achieved subject to the operational and business constraints that must be met. An abstraction of the problem is then developed in mathematical terms and solved with a range of techniques, such as linear or non-linear optimization, stochastic processes, etc. OR projects are data intensive, and special efforts are made to ensure data quality and accuracy.
The Shopping Experience
When a customer plans a trip, the first step is to explore available travel options to and from the desired destination. This shopping process requires the generation of outbound and inbound schedules, which involves selecting a set of relevant itineraries from thousands of available options and pricing them based on real-time availability of seat inventory. Conservatively, a single shopping request may require the evaluation of more than 100 million schedule and fare combinations within seconds to arrive at an answer.
According to a study by Topaz International, Sabre’s Air Traffic Shopping Engine was 15 times more likely to find a lower fare than two other leading companies. Besides low-fare efficacy, Sabre’s shopping diversity — the ability to display a selection of itineraries based on departure time, elapsed time, non-stops versus connections, interline and carriers — relies on sophisticated consumer-choice models to select itineraries that meet customers’ needs.
To enable airline shopping for customers utilizing various selling channels, such as travel agencies, online travel agencies (OTAs) or individual airline websites, Sabre deploys a combination of fare-, availability- and schedule-led algorithms. The shopping infrastructure now has 700 million fares that are updated hourly, 300 million fare rules, 50 million active availability records and 5 million schedules and can support 20 million shopping queries per day and 550 shopping queries per second at peak times.
Airline Pricing And Revenue Management
Today’s fare-management systems eliminate latency and improve accuracy of the fare-filing process through fare distributors, ATPCo and SITA. The timely and accurate distribution of fares generates an estimated US$1billion in incremental revenue annually for airlines.
When availability and prices are determined during the shopping process, revenue-management solutions generate 4 percent to 8 percent in incremental revenues for an airline by selling the right seat to the right customer at the right price at the right time. The natural evolution of leg/segment revenue management was the introduction of origin-and-destination (O&D) revenue management to effectively manage connecting traffic for network carriers by Sabre in 1987 with virtual nesting controls and subsequently with more granular continuous-nesting (“bid-price”) controls.
This area continues to be refined with more sophisticated models such as consumer-choice-based, network revenue-management demand and optimization technologies that jointly consider upsell and cross-flight recapture effects to optimize traffic flow through an airline’s network. These models are especially critical with restriction-free or lightly restricted tariffs advocated by many low-cost carriers.
To ensure an airline’s offer best matches demand for travel, evolving solutions optimize the schedule by combining advanced consumer-choice models to characterize consumer preferences, such as departure time, elapsed time and carrier preference, with large-scale network optimization models. Network-planning, flight-scheduling and capacity-allocation applications from Sabre generate an estimated US$8.5 billion annually for airlines by scheduling 3 trillion available seat kilometers (ASK) for approximately 80 percent of the world’s top 100 airlines.
Today, schedule-management solutions with decision-support capabilities can support all facets of the schedule-development process, including analysis of codeshare partners’ networks, from schedule creation 5 years in advance until the day of departure.
For airlines with multiple fleet types and capacities, capacity allocation models determine the optimal allocation of aircraft types to flights based on demand forecast, revenue and operational costs. The resulting aircraft assignments are feasible with respect to down line processes, such as crew scheduling, maintenance planning and airport operations.
Collectively, the advanced network planning, flight scheduling and capacity allocation applications generate more than 5 percent to 8 percent in incremental revenues over traditional processes employed by airlines.
Close-in Re-fleeting Benefits
The consulting team at Sabre Corporation has conducted studies with several airlines that practice close-in re-fleeting. Results show that airlines can achieve between a 1 percent and 1.4 percent increase in revenues, higher load factors of 0.5 percent to 0.7 percent and increased yields of approximately 2.1 percent.
Another potential source of revenue improvements for an airline results from close-in re-fleeting, the process of adjusting fleet assignments based on tactical demand forecasts by flight leg and revenue estimates from revenue management. These factors are used by fleet assignment models to improve network revenue performance.
Close-in re-fleeting is gaining acceptance in schedule planning and operations due to its potential benefits. Consulting studies at Sabre with airlines adopting close-in re-fleeting have shown that the approach results in higher revenues (1 percent to 1.4 percent), load factors (0.5 percent to 0.7 percent) and yields (approximately 2.1 percent).
Flight Planning And Fuel Optimization
Consider these statistics:
- Today, approximately 33 percent of an airline’s operating costs are spent on fuel, up from 13 percent in 2001.
- The airline industry worldwide generates around 2 percent of manmade CO2 emissions.
- Jet aircraft currently in service are 70 percent more fuel efficient per seat kilometer than those in the 1960s.
- By 2020, net aviation carbon emissions will be one-half those of 2005.
- The Carbon-Neutral Growth 2020 initiative caps CO2 emissions at 2020, even as demand for air service continues to grow.
While airlines have long sought to find the most efficient ways to utilize their most expensive asset, aircraft, there is now a focus on doing so in a manner that is eco-friendly, as well. Today’s multi-variable flight planning systems generate optimal flight plans that minimize operating costs by 1 percent to 7 percent and optimize fuel burn, which, in turn, means reduced carbon emissions.
The model developed by Sabre calculates emissions for a flight based on carrier, aircraft type and seating configuration. The detailed aviation carbon calculator is capable of providing emissions information to carriers at a higher level of accuracy than before. Additionally, by integrating data and workflow with airport systems, flight dispatchers are provided with data visualization and visual validation of options and selections.
Crew Planning And Flight Operations
These two areas utilize some of the most sophisticated optimization models available today to effectively solve complex problems involving thousands of interdependent variables in the areas of crew pairing, crew rostering and aircraft movement control, resulting in improved on-time performance, reduced operating costs and increased aircraft utilization.
Crew pairing and rostering are large-set partitioning optimization problems that cannot be solved to optimality in polynomial time. For example, for a large fleet type to minimize excess pay and credit, the number of potential crew pairings can run into 1,015, while the crew rostering problem, which can have thousands of crew and tens of thousands of pairings, can run into 1,025 variables, prompting the need for intelligent heuristics to solve the problem that is close to optimal.
The crew management solutions suite from Sabre optimally plans and tracks the daily schedules of more than 100,000 crewmembers, improves on-time performance by more than 10 percent and provides crew cost savings in excess of US$200 million annually. As well, the flight operations suite manages over 6,700 aircraft and 5.7 million flights annually, improves on-time performance by 10 percent and reduces delay costs by US$15 million annually.
In the United States, an average of 23 percent of flights are subject to a range of disruptions that result in long travel times for passengers. In many cases, these disruptions are beyond an airline’s control and are a result of weather conditions or increasing congestion in the national airspace.
Adopting a “wait-and-see” attitude often leads to poor decisions and extended delays for customers. On the other hand, some airlines have become ultra-conservative in their decision making, cancelling large numbers of flights to avoid recent government penalties associated with prolonged tarmac delays — US$27,500 per passenger for ground delays in excess of three hours from the gate. In addition, such disruptions can wreak havoc on crew schedules when duty time-limit rules, which apply after arrival at the gate, stipulate crewmembers have exceeded their legal time to operate a flight.
There are three components to this challenging problem — passenger reaccommodation, aircraft recovery and crew recovery. Thus, solutions developed to help airlines effectively manage disruptions must address these three issues with the goal of minimizing passenger inconvenience and restoring operations as quickly as possible.
However, before passengers can be reaccommodated, a new schedule must be generated and aircraft, crew, gates and ground staff reassigned using modeling and advanced decision-support solutions.
A passenger reaccommodation model developed by Sabre receives schedule change and disrupted flight information and evaluates each passenger’s itinerary based on an airline-defined passenger list that prioritizes passenger attributes such as unaccompanied minors, frequent flyer status, fare paid and class of travel. Passengers are rebooked and notified by an automated alerting process that strives to accommodate highly valued customers first, retain brand loyalty and minimize passenger inconvenience.
In 2012 during Hurricane Sandy, Sabre’s passenger reaccommodation model recommended optimal rebooking solutions for more than 71,000 reservations from Oct. 28-30, for JetBlue and WestJet. The reaccommodation solution automates the entire rebooking process, which significantly reduces the amount of manual work required by reservations and airport crewmembers, allowing them to provide more personalized service to customers during an irregular operations event. Since implementing this solution, JetBlue has realized a measurable increase of positive comments in customer satisfaction surveys. Additionally, JetBlue estimates a 25 percent reduction in call volumes at its reservations call centers, which allows its agents to continue to provide superior customer service, even during operational disruptions.
The operations recovery system takes the disrupted flight schedule, operational constraints (airport curfews, gate limits, air traffic control flow management programs, equipment restrictions and weather restrictions) and data on all available aircraft and crewmembers to generate a proposed recovery plan, including a revised flight schedule, as well as revised fleet and crew assignments. The proposed plan is as close as possible to the original flight schedule, while accounting for scheduled crew assignments and passenger itineraries.
Crewmembers are reassigned based on a revised flight schedule generated by a schedule recovery decision-support tool or manually. Disruptions are resolved at the crewmember level, and alternative solutions proposed with respect to crew availability, crew preference and cost are factored into the complex decision-making process.
OR has also helped airlines improve efficiency in the area of airport operations, specifically, in the creation of decision-support tools for the planning and management of resources at gates and on the ground and utilization of terminal staff.
Recently, Sabre developed a prototype Hub Control Center to address recovery at a hub airport when the inbound flights are delayed. The model finds the most efficient solution to address violations associated with aircraft turns, connecting passengers and cargo. Results with airline data indicate a savings of 2 percent to 5 percent in flight delay costs while minimizing overall passenger inconvenience.
Crew Management And Flight Operations Technology
The crew management solutions suite from Sabre optimally plans and tracks the daily schedules of more than 100,000 crewmembers, improves on-time performance by more than 10 percent and provides crew cost savings in excess of US$200 million annually. In addition, the flight operations suite manages more than 6,700 aircraft and 5.7 million flights annually, improves on-time performance by 10 percent and reduces delay costs by US$15 million a year.
Historically, the airline industry, while vital to global commerce, has struggled to maintain profitability. In recent years, the industry has moved to a more sustainable business model that allows consumers to pay for ancillary products they value, while enabling airlines to reduce the costs associated with providing services that some consumers would not choose to buy. An early enabler of this new business model, Sabre developed the capability for airlines to merchandise and charge separate fees for key services.
Republic Airlines was the first to utilize this capability, in the form of paid-selection for seats with improved legroom. Other airlines, including JetBlue, also employed these tools, which were introduced before the industry was ready for long-term solutions and industry standards, to market and sell their enhanced legroom or preferred-seating products.
More recently, Sabre worked with US Airways to launch “ChoiceSeats,” the airline’s pre-paid seat-offer program. Sabre has moved to industry-standard solutions to ensure airlines can broadly and readily adopt these technologies to drive this relatively new business model.
OR Continues Innovation
Within the past five years, OR has made increasingly significant advances in support of the airline industry and related businesses. These include:
- Network planning applications that maximize demand flow in the network by retiming flights to/from a hub, determining optimal codeshare opportunities and suggesting new markets to enter, as well as frequency of operation. Of particular interest is a unique optimization tool that explicitly models upsell and recapture.
- Hotel shopping algorithm that optimizes screen displays with a calibrated consumerchoice model to maximize market share and conversion rates.
- The first O&D revenue opportunity model for dependent demands.
- The first airline decision-support pricing model that recommends optimal tariff structures to achieve a desired traffic mix.
- Consumer preference-driven air shopping display algorithm that delivers a single itinerary back to the consumer based on a desired schedule, carrier and fare.
- Fare forecasting that predicts when fares will rise or fall by market.
- Gate assignment model that considers multiple objectives and constraints while maximizing overall assignment quality.
- Tactical maintenance planning model that uses tail assignments as input and distributes maintenance events among available blocks for each tail number to maximize aircraft utilization.
- Hub control model to support airport operations for payload and turnaround management.
Big Data And OR
Lately, there has been quite a bit of discussion about Big Data and how this data — structured and unstructured — should be captured and leveraged to produce a competitive advantage. Essentially, Big Data are data sets that are too large or complex to manipulate or interrogate with standard methods or tools.
OR models are typically data intensive and the OR practice has always processed and generated large volumes of data. OR techniques are a horizontal enabler for many of the Big Data value propositions that support revenue generation, improve conversion rates, enhance process efficiency, refine customer experiences, generate target offers and enhance customer service.
A partial list of Big Data applications and actions powered by OR techniques Sabre is currently investing in include:
- Dynamic inventory alerts and availability that reflect prevailing competitive market conditions to generate incremental revenues;
- Pricing opportunity model to fine-tune airline pricing strategy;
- Machine-learning algorithms to predict when air fares will increase or decrease in key markets;
- Twitter data for sentiment analysis, lead generation and variable manning at airports on day of departure;
- Recommend alternate flight plans based on advanced weather data.
As in past decades, OR will continue to lay the foundation for the future of airline technology, enabling airlines to focus on a seamless travel experience for their customers. From airline planning to daily operations, OR enables products and services that offer unique value propositions and return on investment by minimizing costs and/or maximizing revenues.
Sabre's Operations Research Footprint
Key contributions in revenue management from Sabre include:
- First yield-management system for the airline industry launched (American Airlines)
- First virtual-nesting system for O&D control launched (American Airlines)
- First continuous-nesting system for O&D control (Club Med and Holiday Inn)
- Single largest deployment of O&D to control US$50 billion of seat inventory for US Airways and American Airlines in November 1998
- First component-based open-systems version of real-time inventory for O&D control (outside of a traditional reserva- tions TPF/ALCS environment) (Air France)
- First low-cost carrier yield-management solution (bmi)
- First consumer-choice-modeling-based demand-forecast model to generate tactical forecasts for revenue manage- ment in 2007 (GOL)
Highlights of some of the key historical contributions from Sabre include:
- World’s first revenue-management system
- World’s first large-scale crew-planning model
- World’s first O&D revenue-management system with virtual nesting
- World’s first hurdle-rate-based (“bid price”) inventory controls for all North American properties of Holiday Inn Worldwide
- World’s first crew-pairing system with long-haul crew augmentation and down ranking (Singapore Airlines)
- World’s first O&D revenue-management system with continuous-nesting controls (bid-price controls) and 100 percent polling (all flights on O&D control)
- Deployment of Primal-Dual algorithm for crew pairing, crew rostering, ground staff rostering and crew recovery World’s first O&D fleet-assignment model deployed (SAS)
- Modeling of restriction-free (dependent demand) and lightly restricted tariffs for revenue management
- First deployment of the dependent-demand model (British Midland)
- World’s first demand-forecast model based on consumer preferences (consumer-choice model) deployed (GOL)