Sabre's system helps cut the complexity of figuring where American Airlines should fly -- and how often In March 1989 American Airlines Inc. got its first taste of the power of forecasting. Eastern Airlines had just declared bankruptcy and American needed to figure out as soon as possible how it could capitalise on this turn of events and lure its erstwhile competitor's customers away. Using an early mainframe prototype of what is now its Integrated Forecasting System (IFS) software, American compiled a forecast that highlighted new opportunities around the world. Never mind that it took eight hours-IFS showed that it would be more advantageous to open a new hub in Miami rather than Atlanta, where Eastern had reigned.
Just one week later, American's CEO and chairman at the time, Bob Crandall, presented the plan to American's board at corporate headquarters in Fort Worth, Texas. The Miami hub has generated hundreds of millions of revenue since, says Brad Jensen, vice president of application development for Sabre Technology Solutions (the IT division of The Sabre Group Inc., a sister company to American Airlines).
Today, on a Unix system, the IFS runs even faster and handles far more complexity than earlier versions. It can forecast market share for a worldwide network of 4 million services and nearly 400,000 city pairs in three minutes.
IFS helps American manage a daunting schedule of 4,000 daily flights by uncovering ways to cut costs or boost revenue. In a 24-hour period, IFS can calculate up to 60 iterations, analysing the financial impact of adding or deleting flights, starting service to a new city, substituting bigger or smaller planes or even schedule changes by American's competitors.
The system is also bucking conventional wisdom. It was widely believed that if American took jets off certain routes and replaced them with prop planes, it would lose business on those routes. IFS demonstrated that the airline could redeploy those same jets by adding flights to longer, more profitable routes, thereby making up for any losses incurred from the prop planes. Jensen says the end result was an overall decrease in costs per mile.
The ability to act fast, armed with a vast amount of supporting data for making sound decisions, is giving American an edge on the competition. The airline estimates IFS has increased its annual profits by as much as $US500 million dollars since the system was launched in 1996, so much so that American does not make any major economic decisions without first running a model through IFS, says Steve Iverson, managing director of capacity planning for American and one of the system's lead sponsors. "In the '90s, it's not smart to make decisions based on one set of options," he observes.
While Iverson hesitates to make a direct connection between IFS and the airline's financial performance, it's hard to ignore the numbers. American's parent company, AMR Corp., reported 1997 revenues of approximately $US18.6 billion, making American one of the leading domestic airlines. The company beat Wall Street's expectations with a 34 percent growth in earnings in the third quarter of 1998.
An Uncharted Route
Sabre's award-winning system is the result of more than a decade of research and development. Back in the '80s, American had two areas it needed to improve: speed and accuracy in forecasting. Using PC spreadsheets and calculators, analysts were taking days to forecast the impact of a single flight change.
Modelling an entirely new schedule was out of the question. With deregulation, the complexity of airline schedules began to increase exponentially along with demand, and deleting one flight would affect hundreds of other connections.
Executives wanted to be able to respond to industry changes that were occurring on a daily basis from competition, regulation or economic shifts. The mainframe version of IFS, launched in 1989, eventually enabled American to run forecasts in an hour. But once senior management saw the system's potential, it demanded results in minutes.
The challenge to build an engine that could crunch the huge volume of airline data at that speed was staggering. There are some 400,000 potential flight combinations that American tracks, according to Jensen. Sabre knew it needed a parallel-processing architecture to meet the performance goals. The project team chose a client/server platform that would give users more control in launching and managing the forecasts from the desktop; the object-oriented environment would allow developers to cut costs by reusing software objects in future applications. The project was risky back in 1991, says Jensen, because there were few developers who understood the ins and outs of supercomputing, and fewer vendors that had mature object databases. Sabre had to create the system from scratch on new hardware.
Fortunately, the $US16 million project had strong backing from the very top at American, whose Sabre reservation system had revolutionised the travel industry in 1963. When Bob Crandall first heard Sabre's proposal for IFS, he leapt out of his chair and yelled, "I want that system!" "There was a realisation among the senior management at both companies that in order to have significant breakthroughs, you have to take significant risk," reflects Thomas M. Cook, president of Sabre Technology Solutions.
Yet because of the complexity of both the algorithms and the business problem, Jensen says the project would have failed if not for a close cooperation between Sabre and the capacity planning department at American. The development team was housed on the same floor as users, and senior American execs participated in meetings on a weekly basis. When the system was launched in 1996, analysts in capacity planning were ecstatic at how much faster and powerful it was, reports Kevin Fleischood, a project manager from Sabre. Still, the system represented a dramatic shift in the way American made business decisions, and Iverson admits that managers outside of his department were sceptical at first. Once managers understood that the system would complement traditional decision making and business judgement, rather than replace them, those concerns died down.
Today IFS is an integral part of American's semi-annual operational scheduling plan, allowing the airline to predict revenues and market share for any city pair in the world (for instance, San Jose to Boston) under any number of conditions, such as adding a connection or assigning different aircraft.
Through a browser interface, American's forecast analysts feed the system with public and internal data from approximately 60 different source files detailing estimated passenger demand, fares by market, seating capacity of its planes, unit costs per flying hour and flight schedules of competitors. After the forecast is complete, the planning department measures the results against historical passenger data and strategic goals to make the final recommendation.
With IFS, American can look at every change it or a competitor makes to a schedule; the more studies it can run, the deeper its level of planning information. For instance, the system can measure the impact of United adding a flight from Chicago to Syracuse, N.Y., on an American flight between Dallas-Fort Worth and Phoenix. Now, says Jensen, senior management gets answers to hundreds of questions during a given month. Before, they had to prioritise the top two or three.
Iverson says a surprise result of IFS is the way it has changed the daily operations of the capacity planning department. For one, the volume of forecasts IFS can run in a day coupled with the process control enabled by the system has allowed the department to reduce staff, eliminate graveyard shifts and increase the number of scenarios analysts can study. And because the system's design allows users to launch multiple forecasts at once, analysts now spend less time tweaking the system and more time analysing business problems.
Overall, there are less errors and more consistency in the forecasting process.
The bottom line? American is developing an intimate understanding of its marketplace and its customers' preferences. Observes Christopher Hoenig, an awards judge and CEO and president of Washington, D.C., consulting firm Solutions Inc.: "It epitomises what really sophisticated software, combined with good data and good business processes, can do in allowing a very large organisation to look ahead and anticipate many scenarios-it's sort of a 'giants learn to dance' application." IFS continues as a work in progress for Sabre. Sabre has built similar versions of IFS for other airlines, and most recently sold versions to the London subway system and the French railways. Cook calls the technology "a cornerstone of our growth strategy." Still, the technological achievements of the system, while grand, don't hold a candle to the ultimate result: a forecasting system that brings reliability and profitability to a leader in an industry rife with uncertainty.
Senior Writer Polly Schneider can be reached at firstname.lastname@example.org.
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.