American Airlines: Hitting Turbulence

By Deborah Gage Print this article Print

Forty years ago, the Sabre computer reservation system propelled American Airlines' business to great heights. Today, it could contribute to the company's demise.

Captain Doug Pinion has many ideas for ways that American Airlines could save money. As scheduling chairman for the Allied Pilots Association—American's pilots' union—Pinion regularly discusses with management how the airline could use automation to make flight scheduling less expensive and more convenient. For example, back-up pilots could use the Web to bid on openings on flights planned for the following day.

Instead, American calls pilots in order of seniority and waits for each one to answer the phone—a time-wasting and inefficient process.

Pilots' ideas are often rejected as too expensive, Pinion says. That's because the proposals require esoteric computer programming of Sabre, a 40-year-old mainframe system developed by American and IBM that predates modern computing, and was once the backbone of American's information systems.

Former Sabre CIO Terry Jones, now a consultant, describes Sabre as "an idiot savant."

"Ask it if Terry is going to Pittsburgh, it says yes," Jones says. "Ask it if Bob is going to Pittsburgh, it says yes. But it can't tell you how many people are going to Pittsburgh. It can't add."

American says it is working on the problem. But Sabre—emblematic of all that was right at American years ago—today threatens the airline's ability to improve efficiency and survive against lower-cost rivals.

During its heyday, Sabre ushered in the era of computer reservation systems, radically lowering American's costs of processing tickets, and giving it such an advantage with travel agents that the U.S. Department of Justice (DOJ) investigated American for possible anticompetitive behavior. (The DOJ claimed American programmed Sabre to list its own flights first; American claimed the listings were alphabetical.) As a result, all reservation systems were regulated.

Sabre holds such a wealth of data on passenger travel patterns that American developed software to maximize revenue—not just for itself, but to sell to other airlines. In 1996, American made Sabre a subsidiary and spun off about 18% of it to shareholders, completing the process in 2000 and making Sabre independent.

Yet, American cannot escape ties to the old mainframe, even as it tries to improve efficiency and slash expenses.

American's CIO Monte Ford says American has the same older systems as other airlines and is at no more of a disadvantage than any other carrier. "In a lot of ways we're better advantaged because of the kinds of upgrades we've completed or are about to complete," he says.

American is showing some signs of stability. On June 25, citing better revenue per seat mile and lower costs, the airline said it generated cash from operations in May and might do so in June. Still, American acknowledged its revenue is "still depressed relative to historical levels," and it has not ruled out the possibility of filing for bankruptcy. Its situation is further complicated by Sabre, which also was hit by the downturn in the airline industry and has its own challenges. A year after being spun out, Sabre sold its outsourcing business, which it deemed nonstrategic—and which included the contract for American's information technology infrastructure—to EDS, requiring American to enter a three-way relationship to manage its technology.

Changing Old Ways

Over the years, Jones says, many functions have been moved off the original Sabre mainframe onto Unix and Windows systems—he describes them as "a set of ladies in waiting" feeding data to the queen. However, he says, there is still no faster way than the mainframe to process large numbers of transactions. So it houses American's Flight Operations System, which plans flights and crews and must communicate with the pilots, and its Passenger Sales System, which handles reservations and communicates with AA.com.

The mainframe causes problems for AA.com just as it does for the pilots. Sabre's flexibility—its native interface is a blank screen—was once an advantage, Jones says, because it let a travel agent make reservations by entering information—names, flight numbers, and so on—in any order.

Because Sabre has no standard way of recognizing incoming data as a transaction, a reservation coming from an agent may be entered differently and, therefore, understood differently than a reservation made by a passenger through AA.com.

Travel agents, for example, are inconsistent in how they use abbreviations. Programmers say this is one of many exceptions that require American to manually tweak code. Code tweaking is also required for AA.com to communicate with American's mammoth frequent-flier database, another mainframe-based system that feeds information to the Web site.

Despite the labor-intensive code-tweaking, AA.com is a top priority at American because it helps the airline rapidly cut costs. Every ticket sold through the Web site can save up to $15 in distribution costs.

The Web site is also a model for how American's information technology infrastructure could work in the future. American relaunched AA.com a year ago on software from Art Technology Group, which is built on Java instead of the older and less flexible C++ code used in the previous site.

All systems used by customers—the airport kiosks, telephones and so on—are now being migrated to the same code that runs the Web site. This means that new code, such as the ability to print boarding passes, can be shared by different devices such as kiosks and PCs; and it can be added quickly—AA.com rolls in new features every two weeks. ATG's software also modernizes and standardizes connections to American's back-end systems—it uses Enterprise JavaBeans to exchange business rules with the frequent-flier database, XML to receive data on flight deals from the content management system, and Jini, a Java messaging protocol, to send ticketing through the Orbitz booking engine, which in turn handles communications with the mainframe reservation system.

Ford says American is working on a project to eliminate the reprogramming of all older systems, letting applications across the enterprise share and reuse code. A piece of code to check Sabre for available airplane seats, for example, would be available not just to AA.com, but to any application needing that information.

Ford declines to specify how the project will work, calling it "a big secret." But he says that every group associated with American—from pilots to flight attendants to customers—will benefit from "better data and faster access to data."

In fact, the pilots may now get their Web bidding system, says Captain Pinion. However, he's not sure why American's management gave the go-ahead. "You never count on something until you have it, because a lot of times they'll agree and then say it's too much money," he says. "But it's a positive."

Meanwhile, Ford expects that as American's technology infrastructure grows even more standardized, the airline will save "tens of millions of dollars in internal costs alone" across all areas of operation. One area of savings, according to programmers, would be maintenance fees on the old mainframes.

Tens of millions of dollars does not sound like much for a company that lost $3.5 billion in 2002 and $1.04 billion in the first quarter of 2003. But Ford and his information technology team must manage with the resources available. American declines to say how much it spends on technology, but acknowledges its 2003 budget is down from 2001. Its 2002 annual report shows fixed costs for information technology of $175 million in 2003, decreasing to $158 million per year through 2007.

Ford's challenges are not limited to his company's financial troubles. The day after he became CIO, on Jan. 10, 2001, American said it would acquire TWA. So Ford supervised the integration of TWA's employees and infrastructure into American, and then recovered control over software assets—including how much to charge for tickets—that Sabre had transferred to EDS.

He's still supervising the integration of what were originally 42 separate technology budgets, each with its own relationship to Sabre—a leftover from the days when Sabre was part of American Airlines.

With more than 250 technology projects in the works— pared down from 500—Ford says the group stays focused through a technique borrowed from the financial industry by his managing director of information technology planning, Susan Garcia. Called portfolio management, it treats technology projects like investments in a financial portfolio, allowing American to analyze the value and cost of starting new projects against maintaining older ones. In light of what American has been through the last two-and-a-half years, Ford calls portfolio management "a godsend."

Right now, projects are assigned to one of five buckets—Cost Savings/Productivity, Simplicity, Revenue Enhancing, Research and Development, and Mandates—which correspond in definition to the company's business goals and in size to their importance, as reflected by the amount of resources allocated for each one. Cost Savings/Productivity is larger than the other four buckets combined.

Ford re-evaluates the buckets every quarter, although Garcia says she examines them daily or weekly because she supervises 250 people and figures the tool should be more than just a reason for sending reports to the CIO.

One result: a ruthless hunt for ways to cut costs. Garcia says the group found $16 million worth of inefficiencies in the first five months. For example, she describes employees measuring the amount of traffic on a data circuit and then combing through airports to pull out redundant data lines. They rewrote software so that Sabre is checked every 10 minutes for information on connecting flights rather than every 10 seconds. They found that American had five separate software programs just to check Sabre for airplane seat maps—whoever pushed the button fastest, got the seat.

Can American remain the world's biggest airline? It's too early to tell. Low-fare rivals like JetBlue, which began service in 2000, fly without the financial burden of a hub-and-spoke network, which forces American to guess, for example, how many discount fares to sell on the Austin-to-Dallas flight when some passengers will continue on to Tokyo. American had more than 27 times JetBlue's revenue in 2002, and it has highly sophisticated software to tackle these problems.

As former American executive Robert Mann points out, the current economics of the airline industry makes such problems harder to solve. Mann worked at American for eight years and managed the start-up of the AAdvantage loyalty program. "What [American tries to do] … is utilize historical patterns of sales to sell each seat at the right price to the right customer at the right time," says Mann, now a consultant. "But the industry is so far out of its sweet spot that it doesn't know historical demand and travel patterns anymore. Planes are fuller than they have been since the second World War, when the industry was quasi-nationalized. But the market's willingness to pay seems to be stuck [at lower fares]."

So the question becomes whether American's information technology team can work fast enough, and whether its project is big enough, to eliminate the drag of its aging systems on American's financial health.

This article was originally published on 2003-07-01
Senior Writer
Based in Silicon Valley, Debbie was a founding member of Ziff Davis Media's Sm@rt Partner, where she developed investigative projects and wrote a column on start-ups. She has covered the high-tech industry since 1994 and has also worked for Minnesota Public Radio, covering state politics. She has written freelance op-ed pieces on public education for the San Jose Mercury News, and has also won several national awards for her work co-producing a documentary. She has a B.A. from Minnesota State University.

eWeek eWeek

Have the latest technology news and resources emailed to you everyday.