ITA Software: The Travel Company Everyone Uses and No One Knows Reinvents Airline Reservations, Again
In the heart of Cambridge, MA, a few blocks from the MIT campus and across the street from the missile-guidance labs at Draper Laboratories, there’s a 10-story brick office building with a brand-new corporate logo near its crown. It’s the ITA Software building, and the LED-backed logo, with its distinctive airplane-tailed @ sign, was switched on for the first time last Thursday, December 11. Scores of ITA software engineers ventured briefly into the freezing rain to watch the official lighting, then retreated back inside for a champagne toast and a slide show documenting the logo’s installation.
The lighting ceremony, the champagne, and the accompanying speeches were a bit out of character for ITA, which has never been the type of organization to call attention to itself. But for the MIT-bred company, which invented the software that hundreds of thousands of people use every day to buy airline tickets online and which has quintupled in size to some 500 employees over the last four years in order to complete a massive new airline-industry project, the ceremony was a brief moment of self-congratulation.
Jeremy Wertheimer, ITA’s co-founder, president, and CEO, invited me back to his office after the champagne for a rundown on the company, which I’ve been meaning to profile ever since we ran a piece one year ago that described its unconventional campaign to recruit brilliant programmers. The campaign included ads on Red Line trains with Googlesque puzzlers like “Write a program to compute the sum of all the integers between 1 and 1011 both divisible by seven and, when the decimal digits are reversed, are still divisible by seven.”
For all I know, Wertheimer wrote that one himself. He’s got a PhD in computer science from MIT and a huge picture of artificial-intelligence pioneer Marvin Minsky on his wall. He’s won numerous CEO-of-the-year-type awards from high-tech organizations around Massachusetts, and, in one of the largest single venture capital deals in New England’s history, he talked a syndicate including Battery Ventures, General Catalyst, Spectrum Equity, PAR Investment Partners, and Sequoia Capital into investing $100 million in ITA in 2006.
Under Wertheimer’s leadership, ITA has also figured out how to make money in an industry—the airline business—where consistent profits have been elusive. The company’s first big product, and still its anchor, is QPX, the system you see when you shop for tickets at the websites of airlines like Alaska, Alitalia, American, Continental, Hawaiian, United, and US Airways, or when you go to airfare comparison shopping sites like Orbitz, Kayak, Farecast, CheapTickets, FareLogix, Hotwire, Sidestep, or Cleartrip. Whenever you search for flights on one of those sites and get a matrix of itineraries and prices sorted by departure time, airline, or price, it’s QPX under the hood. When ITA licensed QPX to airlines and other ticket distributors, it had the foresight to charge not by the installation but by the transaction—meaning part of the price of nearly every airline ticket you buy online goes to ITA.
Clients are happy to pay ITA’s fee, because QPX doesn’t just show options to ticket purchasers, but balances a staggering number of variables behind the scenes—such as seat availability and demand on various routes—to help the airlines set the most effective price for each seat. “If you look at the way people bought airline tickets about 16 years ago, there was no good way to even say ‘What is the cheapest way to get from here to there,’ because it wasn’t required back in 1960,” when the first computerized airline reservation systems were designed, Wertheimer explains. “You needed to have somebody new come along, who didn’t have a pre-existing system. That was what we did. We solved the hard problem of how to sell this perishable good—an empty seat that, by the time the door closes on a flight, is worth nothing—at the highest price somebody is willing to pay at any given moment. And it turned out that everybody wanted that.”
Now the company has embarked on its big second act. QPX helped the airline industry reinvent the ticket purchasing process and save massive amounts of money on live sales agents and travel-agent commissions. But the company’s new all-encompassing reservation system, code named Polaris, touches almost everything else an airline does, from tracking frequent-flyer mileage and issuing reward tickets to automatically rerouting passengers and bags waylaid by weather or mechanical problems. It runs on PC architectures instead of mainframes, uses standard SQL databases, incorporates Web-style interfaces, is built around Internet-based communication protocols, and is generally designed to bring airlines fully into the Web era.
“The airline industry computerized early, in the early 1960s, because it was difficult to see how you could run a thing with millions of passengers and thousands of connections every day without having a broad computer system to tie it together,” says Wertheimer. The first such computer system, called SABRE, was developed by IBM and adopted system-wide by American Airlines in 1964. The system gave American a huge advantage, and was quickly copied by the other airlines. Incredibly, though, the code written by IBM is still at the heart of all four of the main airline reservation systems in use today. And despite the advent of systems like QPX, the limitations of the legacy systems make it very difficult for airlines to experiment with new pricing models that might increase profits—for example, by giving customers simple forms of flexibility like the option of paying an extra $50 for the right to change the date of a return flight without penalty.
In 2005, Montreal-based Air Canada asked ITA to start from scratch and build a new reservation system that would have this flexibility. That took some courage, because “you have to change almost every process at the entire airline,” says Wertheimer. “Luckily, Air Canada wanted to do that. The benefit is that it supports a more varied model of what your product is. There’s a story in the industry that an airline asked its computer system provider if it could add a fourth cabin to the traditional first class, business class, and coach cabins. They were told it would cost nine figures, because the three-cabin model was hard-wired so deeply into the code.” With Polaris, by contrast, almost nothing is hardwired—it’s all just data, bundled in XML-style packages with easily modified descriptions. “If somebody at Air Canada says tomorrow we’re going to have 17 cabins, no problem.”
Of course, airlines don’t have a stellar record when it comes to changing over to new computer systems—the debacles with United’s automated baggage system at Denver International Airport in 1995 and with British Airways’ Terminal 5 at London Heathrow earlier this year are seared into many passengers’ memories. “No one has ever switched from a mainframe-based system to a PC-based system before,” Wertheimer acknowledges. But he says airlines have plenty of experience switching from one mainframe-based system to another—it just takes lots of training beforehand. He thinks Air Canada will be ready to weather the cutover when the time comes.
Creating a team large enough and smart enough to build Polaris is the main reason that ITA has filled up six entire floors of its building—hiring, until very recently, 10 new software engineers every month, using standards not far short of MIT admissions departments’. Polaris is “another one of those insane kinds of projects,” Wertheimer says. “To me, innovation is doing something where you really don’t know how hard it’s going to be, but you’re willing to try it anyway. We have been able to attract otherwise sensible people to come to work on this problem just because it’s so crazy hard.”
One of those people is Dan Weinreb, a veteran software engineer who joined ITA in 2006 after a stint with BEA Systems. In a blog post from last December, about his decision to take the job, he admits he “didn’t have any a priori interest in the airline software field,” but says that he’s fascinated by how software can help real-world organizations process millions of transactions without error every day. In any case, Weinreb writes, his primary job criteria is “that I get to work directly with extremely good software engineers who work well together.”
The Polaris project fit the bill. For Air Canada, Weinreb says, he and his colleagues spend a lot of time working on “high availability: making the system stay up all the time, despite any kind of failure that we can reasonably anticipate. I have been focusing specifically on the problem that we call ‘hot upgrade': how to install new versions of components of the system, while it’s running, without impacting latency. This is very challenging and a lot of fun.”
Regarding Wertheimer specifically, Weinreb says he’s “the best [CEO] I’ve ever worked for. He knows how to run the business, about the industry, and how to hire great people.”
Wertheimer says the code development for the Polaris project will be done “over the next few months,” but that Air Canada hasn’t announced a specific date for implementing the system. Since nobody these days builds a giant software system for just one client, ITA has a lot riding on the debut.
“It will be interesting to see how long it takes before customer number two signs up,” says Dave Baggett, ITA’s co-founder, who sat in on my interview with Wertheimer (Baggett also went to MIT and was the developer behind one of the very first successful Sony Playstation games, Crash Bandicoot). “Once customer number one is out there, there’s less courage required.”
“We have lots of potential number twos coming by our sales office every week and hovering around,” Wertheimer asserts. “Nobody wants to be number three. In the airline business, it seems that there are only three numbers—one, two, and infinity.”
In practice, of course, there will probably be lots of number threes, just as there were for QPX. “Five years ago, people in the airline business still thought they might get another 20 years out of their existing systems,” says Wertheimer. “But over the past few years, they’ve pretty much decided that they need new ones. Everybody is looking for somebody to be the crazy first person to jump into the water, because it might be pretty challenging. But when oil was at $140 a barrel, companies were under tremendous stress, and they were becoming pretty innovative. They aren’t in such dire straits now, luckily, but the fact that oil is back down to a reasonable level doesn’t mean that your airline isn’t going to go out of business next year. So it’s not a bad time to innovate.”