Failure Is Nothing to Be Afraid Of



Technique follows technology. You may not know right away what to do with what you have, but one thing’s for certain—it’s not all going to work. What matters is what you do with those failures—and how quickly you do it. At Travelocity (a spin-off of Sabre, which was itself a spin-off of American Airlines), our mantra was, “Fail fast.”

PDF DownloadA systems guy told me recently that it took him nine months to implement a relatively trivial product-comparison tool. It took just four-and-a-half weeks to code—and the rest, apparently, was spent building consensus. Everyone was simply terrified that it might be wrong. I told them, “Put it up—on 25% of your servers. If it sucks, your customers will tell you. You won’t go broke in three days. You can take it down if you need to. Just do it.”

Once failure became tolerable, it changed the approach quite a bit. It wasn’t that getting things right no longer mattered, but, with the cycle time so reduced, we were getting instant feedback.

Failure is still failure, but what’s a good batting average? Most of our programmers came from American Airlines, where the culture was “Never fail.” They over-engineered everything, with triple redundancy—for good reason. When you’re talking about planes, you’re talking about uptime—literally. Mistakes have serious repercussions.

It took a while to convince folks who had lived in that never-fail world that, on the Internet, it was OK to fail. At Sabre, for example, we rolled out to travel agents a massive hotel-information product—probably 10 years too early. It died after we’d spent four years (and millions of dollars) on it. We had a lot of good people behind it, but it just became too expensive to support. When your project’s on the Internet, you can make decisions quickly because you’re centralized. But back then, we had thousands of installations of CD-ROM drives, video cards and so on. We’d spent all this money on the CD-production hardware. It was a lot harder to say “stop.”

The first thing we did at Travelocity—before it was even called Travelocity, in fact—was move to a separate headquarters. I found this old, abandoned facility that nobody wanted, which had been a slide-production facility for big marketing shows. We brought together a real rag-tag group of people who had the courage to go. Others wouldn’t come. They thought it was too risky, that failure was too likely. They only wanted to come when it was profitable, when the possibility of failure was almost gone. But by that point, we’d built it into a company with $2.9 billion in bookings and $239 million in revenue, and they were no longer invited. Latecomers hadn’t risked what we had. They hadn’t earned the right to be there.

Sometimes what stops you is completely outside your control. In 1991, we were building a big reservation system for Aeroflot, the Soviet airline. We managed to get a lot of the code done—and then the nation underwent a revolution. Aeroflot’s check was backed by the full faith of a government that suddenly no longer existed.

The Internet led to a somewhat different kind of revolution, but one that was no less drastic. Before, technology was driven by the internal needs of the corporation, but the pace was kept in check by the technology department’s ability to deliver. Those were two essentially equal gears going around. With the advent of the Internet, suddenly there was this huge gear called the market and a little gear called you. When the market turned just a little bit, you spun ’round and ’round, faster and faster. Even today, you have to respond quickly: If you’re putting your business online—and, one way or another, everyone is—customers are going to vote with their mouse-clicks. And your competitor is just an eighth of a second away.

Besides, the Web doesn’t really let you go ’round and ’round. You only get to click forward. If you go back, you get the white screen of death. Real life isn’t like that—failure knocks you backward sometimes, and you need to be able to get moving again, right away.