Things A Little Bird Told Me

by Biz Stone

  I did exactly as he instructed me, and it worked. When I went past the point of no return, it was effortless.

  The same is true for making a big move in your life. Asking a girl on a date, particularly if she brings a fifth wheel, means you’re risking embarrassment and failure. Deciding to quit your job, particularly if it means leaving behind valuable options, means you risk financial ruin and still more failure. But when it works out, isn’t it fantastic? When I successfully executed the back handspring, I was in awe. It was all about being willing to fail, just like in Gattaca.

  One of the primary reasons Twitter was constantly breaking was that it was originally built quickly, as one big, messy program. It wasn’t a distributed architecture, which meant that it was like a house of cards. If one piece slid out of place, the whole thing fell apart. And each time we wanted to figure out what had gone wrong, we had to sort through the entire system. After hours of forensic study, we’d locate the broken piece and then have to determine who wrote that part of the program. If that person was out sick, tough luck for us. We were disappointing people who used Twitter and getting hammered in the press.

  Then one day I was watching an old episode of Star Trek: Voyager—“Demon”—when it gave me an idea. The space station is almost out of fuel. The captain orders that they go into “gray mode.” Gray mode meant shutting off all nonessential systems in order to use as little power as possible. Essentially, they were putting themselves on life support.

  All the systems on the Voyager are compartmentalized. They can shut off parts at will, while the ship continues to operate. (I should add that this is kind of an obvious revelation.) The way we built Twitter was not ideal, but it also wasn’t a mistake. We weren’t planning on the massive success that came in bigger and bigger waves. Nor should we have. It’s better to get it built and out in the world than to take years to make it perfect before you know if it’s going to work.

  I went to work the next day determined to suggest a new approach to Twitter’s failures. Jason Goldman had come over from Blogger to be VP of product development and Evan’s right-hand man. Luckily for me, Jason was also a Star Trek nerd. (The two of us are Trekkies.) I asked him, “Can we make it so that we separate elements of our system into different pieces—like registration, updates, and certain server requests—so that if one piece is going to collapse we can turn off just that section and at least something will work? Then we won’t go down completely every time something goes wrong. You can still see the home page. You can still tweet. Could we create a gray mode?”

  The answer was yes. That very week, we put together a rudimentary version of compartmentalized features. Now we didn’t have to go down completely for every little problem. Star Trek: The gift that keeps on giving.

  One of Twitter’s biggest failures was our so-called platform. In 2007 we released our platform, a collection of APIs (application programming interfaces) that allows third-party developers to make use of Twitter’s technology. We loved the idea of inviting developers to build apps that would enhance or complement Twitter, but we didn’t think it through enough.

  As soon as we released the platform, tons of new Twitter apps sprang into existence. The glut of options muddied the user experience. And allowing all these apps to make essentially unlimited requests to our server hit the service hard. This was a big part of Twitter’s stability issues. The developer platform was heavy, expensive, and often contributed to Twitter’s breaking down.

  When Facebook came out with f8, its platform, I believe it experienced some of the same problems. In the first six months, Catherine Rampell of the Washington Post reported, seven thousand new apps had been launched. It was overwhelming, and Facebook had to pull back, most likely by slowly but surely introducing rules and restrictions. Now most of the apps on Facebook are made by Facebook.

  Designing book jackets, I had learned that the perfect one satisfies multiple criteria. It pleases Design, Editorial, and Sales. Similarly, a successful software platform should, first and foremost, serve the consumers. Second, it should enrich the developer community so they can make a living while creating fun projects using parts of our code that we’d decided to make public. Finally, it should feed back into the overall value of Twitter, making it a better company and service. Those goals should have determined what we released. Instead, we opened the floodgates. When we had to close some of them later, we upset a lot of people.

  We didn’t have our eye on the ball. We could have started slowly, releasing specific options for developers to work with that might help make it joyful and easy for users to discover accounts to follow that they wouldn’t otherwise have found. But we didn’t take a measured, critical approach. The results were damaging for the service, our users, and independent developers. Some failures aren’t risks that didn’t pan out. Some are just plain old mistakes. All we can do is be honest about them and learn from them.

  After SXSW, when I realized Livia and I had been together for so long, I said, “You know what? We should get married.”

  Livy said, “No shit.”

  Apparently she’d been hinting at it for a while. She’d say, “Look, they’re getting married. They’ve been dating for less time than we have.” Subtle, huh? But I was characteristically clueless.

  Nonetheless, emboldened by her promising, albeit obscene, response, I rose to the challenge. After giving a talk at the NASA Ames Research Center, I bought her a cheesy NASA mood ring to serve as a temporary engagement ring.

  Livia and I originally intended to elope. We just didn’t want the hassle of a wedding and had discovered a beautiful bed-and-breakfast on the coast in Mendocino, California. Somehow, a few dozen friends showed up as our “witnesses,” and it turned into something between an elopement and a simple wedding that left out family. This way, we experienced the magic of an easy wedding along with a lifetime of disappointed and angry family members who felt betrayed and abandoned.

  Nevertheless, in June 2007, we had a beautiful service in a garden on a bluff above the Pacific. Within a minute of the ceremony, my friend Dunstan snapped a Polaroid photo. It’s my favorite picture from the wedding. I’m wearing my linen suit, throwing my head back a bit with a huge smile. My wife is wearing a vintage 1920s evening dress, but we can’t see her expression. Her head is down, her face buried in her hands.

  While I look like the happiest guy in the world, my wife’s body language suggests a woman who has just made the worst mistake of her life. Like she’s saying to herself, “What have I done?” Some of the best things in life, I have assured her, come from mistakes. In fact, Ben Franklin once said, “Perhaps the history of the errors of mankind, all things considered, is more valuable and interesting than that of their discoveries.”

  To this very day, my wife and I remain happily married. As far as I know.

  Throughout the first years after Twitter exploded at SXSW, the service experienced major connectivity problems. We crashed. A lot.

  Companies like to put forward a persona of perfection. “We have the best rates!” “We do the best work!” “We’re awesome!” “You should choose us!” “We’re world-famous in Poland!” That’s normal. But it’s also a very safe, contrived path. What if you do fail? Or only sort of succeed? Do you still send out those relentlessly positive messages? You don’t want to advertise your failings, but to hide them is, on some level, deceptive. This brings me to the value of vulnerability. When you let people understand that you are people like they are, passionate but imperfect, what you get in return is goodwill.

  Take the actor Harrison Ford. (Again, but why not? He’s a great actor.) He usually plays the hero. Traditionally, heroes are fearless, strong, and pretty much bulletproof. But Harrison Ford plays it differently. Whenever something really bad is going down and they do a close-up on him, he looks either scared or like he’s thinking, Oh, God, I can’t believe I have to do this now. In Raiders of the Lost Ark, faced with a pit of writhing vipers that he has no choice but to pass through, he fam
ously says, “Snakes. Why did it have to be snakes?” There’s no bravado. The hero he gives us is a regular guy . . . and now he’s in a snake pit. He’d better figure out how to escape, and fast, if he’s going to save his skin. As a viewer, you’re far more invested in his survival and success because he lets you see his humanity.

  For the past decade a large part of my job has been explaining to people why something broke. At Google, when I worked on Blogger in its early days, it broke a lot. I assumed responsibility for explaining to the people using the service what had gone wrong, why it had happened, and what steps we were taking to make sure that particular problem wasn’t going to happen again.

  One time, in 2003, when Blogger crashed, I set about investigating the cause. Someone finally explained to me that what it came down to was electricity. Google was so huge that it required a vast amount of electricity to support its data centers, the facilities where its computer systems are stored and maintained.

  It turned out that when Google ran low on electricity, Blogger wasn’t high on the priority list. So we got turned off. I’m simplifying, but that was the basic idea.

  When I found this out, I wrote a post on the official Blogger blog explaining that Blogger had gone down because Google was so huge that we’d run out of electricity.

  Posting to the company blogs was a big deal, but I always treated it with some irreverence. One of my proudest achievements was that when Blogger launched photos, I used a picture of my cat Brewster as an example. Google was a big, fancy company about to go public, and yet I managed to put a cat picture on its official Google Inc. blog. I wasn’t just amusing myself. I saw it as my role to put a human (or cat) face on our technology.

  The Brewster post went by without comment, but my post about electricity was a bigger deal. Where was Google going to get more power? What I didn’t know when I posted about the electricity was that Google had been involved in a secret project. Through a third party, it was acquiring huge swaths of land east of Portland, Oregon, where it planned to create its own power centers. Investors and the media watched the company with hawk eyes. When my post went up, the sleuths of the internet took interest and pieced together what Google was up to. Luckily, I didn’t get sent to the principal’s office on that one, but sometimes being completely truthful has consequences.

  Even so, I believe in honesty, and I believed that explaining mistakes to the people who used our service was the best way to create a long-term relationship with them.

  I brought this philosophy to Twitter. At first, I didn’t have a master communications plan other than my instincts. I wanted us to let everyone in the company know what we were doing and what we were planning to do. I wanted the external communications to be the same as the internal, minus anything we couldn’t legally discuss, or stuff that wasn’t classy, such as specific amounts of money raised in financing. I wanted the opposite of a PR team that would spin everything. There should be a universal truth.

  Then it became painfully apparent that the service we’d built could not handle its rapidly growing audience. Having made it my job to communicate when things went wrong, I found myself very busy.

  I explained any problems to the people using the service (assuming the service was working). If the system crashed, I’d walk down to the engineers to investigate what had happened. Then I’d go on the Twitter blog to report what I’d discovered. Most of the time, I found a way to present it as good news—in the sense that we’d figured out what was wrong and could therefore promise that it was unlikely that particular, exact problem would happen again. (Something else was definitely going to cause the system to crash, but probably not that.)

  It didn’t take long to see the results of my approach. Apple has an annual worldwide developer conference. The first people using Twitter were the same people who were eager to hear about new products and technology Apple might launch, and before the June 2007 conference, there were rumors flying that Apple was going to announce an iPhone.

  The day before the conference, all the chatter about the iPhone put a strain on our service. There were intermittent outages, and we—and our users—started to worry that Twitter wouldn’t hold up for the announcement the next day.

  That night, we stayed late to work on the problems. Our users knew us, and they assumed, correctly, that we had to be killing ourselves trying to bolster the service for the next day. Late at night, a few pizzas arrived, and then a few more. But nobody in the office had ordered pizza.

  Then a user tweeted,

  Did you get our pizzas?

  Oh my God. We were getting support from the Twitter community.

  Instead of complaining that the site was down, several different people had sent pizzas over to our office to lift our spirits and support our efforts. We weren’t some anonymous robots who frustrated them with our bugs and glitches. All our honesty had revealed our humanity and brought us goodwill.

  Twitter continued to fail. It was up to me to decide how to handle this internally, and with our users. I wanted people to know that we were doing our best, but I didn’t want to try to hide or downplay our flaws. I decided that we would own our many imperfections.

  In an early version of the system, when you sent a text, a “success screen” came up. On most websites, that screen would have read, “Thank you. Your message was sent.” On Twitter, our success screen read, “Great, that might have worked.”

  I’d always wanted to acknowledge the feelings of the person on the other side of the screen. Back at Odeo, when the system crashed, a dialogue box came up where you had to click OK in order to continue. I asked Jack to add a checkbox. Then, in addition to clicking OK you could also put a check next to a line reading: “But I’m not happy about it.”

  Eventually, in order to soften the blow of Twitter’s outages, I poked around on a stock photo website and found an image of a whale being lifted up by a bunch of birds. Perfect! I put it on the Error page.

  The Fail Whale, as it came to be called, was a happy, positive image, and it portrayed us as a small but committed group of birds managing, as a team, to carry the weight of an impossibly huge whale. We were small, but we were determined to succeed.

  And here’s the kicker. There was so much fuss about Twitter’s outages that the Fail Whale became a meme. There were fan clubs. There were Fail Whale parties. A guy even got a tattoo of the Fail Whale on his ankle. There was a Fail Whale conference, and I was asked to keynote! People who’d never heard of Twitter started hearing the complaints. What was this service that people liked so much they couldn’t bear to live without it? I don’t have any scientific evidence, but I believe all the fuss caused more people to check Twitter out. I wouldn’t be surprised if the Fail Whale actually contributed to our growth.

  Our failures became an asset.

  Sometimes we got nasty complaint emails at Twitter. They often said, essentially, “You assholes don’t know what you’re doing.”

  My favorite thing to do when I got these mean emails was to respond with kindness in equal amount: “Dear Joe, thank you so much for your feedback. I’m as frustrated as you are when the service goes down. I’m so glad you sent me this note. Here’s what the guys are doing. Please let me know if it doesn’t work for you in four hours.”

  Inevitably, I would receive a sheepish, apologetic note saying, “You guys are awesome. I only wrote that email because I really like you.”

  Receiving these email responses helped me realize that the loudest complainers were often our biggest fans. The only reason they took the time to write was because they had passion for our product. By responding personally and honestly, I was letting them know that we cared, too, and that there were real people behind the scenes doing everything they could to help those poor birds move that whale. It doesn’t pay to act bulletproof. Nobody is flawless, and when you act as if you are, it always rings false.

  Not only did we encourage angry people to email us, but I put my cell number on the home page of the site and answered t
he phone when it rang. People would call me for basic support stuff, asking how to log in, how to change an avatar, or whether they could switch usernames but keep their Tweets. One Saturday morning at six o’clock, the phone woke me up. When I rolled over in bed and answered it, an old man’s voice said, “Yeah, my church told me we’re supposed to use Twitter.”

  I said, “Okay . . .”

  He said, “So I solved the word puzzle.”

  This confused me. The day before, I’d had an idea for a multi-player word game that could be played over Twitter. I wanted to call it Wordy. You’d text “play wordy” to 40404, and you’d be given seven letters with which to spell the longest word you could. I’d told Evan the idea the day before, but how did this guy know about it?

  “You solved the word puzzle,” I repeated.

  “Yeah, so now what am I supposed to do?” he said.

  Slowly waking up, I realized that he was talking about having filled in the CAPTCHA, the distorted word image you had to match at registration to verify that you weren’t a bot.

  Caring about our customers meant caring about every single customer’s experience, day and night. I explained to the good man how to use Twitter, how to follow people, and how to explore the site.

  My contact info was up on our home page until the phone calls were almost exclusively reporters. Then I changed my number.

  Over and over again, through the Error messages, the blog postings, the Fail Whale, and responses to user mail, I told people that we were human. We knew we were failing. We didn’t like it. But I wanted us as a company to believe what I had seen in Gattaca, what I had learned when I took a chance with Livia, and what the success of Twitter would eventually prove to me: Failure was part of the path. It was worth the risk. In fact, it was a critical component of growth. By sharing it with our users, we were showing our ultimate confidence in ourselves and our success. We weren’t quitting, and we hoped our faith would inspire theirs.