At 8:57, Te tapped my shoulder and said, “Remember, you’ve got a call in five. So you can watch the ball drop, but then you gotta be by the phone.”
I nodded, and out of the corner of my eye saw the door open and then close. It was Reed, slipping in just before launch. I hadn’t expected him to come, but I was glad to see him—and somewhat relieved that we were on schedule. He gave me a brief nod when he walked in, but he didn’t say anything, just stood somewhat awkwardly behind the huddle of employees in front of the computer.
By 8:59 the office was so quiet that I could hear my watch’s second hand. At 9:00 on the dot, Eric leaned over, punched a few keys, and we were live. We held our breath. Eric had hooked up a bell to his computer—not unlike the kind that businesses leave on the counter to alert employees that a customer needs help—and rigged it up so that it would ring each time an order came in. I filled out the day’s first order as a test: I, Marc Randolph, requested a copy of Casino, to be delivered to my address outside Scotts Valley. I hit Enter to place the order, and moments later the bell rang. Almost immediately, we had three others in the queue, each sounding the bell as credit cards were authorized, inventory decremented, and packing slips printed. I patted Eric’s computer for good luck and walked back to my office for press calls.
Within minutes, the bell sounded like a machine gun. Even with the door closed, even as I carried on a conversation with Steve Perez at the Santa Cruz Sentinel, I could hear it, pinging in the next room.
We got fifteen minutes.
For fifteen minutes, customers chose movies, inputted their personal information, gave us their credit card numbers, and hit the red button that said confirm. For fifteen minutes, the bell rang, orders were printed out on the pair of laser printers at the back of the office, and Jim’s team took them to the vault. For fifteen minutes, each order was matched to its movie, the disc was slid into the mailer, and an address label slapped on. For fifteen minutes, the completed orders grew into a small stack in a box by the door.
Several months ago we recognized an opportunity to create a major commerce brand in a billion dollar market, as well as to be a critical catalyst in the growth of one of the fastest-growing consumer electronics categories. This morning, Netflix opened the world’s first Internet DVD rental store: NetFlix.com. The NetFlix store carries every DVD movie—all of which are fully available for rental.
I watched it all through the glass of my office window, giddy with excitement. I’d asked Te to stay in the room with me, and to write down any questions from the press on the whiteboard I kept in my office—the same board we’d used to decide on a name. I liked to use journalists’ questions as a jumping-off point for longer, more in-depth stories—even though the beginning of each call was canned, I wanted my answers to be improvisatory riffs that really got at the heart of what we were trying to do. I’d weave in American history, pop culture, and even stories from the outdoors. But I needed an anchor, a handhold to grab—hence the questions on the board, and Te standing next to it, marker in hand, like a Silicon Valley Vanna White.
Despite phenomenal growth in the DVD market, most of the nation’s video outlets don’t yet carry DVD—and those that do, carry only a limited selection of titles, often a single copy of each. Netflix, on the other hand, carries virtually every DVD. While we don’t carry X-rated titles, we do—as of this morning—list 926 titles for rent, the largest selection available anywhere. And our warehouse contains hundreds of copies of the most popular movies, virtually guaranteeing that our customers can rent the movie they want, when they want it.
It wasn’t hard to get excited, riffing about our business, that day. I could see it through the window, right there in front of me—the dream I’d been working on, in full color.
At the NetFlix web site, we make it fast and easy to find the right movies, and we deliver them in two or three days. Customers keep them seven days and get to watch them as often as they like. When they are finished, they simply replace the discs in the envelope we provide and drop it in the nearest mailbox. We even pre-pay the return postage.
Slowly, however, I started to realize that something was wrong. Eric was frowning at his computer. Boris and Vita were typing furiously. Suresh was down on his hands and knees, grasping at something underneath the servers. Kho was unplugging and replugging cords into the wall, and tracing their looping trajectories up to the ceiling.
Eventually, Christina edged into the office, biting one of the few fingernails she had left. I’d just finished chatting with Jon Swartz at the San Francisco Chronicle.
This is a tremendously exciting prospect for us, for our customers, and most importantly, for the entire DVD community.
I set the receiver down in the cradle. That’s when I noticed it—the bell wasn’t ringing.
“What happened?”
Christina rolled her eyes. “Servers crashed.”
This is another problem that current startups don’t really have to deal with. Now, almost every web company runs their business in the cloud. Rather than the long, laborious, and capital-intensive setup that Eric and Kho had to deal with, now companies simply write a check, buying access to somebody else’s computers, stored in air-conditioned warehouses with backup power and plenty of storage. But back in 1998, cloud services didn’t exist. If you wanted to run an e-commerce site—or any website with high traffic, for that matter—you had to own the means of serving up web pages, storing data, and keeping track of customer information. That meant racks of computers in your own office dedicated to hosting your website.
We’d gone into launch day with a grand total of two of them. Corey, who had spent two years at Netscape, had tried to tell me to stock up on extras. “You’re going to need them,” he’d said. “If not for the launch, then soon after. Why not buy in bulk, ahead of time? Don’t you want to anticipate the best-case scenario?”
I did. But I think a part of me was still superstitious, worried that I would jinx the whole thing. Christina had said it best—launching the company was like throwing a party, one that you weren’t sure anyone else would attend. You didn’t want to buy extra kegs if no one was going to come.
But Corey had been right, of course. Two servers was like trying to cross the old West with a single mule. Wasn’t going to cut it.
When I walked out of my office, Eric and Boris were gearing up to make a trip to Fry’s, the electronics store over the hill in Campbell, where they’d buy eight new desktops with a whopping 64 megs of RAM each.
“That should do it,” Eric said, looking unconvinced.
“What do we do in the meantime?” Christina asked. “We could be losing dozens of people.”
“What a nightmare,” Te said. “All these press guys are gonna go to the site and there’s gonna be nothing there!”
Then Reed spoke up. It was the first time he’d said anything all morning. “Can’t you just put a STORE CLOSED, COME BACK TOMORROW sign up?”
We’d grown accustomed to calling Netflix “The Store.” It made sense—what we were trying to offer was an e-commerce version of what Mitch Lowe and his family were doing at Video Droid. But unlike a storefront, a website can’t hang an OUT TO LUNCH sign on the door. The internet doesn’t have business hours.
“Did we build an error page?” I asked.
Christina’s face fell. “No,” she whispered.
“Well, let’s do it,” I said. For the next forty-five minutes, while Eric and Boris were buying new servers, we built a cheeky “We’re down but not out” page that would reassure customers that they were in the right place—and that we’d be right back.
That page got a lot of views that day.
An hour later, Kho hooked up the eight new servers, essentially quintupling our capacity for new orders. Everything worked fine—the site was up and running, the orders were flying in—for about forty-five minutes. Then the servers crashed. Again.
And again, Eric and Boris took off for Fry’s. I didn’t go with them, but I can imagin
e it even now—the two of them riding sternly to the store in the rusty pickup truck that belonged to Greg Julien, Netflix’s company controller. Pushing a shopping cart directly to the computer department, then dealing with the same checkout person as they discussed among themselves whose credit card to use. That clerk had probably seen this exact thing happen dozens of times, with dozens of startups. We were in Silicon Valley, after all.
The site crashed all day. And because we had no way to measure site traffic yet, we didn’t know how many potential customers we were missing.
It was a disaster. But at the same time, these were good problems—we had visitors to the site, we had orders coming in.
“People are coming!” I found myself saying in amazement. “They’re coming to the site, and they’re giving us their credit card information!”
When we’d moved into the offices, I’d bought a bottle of 1995 Ridge Estate Cabernet Sauvignon—a California wine that was about a hundred dollars more expensive than the bottles Lorraine and I usually bought. (Translation: It was $120.) I’d told everyone that we’d open it once the website had gotten a hundred orders, and took a straw poll about when the rest of the office thought that would happen. The shortest guess was from Suresh, who was working on inventory and order entry. He guessed less than a day.
I guessed a month or two.
Guess who was right?
“Good call, Suresh,” I said sometime after 2:00 p.m., when the hundredth order came in. I flipped him a silver dollar, which he caught without even taking his eyes off the screen.
It’s what we’d all hoped for, of course. But it still was astonishing, in the moment, to see it. As I watched the orders come in, and listened to the printer printing them out, I had an enormous sense of relief. Our big reveal hadn’t been greeted by an orchestra of crickets.
It was popular. In fact, it was a bit too popular.
We ran out of boxes. We ran out of tape. We ran out of paper. We ran out of ink.
After the fortieth printer jam of the day, I walked to Corey’s desk and asked him if he could slow things down a little. The servers were down, the printers were jammed, and we had Christina’s entire content team typing individual confirmation emails for the orders we’d already received. (Turns out an automated email should have been higher on the to-do list.)
“Think you can hold off the nerds for a little while?” I asked.
Corey laughed. “I’ll try.”
He paused.
“But they’re really into it,” he said.
Increasingly, as the day went on, one deadline began to loom large: 3:00 p.m. That’s when the Scotts Valley post office packed all their mail into their trucks and headed over to San Jose. If we wanted our DVDs to go with them, we had to have all of our shipments there—processed, packaged, and addressed—by 3:00, or the same-day shipping we’d promised our users would turn into next-day shipping.
That was unacceptable to Jim. But as the day went on and the orders rolled in—as the servers crashed, and the printers jammed, and Christina’s team gave themselves blisters writing confirmation emails to everyone who ordered a DVD—he was getting nervous.
“If we get jammed, we can drive everything down to Santa Cruz,” Jim said. “Their last pickup is at four.”
Jim had done research for weeks on pickup times, post office hours, and routes. He knew that the DVDs we dropped off, presorted by destination, would travel first to San Jose, then to all of the destinations we’d seen on orders that morning—San Diego, Seattle, San Antonio. But first they had to get out of our hands.
“If I leave at two fifty-two,” he said, “I can get to the Scotts Valley PO with a minute to spare. If we’re not ready by then, I can go to the Santa Cruz post office—but it’ll take me twenty minutes to get there, and who knows if there will be parking. So I’d have to leave at three thirty to be safe.”
I knew that Jim was just thinking out loud. He’d driven to the post offices a half-dozen times in the weeks before the launch, trying to find the quickest route. Once he was there, he’d familiarized himself with the parking lot and the drop-off location. In an extreme gesture of optimism, he had even put a handcart in the back of his pickup that he could use to wheel up boxes if the orders were too heavy to carry. He’d already scouted out where the handicap ramps were, in case he needed to use them.
“You do whatever you think is best,” I said. “But it would be kind of nice to use the hometown PO for the first run, wouldn’t it?”
Jim nodded. We’d stepped into the vault by then, and two of Jim’s team were busy flipping through the hanging discs, searching for movies to fulfill recently placed orders. I picked up an order slip from the table near the door and joined them, searching the alphabetized wall for a DVD copy of Heat. I walked past it several times before I saw it. I bumped into Jim’s workers at least twice.
“You’re hopeless, Marc,” Jim said, grabbing the DVD from me and expertly shoving it into a mailer. He affixed an address label with finesse and expertly sealed the flap. “Now get out of here. We’ve got forty-five orders more to fill before the post office closes.”
The clock on the wall of the safe read 2:24.
It was stressful until 2:52, when Jim left for the post office. Then the whole office relaxed. The day’s deadline had passed. Now it was just time to figure out how to make things work better the next day.
We’d expected 15 or 20 people to use the site to order a DVD. We’d gotten 137—and potentially we’d gotten more than that, since we didn’t know how many people had tried to access the site when it was down.
It was an enormously promising start. But that’s all it was: a start. There were hundreds—no, thousands—of changes we still needed to make.
Did we open the bottle of wine? We didn’t have a corkscrew, so I had to push the cork into the bottle using a ballpoint pen, then decant into an empty liter bottle that used to hold Diet Coke. And we had to use red Solo cups instead of wineglasses. But we opened that bottle, and all shared a brief toast in the conference room. I looked for Reed but didn’t see him—he’d slipped out sometime in the afternoon.
“To beginnings,” I said. “To the work ahead.”
And there was a lot of it. We needed to automate confirmation emails. There were dozens of problems with the online ordering form, which, it turned out, was fine at catching bad state codes. Not so good at validating zip codes. And clueless for international orders. (Who knew people would try to order from other countries!) We still needed an algorithm that would ensure that high-demand titles were always in stock—and figure out how to steer customers to lower-demand titles in a way that made them actually want to rent them.
There were thousands of puzzles to solve, and we all knew that we’d spend months solving them. So after the toast, we all crumpled our Solo cups into the recycling bin and got back to work.
Around six or so, someone ordered pizza. I left around ten. The engineers would probably be in the offices all night, working to ensure that the next day’s traffic wouldn’t crash the site. And of course the site didn’t shut down overnight—you couldn’t just turn the neon sign off and get back to work in the morning. All of us realized, then, that our work with Netflix was entering an entirely new stage.
That night, I sat again at the table in my unfinished kitchen. The kids were asleep, and so was Lorraine. I was still antsy, high on the adrenaline of the day. When I’m like that, there’s no point in trying to sleep. So I pulled out my notebook and started writing down a list of all the things we needed to work on:
Site Redundancy—how do we gracefully recover from crashes when a server goes down?
Get better Packing slips—peel off labels keep coming off in the printer.
More inventory? How many is too enough? How many is too many?
Need Metrics! Get Suresh to report on today’s orders by source and title. What else?
As I thought of possible solutions, I idly lined up a few slats of wood we’d left on the tabl
e. They were 120-year-old pieces of redwood that we had reclaimed from some of the house’s original flooring, and Lorraine had been thinking of using them as shelving. I lifted one, felt its heft, the lines of the wood grain. I tried to imagine it on the wall behind me, which was covered in paint samples for the eventual renovation. I could almost see it.
We were still building our kitchen, even as we lived around it. Just like Netflix, I thought—we’d built it, but it wasn’t finished yet. It would probably never be finished, truthfully. Every day, we’d have to work to keep it upright—to keep the water flowing, to keep the cabinets filled. To keep the burners clean and the gas bill paid.
But it was there, now. It was out in the world.
Years ago, on a climbing trip, I was hiking across a snow plain just below the summit of a mountain when I felt a peculiar static buzzing around my head. My hair stood on end, and around my helmet there was an ultraviolet glow. It was Saint Elmo’s fire—a positively charged electromagnetic field that was about to discharge itself to earth. It was lightning, just before the strike.
That’s how Netflix had felt, all that spring—a blue haze buzzing around all of our heads. But starting on April 14, Netflix wasn’t just potential energy anymore. It was a live current, positive meeting negative. It was a lightning strike.
And now we had to figure out how to manage it.
9.
A Day in the
Post-Launch Life
(summer 1998: eight weeks after launch)
5:00 a.m.
“IT’S YOUR TURN.”
After this proclamation, Lorraine rolls over and folds the pillow over her head.
That Will Never Work Page 12
