Netflixed

by Gina Keating

  Although most customers ordered three disks at a time, gathering and shipping each order in a single mailer proved so time-consuming that Dillon’s analysis showed it was cheaper to ship each disk separately, as it became available, instead of waiting to accumulate complete orders.

  Dillon improved on Cook and Meyer’s software program, which assigned each order to a numbered mailbag based on its extended zip code in order to speed up the distribution of orders without costly automated postal sorters. The warehouse staff, then numbering about twenty, used a handheld supermarket scanner to read a bar code on the order form and then dropped the mailers into the correct bag. The system was rudimentary, but it soon cut total delivery costs below the two dollars per disk that Dillon had targeted.

  Although Netflix continued collecting detailed information on customer behavior, a flawed analysis of that data nearly caused operations to miss a critical link between delivery times and customer acquisition rates. Randolph had assumed that overnight delivery would boost customer retention, but test after test showed no link between the two.

  Since fast delivery seemed not to matter, Dillon designed a distribution system based on a single, giant, fully automated hub in San Jose. But before he locked down the lease on the extra-large space, the marketing team discovered an undeniable correlation between one-day delivery in the San Francisco Bay Area and the rate of new customers signing up. One-day delivery seemed like magic, and people started telling their friends about it. To test the one-day theory in a new market, they created a “hub” in Sacramento, in 2000 by driving the mailers there and back each day from the San Jose distribution center, and the sign-up rate bounced up dramatically.

  The discovery fundamentally changed Netflix’s distribution and marketing plans. The costs of attracting new customers came down as people in the overnight delivery zones recommended the service to their friends; Dillon had to rethink his distribution system.

  Dillon used data he hacked from a post office compact disk that contained delivery data for all U.S. zip codes and designed a software program that pinpointed the best locations for Netflix distribution centers. The program plugged in customer addresses and arranged them all within the overnight delivery radius of the closest post office. Then Dillon sited the distribution center as close as possible to the post office’s Area Distribution Centers, where each zip code’s mail was sorted and sent out.

  Dillon calculated that it took at least fifteen thousand Netflix customers to support a single distribution center. He developed a “satellite hub” system to achieve faster delivery outside the one-day delivery zones in which trucks delivered outbound DVDs from distribution centers to remote post offices that hand carried them to customers’ homes the next day.

  The beauty of Dillon’s program was that it continually evaluated and adjusted the optimum locations for distribution centers as Netflix’s customer base grew and changed. The program meticulously tracked postal carriers’ daily rounds by asking customers to disclose when they mailed and received Netflix mailers, leading to a close collaboration between Netflix and the postal service. Dillon first shared his data with postal inspectors in hopes of improving delivery times and tracking occasional bubbles of DVD theft. The relationship later extended to helping postal inspectors nab mail thieves intent on skimming gift cards and checks. As efficient as the distribution system was, the intensive programming demands that attended its rapid growth led to the occasional glitch: Dillon once accidentally ordered the system to deliver every disk in customers’ queues simultaneously. Some delighted subscribers received as many as three hundred disks in a space of a few days.

  Dillon also engineered Netflix’s abrupt exit from soft-core pornography after Hastings delivered the news at an executive staff meeting in early 2000 that he was about to be appointed to the California Board of Education. Hastings believed that distributing adult films could be a political liability. Either the porn had to go, Hastings told Dillon, or he could not take the board seat. Dillon and the engineers worked the rest of that day and through the night delisting each objectionable movie from customers’ movie lists and from the inventory.

  • • •

  IN JULY 1999, Randolph’s marketing team scheduled tests of new concepts designed to capture customers for several “rental turns,” to try and instill in them the habit of online rental. These included a concept for a subscription program they dubbed the Home Rental Library and another they called Serialized Delivery, which allowed customers to have movies sent to them automatically, as soon as they returned a previous rental and the company billed their credit cards.

  As originally conceived, Home Rental Library let customers rent up to six movies at a time for twenty dollars a month and keep them as long as they wanted without late fees. As soon as they returned all of the movies, they could get six more disks.

  Serialized Delivery featured à la carte prices but gave each customer an account on the Netflix site, where they could store a list of desired DVDs. Rather than asking customers to simply designate a few extra titles at each rental turn, they decided to build the sortable, searchable list of desired films that Kish and Meyer had envisioned before launched. This third feature that Randolph tested that summer as part of Serialized Delivery became known as the Queue.

  Although Randolph and Kish wanted to test the three features separately, Hastings insisted that they be evaluated at the same time. The serendipity of having these three particular concepts—Home Rental Library, Serialized Delivery, and the Queue—lined up together for testing literally saved Netflix.

  Focus groups conducted that summer showed that consumers loved each program so much—all-you-can rent subscription with no due dates or late fees from Home Rental Library, the Queue, and the no-hassle exchange of DVDs from Serialized Delivery—that Randolph combined them as one offering.

  One August evening, about a week before he officially joined Netflix as its new market research director, Joel Mier looked in on a focus group in progress at a market-testing facility on Sutter Street in San Francisco as Randolph, McCarthy, and Hastings intently watched consumer interviews from behind a two-way mirror.

  Each time a test subject dismissed the combined plan Hastings passed a note into the room instructing the interviewers to lower the price or raise the number of monthly rentals allowed, and to ask again. McCarthy, watching this, dropped his head into his hands as he saw profitability recede farther into the distance.

  On September 17, 1999, the combined plan debuted as a test on the Netflix Web site, offering customers the chance to choose and receive four movies for $15.95 per month. Only a fraction of the consumers who visited the Netflix page saw the offer, but the rate of sign-ups, or conversions to the new plan, told Randolph and Hastings they had a winner.

  Netflix publicly rolled out the Marquee Plan, as they named the new program, to all customers as an alternative to à la carte rental about a week later. The company ran the two programs side by side, and watched with mounting excitement as Marquee subscribers pushed up the site’s volume by 300 percent in just three months, to one hundred thousand disks shipped per week.

  Hastings lauded the new program as a “near DVD-on-Demand service” in a news release, and made one of the company’s first public knocks against Blockbuster. “Movie renters are fed up with due dates and late fees,” Hastings said. “Netflix.com’s Marquee program puts the joy back into movie rental. With no due dates, our customers can stock up on rental movies and always keep a few on top of their television, ready for impulse viewing.”

  • • •

  BY LATE 1999, entertainment industry analysts were predicting big gains for DVDs, including sales of six million players and twenty-five million movies in the coming year. Movie rental stores still hesitated to carry movies on the eighteen-month-old format, so tiny Netflix, with its pledge to offer a comprehensive selection, seemed best positioned “to revolutionize the DV
D rental market that we estimate will grow explosively to over $1 billion by 2002,” media analyst Tom Adams told investors.

  But the transition to the new digital format, with its tricky logistics and heavy inventory costs, was no sure thing, despite its soaring popularity, as some of Netflix’s better-funded rivals soon learned.

  Circuit City had pulled the plug on its DIVX format earlier that year, incurring a $200 million loss, and the Reel.com rental operation began foundering beneath heavy overhead costs in early 2000, forcing its parent company, Hollywood Video, to unload it and take a $48.5 million loss. It was a sobering lesson for Netflix, which closed out its 1999 fiscal year with a $29.8 million loss.

  The end of the short-lived format war cleared the way for record DVD releases in 2000, as all six major studios finally embraced the surviving format and the wide profits margins it promised. Although the studios still had every intention of undermining the movie rental business with low DVD sales prices, they recognized that Netflix provided the best avenue for customers to test the new format.

  The studios’ home entertainment divisions had long detested the hard bargains Blockkbuster drove on inventory, especially its famously profane general counsel, Ed Stead, who seemed to make a sport of demanding ever-greater concessions for the huge and powerful chain. Helping Netflix spread awareness of the DVD format through joint promotions represented a small risk for the potentially great reward of shifting home entertainment toward lucrative DVD sales, and it offered the bonus of antagonizing Blockbuster.

  Lowe also hoped to press the studios to accept revenue-sharing agreements patterned on deals Blockbuster’s new chief executive, John Antioco, had cut to mitigate store inventory costs. Netflix suffered from inventory shortages as the subscription plan grew, caused and exacerbated by the fact that subscribers could keep disks as long as they wanted. The costs of buying enough DVDs to satisfy the growing subscriber base would eventually crush the company unless Lowe could persuade studios to drop DVD prices drastically in exchange for a share of rental revenues.

  In the meantime, Netflix engineers had been hard at work since shortly after launch on a recommendation engine—an in-house solution to DVD shortages that would theoretically drive up retention and get more of the company’s catalog into circulation by directing customers away from the most popular films toward more obscure titles that they would like just as much.

  As a result, the recommendation engine took over the editorial team’s tasks of determining which movies to feature on certain themed Web pages, using machine logic rather than human intuition. The program ran through several criteria when considering which movies to show consumers: How many copies do we have in stock? Which titles have the most advantageous financial terms? Was it a new release?

  Eventually they decided the algorithm should choose the films individual customers would like best and display them on personalized pages Randolph had envisioned. They quickly realized that the better Netflix could predict consumer demand, the better they could manage inventory.

  Making accurate predictions proved terribly complicated, because the engineers could not at first find a mathematical means to determine why people liked a film by a certain director or actor, but disliked a seemingly similar film with the same cast and crew. Several in-house engineers, including Hastings, attacked the problem from different angles. Eventually Hastings brought in mathematicians to help formulate the underlying algorithms.

  They named the recommendation system Cinematch, and launched it in January 2000 with a “Movies for Two” promotion that promised to help couples find common ground in their movie choices. Rather than using their initial approach of assigning traits to each movie and trying to match similar films, Cinematch created customer clusters—people who had rated movies similarly.

  Cinematch noted the overlap in certain subscribers’ tastes through a five-star ratings system, then presented films highly rated by cluster members to others in the same cluster who had not previously rented or rated them on Netflix.

  The recommendation engine scoured Netflix’s inventory hourly to take into consideration how many copies of each movie it had before issuing recommendations. A Not Interested rating quickly weeded out anomalies: for example, children’s movies a customer rented occasionally for a grandchild, or an anime film requested once for a homework assignment.

  Cinematch actually built on an electronic matchmaking concept originated by the ill-fated Reel.com, in which customers could type in a favorite actor or film and get a list of related movies. Reel.com’s online catalog of eighty-five thousand VHS movies included, in addition to the movie comparisons: synopses; ratings of sex, violence, and cinematography; a summary of reviews; and links to star and director filmographies. Studio executives heralded it as “a complete electronic video store” that was bound to drive movie sales. Netflix wanted the same role in online DVD rental.

  Netflix’s board decided in 2000 that the time was ripe for an initial public offering. Hastings and McCarthy hit the road to convince investment banks of the company’s viability as a movie-oriented portal for renting DVDs, marketing theatrical releases, and even selling movie tickets. They hoped that stating a broader business strategy for Netflix would persuade investors, now cooling to dot-coms, to overlook an unbroken string of losses and fund an $86 million marketing and subscriber growth plan.

  By the time the company filed its prospectus with federal securities regulators in April, Netflix’s subscriber base had grown to 120,000, and its San Jose warehouse was shipping eight hundred thousand DVDs per month. Its losses had ballooned from $11.1 million on revenue of $1.4 million in 1998 to $29.8 million on revenue of $5 million in 1999.

  To improve Netflix’s finances ahead of the IPO, Hastings homed in on the Marquee program as the best bet for a steady revenue stream and eventual profit. He insisted that supporting both subscription and à la carte rental wasted manpower and resources and needlessly confused consumers. He argued strongly for abandoning à la carte rental in a tense meeting in late 1999 or early 2000 with the company’s executive and engineering teams. Several members of both teams had serious doubts about whether the sign-up rate for the subscription service could compensate for the loss of the à la carte business. Kish and Randolph feared the change would anger electronics manufacturers, who were still advertising Netflix coupons on their DVD player boxes. Instead of getting the ten free rentals they had advertised, the manufacturers’ new customers would have to hand over their credit card numbers to receive a free one-month trial subscription. Randolph hoped consumers and manufacturers would not perceive the change as a bait-and-switch tactic.

  “It feels like a big, risky step, and nobody knows if it’s going to work,” Boris Droutman said later, voicing the concern the dissenters felt.

  Hastings brushed aside their objections. To survive, Netflix needed to focus its resources on the model that worked, even if that meant betting the company on incomplete data and a gut feeling.

  On Valentine’s Day 2000, Netflix dropped à la carte rental, renamed the Marquee program the Unlimited Movie Rental service, and raised the price to $19.95 per month. Customers got a free month of service to test the program, which now included all the movies they could watch in a month, four at a time. Hastings told the financial press he was confident that most would make the monthly financial commitment. “We’re really only risking 3 percent of our customer base—so it’s worth it,” he said, adding, without elaboration, that Netflix expected to offer video downloading when the technology was ready.

  • • •

  THE LAST SENIOR executive team hire Hastings made, in spring 2000, was that of Leslie Kilgore, a former marketing executive for Procter & Gamble and Amazon. Kilgore, then thirty-three, took over many of Randolph’s marketing duties. She brought the same reliance on market research to the job, and in one of her first initiatives, reimagined Netflix’s logo and image ahead of it
s first big advertising campaign.

  Kilgore was an attractive and vigorous woman who seemed to pour every ounce of energy and focus into her work, in a manner very similar to Hastings’s own. While they shared a social awkwardness and apparent lack of understanding of human emotions, some considered Kilgore to be more brilliant than Hastings.

  The favor with which Hastings treated Kilgore set up a rivalry with McCarthy, who correctly saw Kilgore as the competition for the top job at Netflix. Hastings exacerbated the tension by baldly stating at a board meeting, in response to a director’s question about succession, and in front of a surprised Randolph and McCarthy, that he considered Kilgore his successor.

  Hastings invariably seemed to come down on Kilgore’s side when she demanded more money from the company’s tight budget for marketing and “business intelligence” programs. He also allowed her to consolidate marketing and public relations functions under her control, creating a fiefdom inside of Netflix that went unchallenged for more than a decade.

  Her loyal number two was Jessie Becker, a fellow alumnus of the University of Pennsylvania’s Wharton School and Stanford University’s business school who unquestioningly carried out Kilgore’s dictates, just as McCord carried out Hastings’s wishes.

  Kilgore’s work ethic was legendary, and she expected the same level of sacrifice from every marketing team member. After learning that Blockbuster had created an in-store subscription program, which she considered a frontal attack on Netflix, Kilgore spent several evenings after work personally interviewing shoppers emerging from Blockbuster stores near her Redwood City home to gauge interest in the program.

  One marketing team member recalled spending the bulk of a friend’s destination wedding in an expensive hotel room completing projections that Kilgore had urgently demanded. Underlings marveled at her terrifying ability to spot minor flaws in multipage spreadsheets from merely glancing at them. More than one of her employees sought counseling for job-induced stress.