Demand Media's Planet of the AlgorithmsBy
"We are thinking through the implications of, 'How do you editorially program the planet?' " says Byron Reese.
It's March 2010, and Reese, the chief innovation officer of Demand Media (DMD), is accepting a "game changer" award for innovation at the We Media conference in Miami. Several dozen people listen as he invokes Demand's plan for cranking out Web pages and videos to quench every last bit of human curiosity wherever it springs up. The plan's code name: Little Brother, a nod to George Orwell.
Reese does not share many details about how Little Brother will work. Instead, he shows an image from The Terminator. "Perhaps at this point you're thinking, 'I know where all this is going,' " says Reese. " 'I've seen this movie before. The machines are going to start making the decisions. We're going to be ignorant. They're going to take over.' " He assures his audience that Demand Media will not annihilate humanity.
Over the past decade, Reese has quietly pioneered a new breed of media company, colloquially called a "content mill." Where traditional media companies rely on creative professionals to generate ideas aimed at loyal repeat readers, content mills are far more transient. They rely on crowd-sourced stories and search engine optimization, the art of gaming online search results to ensure one appears at the top, to rope in drive-by users looking for a quick hit of information—how to hang a door, say, or make sourdough starter. The pedigree of the source providing it is not important.
Large media companies, including Yahoo! (YHOO) and AOL (AOL), are introducing variations on the content mill model, hoping to draw more traffic, and in turn more advertising revenue, while lowering the costs of content creation. Demand Media is the fastest-spinning mill of all, currently producing more than 5,000 articles and videos for the Internet every day on topics often selected by computers. A decentralized horde of 13,000 freelance writers, editors, and producers earn minimal fees (around $15 for writing an article of a few hundred words) to keep the production line humming. The resulting stories are distributed across Demand-owned websites such as eHow.com, Answerbag.com, or Livestrong.com and are hyper-engineered by specialists to appear at the top of search engine results for a vast number of queries. The stories, which are not tied to news events and are thus designed to have long shelf lives, tend to be coherent but uninspired. The potential for monetization is said to be vast.
"It's a great business model," says Ken Doctor, an affiliate analyst for Outsell. "They've been pioneers in defining a new content creation model based on harnessing so much low-cost content and matching it better than anyone with targeted advertising."
"They really understand consumer behavior on the Web and how to build businesses on it," adds Sheryl Sandberg, Facebook's chief operating officer.
"I feel privileged to be an investor," writes Aviv Nevo, head of venture capital firm NV Investments and one of the largest holders of Time Warner (TWX) stock, in an e-mail. "I am confident that the company will continue to be the premier leader in the creation of rich, focused and tailored content."
On Jan. 25, Demand Media sold 8.9 million shares at $17 each in an initial public offering. The following day, the price rose 35 percent to $22.61, which would give the company a market capitalization of $1.9 billion, greater than New York Times Co.'s (NYT) value of $1.5 billion.
The company spent the past few months in a pre-IPO quiet period. As a result, Reese and Chairman and Chief Executive Officer Richard Rosenblatt, the company's press-friendly face, were unable to comment at press time. According to the prospectus, Demand plans on using the influx of capital for product development and international expansion. Executives have said in the past that someday Demand hopes to crank out as many as a million articles and videos every month.
At the Miami conference, as Reese finishes up his Terminator presentation, he makes it clear that, while easily parodied, he doesn't view himself or Demand as a bringer of intellectual destruction. "This is a frightening notion of the future—that somehow in doing this we lose our fundamental humanity," he says. "I don't think this is the case."
One of Demand's primary selling points to investors is that in just a few years of existence it has already built a huge Web audience. Its prospectus promotes the fact that its sites attracted 105 million unique visitors in November 2010, according to ComScore (SCOR) data, making it the 17th-largest Web property in the U.S.
Raw traffic numbers tell only part of the story. Many analysts now say the best way to gauge a Web company's financial hardiness is to look not at page views or monthly unique visitors but at how much money it can generate from each user. Matthew Shanahan, senior vice-president of strategy for Web consulting firm Scout Analytics, says thriving digital ventures typically have a high ARPU (average revenue per user). Amazon.com (AMZN), for instance, makes on average $189 per unique user. Google (GOOG) takes in around $24. Web publishers, Shanahan says, tend to become reliably profitable at about $10 a user. Demand's average revenue per user currently hovers around $1.60.
Since launching in 2006, Demand has yet to turn a profit. In 2009 it lost $22.4 million on $198.4 million of total revenue, according to its prospectus. In the first nine months of 2010, it lost $6.3 million on revenues of $179.3 million. While it has a lot of unique visitors, they tend to have minimal exposure to advertising, because the average user bounces into a Demand site—and bounces out just as quickly.
According to data from Web research firm Alexa, the average visitor to eHow, Demand Media's most visited site, spends only 2.3 minutes and clicks on 1.7 pages before departing. By comparison, Facebook users average 31.9 minutes and 12.7 page views per visit.
"There is no brand loyalty there," says Shanahan. "They are relying almost purely on bringing in new customers using search engine optimization, which attracts a fly-by audience. As a result, you have this huge revenue risk: Can they keep winning new people? If I were a potential investor, that's the question I would be asking."
In the years ahead, a number of factors could make it tougher for Demand to keep expanding its audience of unique visitors. Considering its voluminous output, Demand eventually could run low on new, lucrative, and underserved how-to subjects. The rise of social media could negatively affect a business model that is unlikely to generate much traffic from recommendations from "friends." Or search engines could tweak their algorithms, making it harder for Demand to score high in search results across such a broad range of subjects. "Looking at the current model—can you really keep doing that for five years and grow the top line the way you're trying to grow it?" asks Shanahan. "It doesn't seem plausible."
A disproportionate chunk of Demand Media's revenue comes directly from Google—making the search giant a life raft for the company. During the first nine months of 2010, according to Demand Media's Securities and Exchange Commission filings, 28 percent of total revenues came from Google. Demand is aware of the risk of piggybacking so heavily on one source for traffic and revenue, warning potential investors in its SEC filings that the company is "dependent upon certain material agreements with Google for a significant portion of our revenue" and that a "termination of these agreements, or a failure to renew them on favorable terms, would adversely affect our business."
"For Google to make a change that would wipe out Demand Media, I'd be surprised," says John Andrews, an SEO consultant in Seattle. "They are carrying Google's ad inventory." Yet on Jan. 21, Matt Cutts, the engineer overseeing Google's search quality team, posted on the company's official blog, explaining that as pure Web spam decreased, the company had shifted its focus and made two major "algorithmic changes" over the past year targeting mills. "Nonetheless," wrote Cutts, "we hear the feedback from the Web loud and clear: People are asking for even stronger action on content farms and sites that consist primarily of spammy or low-quality content." Does Google consider Demand a content farm? The company did not respond to several e-mails seeking comment.
Gabriel Weinberg, the founder of DuckDuckGo, a niche, privacy-oriented search engine, says the blog post looked to him more like a "PR offensive" than an actual change in policy that would affect Demand. Last year, Weinberg says, after receiving numerous complaints about the quality of Demand articles, he tweaked his algorithm to strip Demand content from his search engine's results. He doubts Google will follow his lead. "They'd be editorializing," Weinberg says. "I think they would be accused of censorship."
Over the past year, Demand executives have taken steps to diversify the company's revenue stream. They have struck a handful of partnerships selling Demand content wholesale to third parties, including the websites of USA Today, the San Francisco Chronicle, and the National Football League. And in March 2010, Demand hired away Yahoo's Joanne Bradford, an experienced digital ad sales executive, to become Demand's chief revenue officer. Bradford now oversees an ad sales staff charged with bringing in brand advertising independent of Google.
In the meantime, the competition among media companies to engineer their content onto the all-important first page of Google search results, no matter how obscure the query, is intensifying. Outsell's Ken Doctor says that businesses such as Examiner.com and Yahoo! Contributor Network are elbowing in on Demand's territory. "Other companies are organizing their own stringer freelance networks using the same kind of technology," says Doctor.
As a result, Demand Media will have to devise new and better ways of extracting revenue from each of its customers in the months to come. For that they are likely to lean heavily on the man who swears he's not the Terminator.
Reese, 42, was born and raised in Texas and graduated from Rice University in Houston in 1991 with a degree in economics. According to his official bio on the Demand Media site, he ran his first business in college, orchestrating "elaborate practical jokes." Reese was unavailable to speak with Bloomberg Businessweek for this story, but interviews with a handful of former colleagues, who say they spent rewarding stretches of their careers working intimately with him, yielded a strong impression. Over the past decade few entrepreneurs have dreamed up more effective and improbable methods of squeezing profits out of obscure corners of the Internet.
Reese lives in a restored Victorian house in the small town of Georgetown, about 25 miles north of Austin. He and his wife, both devout Christians, homeschool their four children and have decorated their home with Biblical paintings. He is an avid deer hunter, has a thing for Byzantine history, and not long ago visited North Korea. According to former colleagues, he has an endless number of business ideas.
"He's an idea generator," says David Hehman, an angel investor in the San Francisco area who helped to back Reese's first major Web venture. "The guy just has a gazillion ideas," says Johnny Anderson, a business executive in Texas who worked alongside Reese at a software company called HotData in the late '90s. "He comes up with off-the-wall ideas and then he executes them," says Zach Hotchkiss, Reese's former head of e-commerce and marketing.
One of Reese's longtime passions, according to his former colleagues, has been making money by identifying and filling in what he sees as the vast gaps in humanity's collective knowledge, as represented by pages on the Web. Over the past decade, Reese has made a small fortune running an eclectic series of Internet ventures somehow related to this idea—and that collectively set the framework for much of Demand Media's current business model.
Around 2000, after roughly a decade working as a marketing executive for various software companies in California, Oregon, and Texas, Reese decided to strike out on his own and wrote a business plan for a company called Brilliant Pebbles, borrowing the name of a Reagan-era missile defense system. At the time a number of media giants—AOL, Yahoo, Microsoft (MSFT), and others—were battling to create the dominant portal. Reese saw an opportunity: What if there were no portal? What if search engines, particularly Google, became so efficient and well-trafficked that users bypassed gatekeepers and went right to the content?
Reese's idea, according to Anderson, was to create a media company catering directly to Google users. The articles could be largely unbranded, because people wouldn't be coming to you for your news judgment, editorial style, or the number of Pulitzers your staff had won. They'd be coming to your page to learn, say, how to treat an allergic reaction to a bee sting because they stumbled on it immediately after they got a bee sting and turned to Google for answers.
In 2000, Hehman decided to help back Brilliant Pebbles. "I liked Byron's raw intelligence," he says. "This guy is probably the smartest guy I know, and it's on a variety of topics: everything from the world's religions to history to technology. He's a really eccentric individual."
Reese opened an office in Austin and renamed the company PageWise. To make the venture work he needed lots of low-cost content, so he launched a feeder site called writeforcash.com. "From needlecrafts to networking, we need your articles," read a version of the site's home page. "We are seeking articles on specific topics that we know to be popular and sought after by people searching on the Internet. Some examples include, 'How to program a universal remote,' 'How to make a sock monkey,' and 'All about Louis Vuitton Handbags.' Get started today!"
Potential candidates flooded in, and Reese gathered a collection of writers willing to fulfill assignments for minimal fees (typically $10 or $15 per article) and no benefits. He used the incoming articles to stock a hodgepodge of sites, such as AllSands.com and WebGuru.com, which promised "to mine all the knowledge of the human race and make it available for free to everyone."
Reese's plan was to monetize the resulting Web traffic using advertising. Then the Internet bubble burst and ad rates plummeted. "We laid everybody off," recalls Hehman. "It was just Byron, his ideas, and the chief technology officer."
Reese reconsidered his business model. Thanks to the churning content mill, he had lots of article pages and plenty of eyeballs, but no significant advertising. So, in a period of aggressive experimentation, he decided to try to monetize his own audience with his own advertisements for his own products.
Every week, according to Peter Handsman, the former CTO, Reese would come up with an idea for something new to peddle. They would draft a business plan, launch a website, and measure consumers' subsequent interest in a product. Efforts to sell coins and watches failed. At one point, Reese tried manufacturing family portraiture using inexpensive subcontractor artists in places such as Russia. The concept wasn't easy to expand. "A lot of people have ideas," says Handsman. "Byron has the discipline to actually measure them. He was willing to come up with a ridiculous number of ideas, but he was also willing to abandon them if they were proven not to work."
Eventually some experiments succeeded. One early hit was a genealogy website where customers could create collectible items decorated with customizable family crests. Reese bought a laser engraver and began shipping out everything from scrolls to plaques to wine glasses decorated with ornate insignias. At around the same time, PageWise happened onto a surprisingly lucrative market for customizable Italian charm bracelets.
In 2002, PageWise launched santamail.org. The site sold parents personalized letters from Santa Claus for their kids. Each letter was postmarked North Pole, Alaska, and sold for $9.95. "The margins were ridiculously high," recalls Hotchkiss, the former head of e-marketing for PageWise. "The volume was insane. It became a cash machine."
Shortly thereafter, Reese's content mill prototype started producing serious revenue, too, thanks to Google's AdSense. Under the program, which launched in 2003, Web publishers could sign up to be Google partners. Google would then go out and sell ads, place them alongside related content on partners' websites, and share a cut of the resulting revenues. "We were selling $6 million worth of jewelry a year by then," says Hehman. "The content money was just gravy."
Reese kept tinkering. In the summer of 2005 he launched happynews.com, which aggregated sunny news stories from around the Web. The site was popular with readers and a punching bag for critics. "Prozac for the eyes," wrote Paul Farhi in The Washington Post.
Reese had ambitions beyond text. He told colleagues that as the price of bandwidth fell, people would begin spending more time viewing Web videos. According to his calculations, Hollywood was spending an average of about $1,000 per one-thirtieth of a second of video. He set out to see if he could produce video for a penny a frame. Not long after jumping into the happy-news business, Reese launched a how-to video site called ExpertVillage.com. In August 2006, the company announced the official debut of its Filmmaker Program. The idea was to recruit some 5,000 freelance filmmakers. Once registered, they could claim assignments from a long list of search-friendly topics. Reese budgeted $2 million to create 75,000 how-to videos over the next 18 months. "We looked at the production of video not as a creative process as it has always been viewed but as a manufacturing process," Reese said in a radio interview with the Christian news organization American Vision. "When it's a manufacturing process, you learn how to optimize certain steps of it and lower costs."
In 2006, Richard Rosenblatt, a polished Southern California media executive, founded Demand Media with an investor named Shawn Colo. Rosenblatt had previously run Intermix, the parent company of Myspace, and orchestrated a sale of the company to News Corp. (NWS) for $580 million. Rosenblatt and Colo raised $355 million and went on an acquisition spree, eventually purchasing eNom, now the world's second-largest Internet domain registry; eHow.com, an online instructional content provider; Pluck, a social networking business; and a hodgepodge of niche content sites, such as Cracked.com and GolfLink.com.
What attracted Rosenblatt to PageWise, according to former CTO Handsman, was not the Italian charm bracelets, the family emblem scrolls, the letters from Santa, or even the happy-news stories. Rosenblatt was interested in Reese's model for manufacturing high-volume, low-cost videos. In June 2007, Demand Media bought PageWise for $15.8 million.
To make cheap content, Demand relies on its seemingly ever-expanding army of 13,000 freelance writers, copy editors, and filmmakers, who have passed the company's vetting process to become members of something called Demand Media Studios. Once accepted into the "studio," writers and producers can log onto a website and choose assignments from a list of possible topics.
The assignments are generated via proprietary black-box technology, which sifts through data on what people are typing into search engines, compares the search queries with keyword advertising rates, and then analyzes how many websites are already serving that particular market. If after mining the data the algorithm determines that a certain headline for a specific topic, no matter how obscure, will be profitable, it pushes the idea into the pipeline. Last year, Demand Media executives told Wired magazine that the mechanized assignments generate 4.9 times more revenue than ideas generated by humans.
In the spring of 2007, a few months before selling PageWise to Rosenblatt, Reese delivered a lecture at Austin's annual South by Southwest festival. His topic: how to start an Internet business for $100.
That afternoon, Reese recommended the crowd get into Web publishing. "The Internet has six billion pages on it right now," Reese said. "Believe it or not, it needs a whole lot more. There's a huge, huge, huge shortage on the Internet of the kind of information that people are actually searching for." The key, he said, was to keep costs low. If possible, don't pay for the intellectual content. Look for material, he urged, on which the copyright has expired. Any book published in the U.S. before 1923 was available.
He said he was in the process of turning vanloads of old books into websites. With a few hours of labor, for example, you could take a turn-of-the-century Creole cookbook and transform it into the definitive site for vintage Creole recipes. Google's AdSense program would then load the thing up with ads for shrimp and cooking pots and spices and direct people looking for shrimp recipes to your website. Every time someone clicked on an ad, you got a cut of the resulting revenue. On a click-by-click basis, those payouts were likely to be minuscule. Don't scan one book. Scan a thousand.
Reese said that years earlier he had created a site called Essortment.com. Over the years the cheaply constructed nonperishable articles stayed more or less the same, while fresh Google ads kept pouring in. "To this day, Google sends me over a hundred thousand dollars a month off this one site," said Reese. "It's like the Energizer Bunny. It goes up to $200,000. It drops off to $70,000. But it never fluctuates out of that range. Year after year after year."
Demand Media executives hope the same is true for their growing library of videos and articles—an ageless future, supported in perpetuity by Google paychecks. They are not the Terminator. They are a pink, fuzzy, hardworking bunny.
As he wound down his talk in Austin, Reese explained the key to success on the Web. He told the crowd of conventioneers to have a bias toward action and not to try to be too smart. "Be as agnostic as possible," he said. "Don't be a true believer in anything."