Life On The Web's Factory Floor

Who do you think turns all those words into an easy click?

Barbara Kempf used to work on an assembly line for words. Her job: take a company name (say, Joe's Diner), think up several words that describe it (restaurant, dinner, burgers), come up with other words about its location (Cincinnati, downtown, Main Street), then create more than 200 combinations of those words and write six- to eight-word sentences associated with each combination. The finished product? Paid search ads, or those "sponsored links" that show up when you search for things like "burgers in Cincinnati" on Google (GOOG ), Yahoo! (YHOO ), or (IACI ).

Over eight-hour shifts, the 51-year-old Kempf "assembled" thousands of these ads every day for $15 an hour for Marchex Inc. (MCHX ), a Seattle-based company that runs online advertising campaigns for local businesses all over the U.S.

With the Internet advertising explosion, advertisers spent $5.1 billion last year to place text ads like Kempf's next to search results online. Several major ad agencies and search engine marketing firms say they plan to double the number of paid-search staff over the next year. Companies such as General Motors Corp. (GM ) now buy ads for upwards of 1 million search terms. The sheer growth of Internet advertising is requiring advertisers to scale and streamline their operations with an almost manufacturer-like mentality.

In the process, a new category of work is emerging: the digital factory job. Behind the seemingly magical offerings of the Internet are thousands of human beings madly inputting data around the clock. The work ranges from the slightly creative, such as Kempf's job of crafting sentences for ads to snag search traffic, to the rote -- typing in descriptions of hamburgers for online menus.

These digital bricklayers are in a sense building the new information pyramid. In Madras, India, "editors" making a fifth of U.S. pay work 24 hours a day, 7 days a week to digitize archived American newspapers from the 1700s to the 1980s. In Boston, New York, and Palo Alto, Calif., Google Books workers manually turn each and every page of millions of library books so they can be scanned and made available to any visitor to the Google Web site.

In Hyderabad, India, typists for startup type the menus of thousands of U.S. restaurants so Web surfers can browse for reservation ideas or takeout. "Internet companies are realizing that you don't need to be a massive company to manage such operations," says Ravi Aron, assistant professor of operations and information management at the University of Pennsylvania's Wharton School. He adds that process work is moving out of traditional places -- insurance claims processing, say -- and onto the Web.

Just like with a display of fresh oranges in a supermarket, far more labor goes into getting the digital product there than most people fathom. Take ProQuest (PQE ) Historical Newspapers archive. Subscribers can type in an author name, headline, or any keyword and access any original image of any article from any historical issue of nine major U.S. newspapers, including The New York Times, The Washington Post, and The Wall Street Journal. The process starts at the company's headquarters in Ann Arbor, Mich., where an operator scans reams of microfilm through a machine that creates a digital copy of each issue. That copy is sent overseas electronically to the shop floor at Ninestars Information Technology Ltd, a digital conversion company in Madras.


There, workers like 27-year-old Virginia Deepa digitally cut up and sort the images into individual articles. Then Deepa tags the headline, first paragraph, caption, and author byline and runs a program to convert each into text. (The rest of the article remains an image file.) Next her work heads to a quality assurance employee who reads over the image file to make sure the words match what is on the page. Then it's ready to be added to the online archive for Web surfers to explore. That process is repeated by Ninestar's 850 employees, 24 hours a day, over three shifts.

Such menial work with data and information is hardly new. But the growing fruits of such operations -- gaining online access to historical papers, for example -- have never been so close to the fingertips of the average person.

Nor has such data entry ever been easier for companies to take advantage of, including small entrepreneurs. The number of third-party, offshore companies that will perform contract work has more than doubled since 2002, according to Wharton's Aron. With Internet connectivity pushing farther into rural areas of China and India, the cost of such work will fall even lower.

That means even more data will be flying online as startups look to fill in the niches, leading to still more digital burger-flipping-type work. When Greg Barton set out to start Manhattan-based, he contracted with a small outfit in Hyderabad. Restaurant menus are too varied for character-recognition software to correctly translate all of the words into text, and mere image scans of each menu wouldn't allow users to search through the dishes by ingredient or type of protein.

So Barton has the workers in Hyderabad sort through thousands of menus, typing up the names and prices of appetizers, entrées, and drinks. Then the offshore staff digitally categorizes each restaurant according to cuisine type, price level, and neighborhood. The operation supports a small ad- sponsored Web page, and 34 employees help him rapidly build out the company's service to include other cities. In the past six months Barton has launched sites for San Francisco, Chicago, and Boston, and he estimates his business can process more than 2,000 menus a month.

Such offshore outfits can also do smaller custom jobs so that even individuals can participate. For 5 cents to 10 cents a business card, typists at nonprofit data processor Digital Divide Data in Laos and Cambodia enter and correct contact information and load it onto customer-relationship management site Inc. (CRM ). Founded by American Jeremy Hockenstein in 2001, DDD hires poor and disabled workers in those countries to perform all kinds of data entry for six hours a day and then get English lessons and computer training for another six. Lately that has included work for executives at International Finance Corp., the World Bank's private-sector arm, as well as for individuals who send in stacks of business cards.

Strong Web access and the low wages of a Cambodian typist -- about $45 a month -- mean there's almost no barrier to entry, so almost anyone can contract for rote work, says Hockenstein. Although DDD started with primarily large jobs -- in 2001 its workers typed the 133-year archive of The Harvard Crimson -- it is adding more small custom work.

These jobs are creating opportunities for some of the world's poorest. One of Digital Divide Data's newest hires is 22-year-old Nouanta Bouthavong. The daughter of farmers from Khammoane Province in Laos, Bouthavong sold vegetables door to door for her family until a sudden illness in her teens left her unable to walk. She attended a vocational school for the disabled and now, in addition to studying English and computers at DDD's school, is in training as an operator.

In the U.S., such line work has high turnover. Google uses a temp agency to keep Google Books staffed. Kempf, the former librarian who worked at Marchex, quit after nine months because the "factory-type atmosphere" got to her. Ad writers worked on a strict weekly quota system that required them to punch out sentence after sentence while their manager, an ex-Marine, barked at them to type faster. "I started having anxiety attacks," says Kempf, who has worked as a freelance index compiler and prefers a slower pace. "I had to get out." A Marchex spokesperson says: "We have a very satisfied base of employees... and allegations about being yelled at have no merit. There's a high degree of accountability, but sometimes that's not for everybody." Either way, there are plenty of people willing to work behind the digital curtain, one keystroke at a time.

Corrections and Clarifications "Life on the Web's factory floor" (Working Life, May 22) incorrectly stated the nationality of Digital Data Divide founder Jeremy Hockenstein. He is Canadian.

By Burt Helm, with Manjeet Kripalani in Bombay

    Before it's here, it's on the Bloomberg Terminal.