Under fire after a series of well-publicized schemes revealed flaws in the effectiveness of its search results, Google is taking action. The search giant recently overhauled its ranking procedures in what company executives called an effort to favor “high quality” sites over those that are “not very useful,” a move Google says will noticeably impact the ranking of results in nearly 12% of all Google queries in the United States.
The change, which was announced on February 24, went beyond the minor tweaks Google frequently makes that are often invisible to the average user. The company is trying to address fallout from issues that have recently come to light, most notably in a New York Times report that showed how department store chain J.C. Penney became a top search result in many queries by “gaming” the search engine. In a separate incident, retailer Overstock.com was punished by Google for going against policies that prohibit companies from falsely boosting their ranking in the search engine, The Wall Street Journal reported.
Many technology commentators also have recently criticized Google for allowing spam and low-quality content into its search results. Google has responded by saying it would downplay results from sites often known as “content farms” — that is, sites such as Demand Media and Associated Content that produce large volumes of quickly written articles constructed according to how they will rank in search results.
These incidents, and Google’s responses, raise a bevy of questions. How much does search quality matter — and to whom? Does Google have too much market power? Should the company be more transparent about the algorithm that selects its top search results? What is the future of search?
“Google’s search business depends on satisfying two constituencies: users and advertisers,” says Kevin Werbach, a legal studies and business ethics professor at Wharton. “Google is the dominant search engine, but if users find they can’t get to the results they want, they will explore other options. And perhaps more financially significant, advertisers won’t bid up the prices of search keywords. Google’s value proposition is based on being a neutral ‘honest broker’ of good search results.”
Experts at Wharton say the increasing attention being paid to Google’s search results highlights the cat-and-mouse game between the company and those who attempt to game its secret algorithm that determines the order in which search results are displayed. The practice, known as search engine optimization (SEO), uses a range of techniques for creating web pages that increase the likelihood the site will appear among the top results displayed in response to search engine queries. There are plenty of legitimate SEO techniques which adhere to Google’s best practices. There are also “black hat” techniques that aim to take advantage of quirks in the search rank algorithms to artificially gain a higher spot in the query results and garner more traffic for a website.
The ultimate impact of Google’s move to prevent the gaming of search results is yet to be determined, according to panelists at the PaidContent 2011 conference in New York on March 3. According to Luke Beatty, a vice president at Yahoo’s Associated Content, one third of the site’s content ranks higher on Google as a result of the change, but two thirds has garnered a lower position. Jason Rapp, president of search engine Mahalo, told conference attendees that the company laid off 10% of its workforce and changed its strategy after the algorithm adjustments. “We had a body of general-interest content that was affected rather significantly,” Rapp noted. “When we see a hit like that, we react accordingly.”
According to Kartik Hosanagar, a Wharton operations and information management professor, the focus on Google’s search quality is a bit ironic considering how the company originally rose to prominence. “If you look at how Google started out and became so successful, it was because of its PageRank algorithm,” says Hosanagar, referring to the analysis of links across the web the search giant uses to determine a site’s importance, and to decide where pages appear in search results. “PageRank allowed Google to improve search results. Previous search engines at Yahoo and elsewhere were easily littered with spam and low-quality content.”
Now Google may be facing the same issue that its predecessors grappled with more than a decade ago. “Search quality is critical to Google’s reputation,” Hosanagar notes. “Quality is really fundamental to Google’s dominance in the market.”
In January, Google accounted for 68.2% of U.S. searches, according to research firm comScore. Bolstered by its partnership with Yahoo, Microsoft’s Bing search site had a 25.6% share. “People wouldn’t put so much effort into gaming Google if it didn’t matter as much as it does,” says Kendall Whitehouse, director of new media at Wharton. “Because Google is so dominant in search, if you’re not easily found there, for all practical purposes, your website is invisible.”
For that reason, Google’s algorithm overhaul is quite significant, notes Wharton operations and information management professor Eric Clemons. “Anything that weakens the combination of content farms and AdSense [Google's ad serving application for websites] is a good thing. The consumer is better off without content farms. That makes Google better off without them.”
For now, Google is trying to resolve the issue by instituting set rules to weed out problem content from popping up in searches. “When we try to address challenges like this, we try to do it in an algorithmic way. There may be one-off situations where for legal reasons … we will intervene manually,” Neal Mohan, Google’s vice president of product management, said on February 28 at the Morgan Stanley Technology, Media and Telecommunications Conference. Wharton’s Whitehouse underscores the difficulty facing Google in distinguishing between good quality content and bad. “We tend to blame Google” for less than perfect search results, “but knowing what [a search user thinks] is the ‘right’ answer isn’t trivial to implement through software.” Because content quality cannot completely be judged by an algorithm, Google added a feature on March 10 that allows users to block sites they do not like.
Does Power Corrupt?
Google is the primary way to find information on the Internet, and that gives the company a lot of power. Naturally, that power attracts players trying to use Google’s system to their advantage. Hosanagar likens Google in search to Microsoft in security: Both are targeted because they are dominant.
Indeed, a company like Demand Media, which marshals an army of freelancers to produce search-friendly articles, cites dependence on Google as one of the firm’s biggest business risks. Demand Media’s initial public offering coincided with Google’s statements that it would crack down on content farms. On Demand Media’s first earnings conference call on February 22, CEO Richard Rosenblatt emphasized that the company is diversifying its traffic sources.
Clemons has argued in various publications that Google has too much power. He notes that Google has the ability to censor results with editorial judgments that are layered onto its algorithm, and those judgments can make or break businesses that depend on the search engine for traffic. According to Andrea Matwyshyn, a Wharton legal studies and business ethics professor, there is a tension in Google’s technology between “preserving the purity of the algorithm” and using editorial judgments “to address social policy concerns and bad actors.”
For instance, Google significantly lowered J.C. Penney’s ranking in several queries after the New York Times article highlighted the scheme that was pushing the retailer’s website to the top of the results for many search terms. The department store chain denied involvement in the scheme, but the company subsequently ended its relationship with its search engine consulting firm. Matwyshyn thinks Google’s reasoning has been sound so far, but warns: “It’s inevitable that Google will label something erroneously and cause economic harm [to a company it downgrades]. The whole question is what constitutes editorial judgment.”
Matwyshyn adds that Google’s legal obligations when tweaking its ranking algorithm and making editorial adjustments are still a gray area. “To date, this topic hasn’t been explored in law, but it needs to be resolved in the next five to 10 years.”
When Google said that it would downplay low-quality content, important questions were raised, Whitehouse notes. “What is the unassailable definition of ‘low-value content?’” he asks. Google publishes rules that govern its policies, but generally does not comment on specific actions against websites. Hosanagar predicts that the search giant will continue to be targeted because there are real business returns in trying to subvert the system. “Search is the primary means to the Internet.”
Does Google have too much power? Indeed, Clemons considers the company to have a near monopoly. Google executives, however, have continually argued that switching to another search provider is easy. “Google is a for-profit enterprise providing a service. No one has to use Google. It can do what it wants,” says Wharton entrepreneurship professor Karl Ulrich. “The risk is that if its results are viewed as incomplete, rival search engines could gain share. Google has a lot of market power because of its market share. However, the switching costs for users are very, very low. Thus, the moment Google does not provide the best search experience, users will migrate to alternatives.”
But Matwyshyn suggests that it might not be that simple for some users. She notes that while those “who use Google purely as a search engine can find a substitute,” Google customers that also use the company’s e-mail, document, photo and video services may find it more difficult to switch to another search engine.
The Ideal Search
Some experts at Wharton wonder if Google is transparent enough. Google frequently revamps its ranking techniques to improve search results. In general, Google tells the public about changes in its procedures, and publishes a best practices guide for web page creators, but does not divulge much information about the details of how search results are compiled, other than noting that the engine uses more than 200 signals, including its original PageRank algorithm, to order websites.
“I’m not sure how I feel about relying on an algorithm for something as important as search if the algorithm is never described to me,” says Clemons. Matwyshyn would also like to see more transparency but notes that the algorithm is “the company’s prime asset. Google has to keep control over the details.”
If Google became more transparent, it might become easier to trick, Hosanagar points out. “Google seems fully justified in its moves,” he says. “When spammers knew the features of search engines at Yahoo and elsewhere, they gamed the system. The low quality content problem precedes Google’s [rise to] power.”
The challenge for Google, as well as other search engines, is essentially reading the minds of users, and coming up with the most accurate set of results. “Search needs a new trick,” Hosanagar notes. “Just like Google came in with PageRank, some other company could find a different and better way to rank search engine results.” The ideal search, according to Hosanagar, would feature artificial intelligence. For example, IBM’s Watson supercomputer, best known for winning on Jeopardy, may make a good search engine, although the use of such technology is unlikely to be generally available in the immediate future.
In the meantime, search engine providers are tinkering with what Hosanagar calls “artificial artificial intelligence.” The general idea is to integrate social networking feeds from Facebook and Twitter “to infer what human beings are doing,” he says, noting that social feeds might work to enhance results until a better system is devised. “The algorithms don’t understand content and can’t guess what the user wants.”
Some of these social search efforts are already underway. Google in February unveiled an application that integrates Twitter feeds into results. On February 24, Microsoft’s Bing announced that it would integrate Facebook “likes” into search results. If a user logs in to Facebook and then performs a search on Bing, tips and recommendations from friends will appear alongside the results.
Social search might improve results, Matwyshyn says, but Google and other companies have to account for users’ privacy preferences and give them control over which results are shared versus which are kept private. “The ideal would be highly individualized results,” she notes. “My search will look different than your search, and be tailored to my privacy levels.” However, “search is a multidimensional problem,” Whitehouse points out. “Is whether a page is the most recent or the most liked by your friends more important than being the definitive statement on a given topic? It depends on what you’re looking for. There isn’t one right answer.”
In the end, Google’s moves to improve search and police any attempts to game its system will be closely watched. “Because of its dominant position, Google has a heightened responsibility, and its actions will be subject to greater scrutiny,” says Werbach. “The key issue is not whether Google has power, but whether it’s abusing that power. So far, the company has been relatively transparent about what it’s doing, and careful in most cases about what it discourages.”