Resume Reading — One Percenters Control Online Reviews

Close

One Percenters Control Online Reviews

Contrary to appearances, online reviews reflect the opinions of the few.

In a recent New Yorker cartoon, a hotel clerk facing two patrons says, “If there’s anything we can do to make your stay more pleasant, just rant about it all over the Internet.” It is a humorous jab that feeds the popular perception that online consumers wield huge power, and that a few strong words on sites such as Yelp or Amazon can make a business change its ways pronto or go kaput.

And, in fact, online review sites have transformed the way we make consumer decisions. The appeal of online forums on Yelp, Amazon, TripAdvisor, and countless other sites, is to offer a wide range of opinions that, taken together, give consumers the scoop about the real quality and worth of a product or service. Ideally the sites allow consumers to make up their minds by sampling a wide range of public opinion, rather than taking the word of advertisers or professional critics.

But recent social research into how these sites work reveals that they may fall short of providing a representative sample of broad opinion. This is due to how the sites display comments or choose not to, how businesses influence what’s written about them, and the fact that only a small fraction of customers write reviews. Says Duncan Simester, a professor of management science1 at the Massachusetts Institute of Technology (MIT), who has studied the impact of online feedback, “We worry that to the extent that customers are using these reviews, they’re not making good decisions.”

When review sites started appearing a decade ago, they promised opportunities to both shoppers on the street and businesses hoping to court them. In 2008, Yelp’s co-founder, Jeremy Stoppelman, explained, “We thought word-of-mouth is the best way to find trusted local businesses.” Small businesses saw that a forum like Yelp, founded in 2004, allowed them to build a positive reputation and connect with millions of individuals in ways that would have been too expensive and time-consuming in the pre-Internet days.

There are review sites for nearly everything these days—from travel accommodations to shoppers looking for bargains, even to apps like Lulu that rate people.

Indeed, being part of the review forums can help bolster a business’s bottom line. One Harvard Business School study of restaurants in Washington found that a one-star increase in Yelp ratings led to a 5-to-9 percent increase in revenue. A March 2013 report from the Boston Consulting Group involving a survey of nearly 4,800 small businesses found that companies who have a Yelp profile yet do not advertise on the site saw their annual revenue increase by $8,000 on average.

For consumers, too, the review sites offered new opportunities. As more shopping has shifted online from brick-and-mortar stores, consumers have become more reliant on the feedback of customers to make purchasing decisions. A 2013 survey by Nielsen showed that four out of five Yelp users said they read customer feedback before making a purchase. And there are review sites for nearly everything these days—from travel accommodations to bargain shopping to best places to live. Apps like Lulu allow users to rate people (in this case, men and ex-boyfriends) rather than products. By 2013, when Yelp held 53 million reviews and averaged 120 million unique visits per month, it was clear that forums have transformed buying, and even dating; decisions that affect our daily lives.

But while getting instant access to unlimited opinions can be a valuable asset, its drawbacks are becoming exposed. For one thing, the sheer volume of reviews can transform a simple purchase into a research project. Buying something as basic as a water bottle online now involves surveying a dozen brands, winnowing them down based on star rankings or popularity statistics, and reading a handful of reviews about the attributes of each model (leakiness? ease of cleaning?), and weighing the conflicting opinions (“Best water bottle ever!” “Do not under any circumstances buy this piece of garbage!”).

While the cacophony of voices may be overwhelming, the percentage of customers who write reviews is actually quite low. In a 2014 study that analyzed data from a private apparel retailer’s website, MIT’s Simester found that only about 1.5 percent of customers, or 15 out of 1,000, write reviews. “And these customers aren’t representative, they tend to buy more niche items,” he says. Simester also discovered that about 1 in every 15 reviews of an item—about 5 percent—are written by people who haven’t purchased it. “And the problem is the other 985 customers rely on the reviews written by these 15 people.”

Some reviews, in fact, may be entirely false. Others may be planted by businesses to burnish their own reputations or tarnish those of competitors. When companies suspect that online criticism is false, they may be willing to go to the courts to preserve their image. This was the case with a Washington, D.C., contractor, Christopher Dietz, who sued Jane Perez, the author of reviews that alleged his work was of poor quality. Demanding $750,000 in damages, Dietz claimed that the negative publicity cost him $300,000 in lost business as well as personal suffering. The case, which went to trial in January 2014, resulted in neither party being awarded damages. The jury ruled that although Perez had defamed Dietz, he had defamed her right back in responses that claimed she had not paid him for his work.

The legal system has intervened more than once to sort out false reviews. In September 2013, New York State Attorney General Eric Schneiderman fined 19 companies a total of more than $350,000 for paying people to write fake reviews of their businesses on websites like Yelp and Google Local. “Astroturfing,” as the practice is commonly known, can take the form of companies paying people to write glowing reviews of their business and products, or paying people to write damningly negative reviews of their competitors, or both. The name is a play on the idea of artificial “grassroots,” since the reviews in question are a sham.

However, punishing 19 companies barely scratches the surface of fake reviews. Bing Liu, a professor of computer science at the University of Illinois, who has studied review sites, estimates that roughly 30 percent of all reviews online may be fraudulent. Yelp claims that it has instituted a system, involving computer algorithms, that help filter out fraudulent comments. “We have 47 million reviews, and obviously our engineers can’t sit there and watch the transaction go down and say, ‘everything in this review is 100 percent factual,’ ” says Luther Lowe, Yelp director of public policy and government affairs. Yelp, which earns revenue from advertising on its site, has battled allegations that it tampers with reviews to favor companies that advertise. It responds that businesses cannot pay to have a bad review removed, and that a firewall exists between its content and the business side.

Search results, or the order of reviews, have a significant influence, as users naturally read the reviews that appear at the top of a list first. Yelp, for its part, claims that its all-important “search results are based on an algorithm that is designed to provide the best results based on a number of different factors including review text, ratings, and number of reviews.” But it has not made the details of those algorithms public.

In a 2013 study, Liu and his colleagues attempted to reconstruct some of the factors that Yelp may use in sorting its search results. They identified several indicators in reviews that had been filtered out of top results. Those indicators included deceptive, or spam, reviewers; users who submit six or more reviews in a single day; reviewers who rate more than 80 percent of their reviews as four or five stars; reviewers who routinely submit short reviews (less than 135 words); and reviewers whose posts are similar.

University of Illinois computer science professor Bing Liu estimates that roughly 30 percent of all reviews online may be fraudulent.

However, the automatic filter creates another problem. While it helps keep down the number of potentially fraudulent reviews, it also tends to suppress first-time reviewers or place their comments less prominently than those of other users, who have built up stronger online reputations. Yelp’s Lowe says that his own mother has complained to him about not seeing her reviews when she first posted them. These filtered reviews are not deleted, or removed permanently, but instead made available behind an additional link on a business’ Yelp page.

To address the criticisms, feedback sites have adopted a host of strategies. Some sites, for instance, require proof of purchase to counter fake reviews. Amazon has a verified purchase label for people who have bought the items they are reviewing. The travel site Expedia has a policy that reviewers must purchase the service they review through the site itself. The verified purchase program is a way to differentiate within the crowd and endorse one customer’s wisdom over another’s.

Some sites give reviewers incentives to write reviews that will be judged as helpful by others. Yelp provides a status badge to “elite” reviewers whose posts get a predetermined number of likes. And Amazon inducts its reviewers into a Hall of Fame. In addition to signaling to others that these particular premium opinion makers can be trusted, these virtual rewards serve to encourage continued participation on the site, and also to boost reviewers’ egos, which seems to be a necessary driver for people to post reviews. A 2011 paper hypothesized that reviewers may be motivated to achieve status by distinct psychological functions that include goal-setting, personal affirmation, and group identification. In other words, aspiring to attain those objectives may induce reviewers to keep putting out their opinion in well-crafted language.

Yet the badge strategy can backfire. Research finds that status signals may spur people to post false reviews. While labels can be viewed as signals to others about the quality of a reviewer’s feedback, they can also be seen as inducements to write more reviews. “If it turns out that reviewers really do distort their behavior in response to these social labels, the firms could start to diminish the role of these labels,” Simester says. “Maybe elite Yelpers or top 100 Amazon reviewers are not good ideas.”

As the stream of consumer feedback keeps growing, it doesn’t resolve the problem at the heart of all the decisions, which is much older than the Internet: how to quantify something as intangible as a person’s—or a company’s—reputation.

While the web made crowdsourcing possible, it also showed that not everyone in the crowd is the same and that different voices strive to be valued differently. Studies of eBay’s early system of rating buyers and sellers demonstrated that eBay sellers who established positive reputations could charge roughly 8.1 percent higher prices for the same merchandise than new sellers or those with lower ratings. Positive rankings from other users turn out to be worth real money.

As Shakespeare put it, “Who steals my purse steals trash … But he that filches from me my good name … makes me poor indeed.”

A good name translates into money online. And as online forums continue to give more people the opportunity to make their voices heard, it’s likely that companies and customers will find even more ways to game the system.


Josephine Wolff is a PhD candidate in the Engineering Systems Division at MIT. She has written for Slate, Scientific American, and Newsweek. 

20 Comments - Join the Discussion