Customer reviews can make or break a product, influencing what we buy and what companies produce. So with this much riding on them, the prospect of deceptive reviews—reviews that are not what they pretend to be—is quite unsettling.

One fear is that employees from rival firms will pose as legitimate customers and post negative reviews about a company’s products or services. Indeed, on some review sites like TripAdvisor, such concerns may have merit. “Some research has shown that hotels from independent chains may write negative reviews about rival hotels in the area,” explains Eric Anderson, a professor of marketing at the Kellogg School of Management.

But Anderson and his colleague, Duncan Simester of the Massachusetts Institute of Technology, have recently discovered a different type of deceptive review—one written not by a company’s rivals, but by its very best customers.

Evidence of Deception
Anderson and Simester examined customer product reviews posted to the website of a prominent apparel company. The researchers first attempted to connect reviews, which could only be written by registered users, to individual purchases—a task made possible because the retailer was the exclusive seller of its products, and held detailed transaction records. These records linked purchases made through multiple retail channels to specific customers via credit card numbers and other identifying information, leaving researchers fairly confident that if a registered user had made a purchase, they would know it.

But then the researchers noticed something unexpected: about 5% of the over 300,000 product reviews they sifted through could not be linked to actual purchases. “So we spent a lot of time trying to figure out who these customers are and what these products were,” explains Anderson. “And after an incredible amount of sleuthing”—ruling out the possibility the individuals had received the items as gifts, among other explanations—“we determined that it was very likely that these customers had never bought the product.” In other words, the customers appeared to be lying.

“It’s actually thousands of customers writing these negative reviews. And so the question then is, well who are these guys?”

Linguistic analyses support their conclusion. Prior research has identified a number of features that can be associated with deceptive language—one of the more prominent of which is the use of elaborate narratives. And indeed, reviews that could not be connected to actual purchases had a word count about 40% higher than other reviews. Another linguistic hallmark of deception: a surfeit of exclamation points—“not just one but multiple exclamation points,” says Anderson. “It turns out that’s also very prevalent in these 5% of the reviews.”

So how do these deceptive reviews differ from other reviews on the site? The most striking quality is their negativity. The researchers found that reviewers for whom a purchase could not be verified were twice as likely to assign the lowest rating to a product than reviewers with verified purchases. And make no mistake—this 5% of disproportionately negative reviews was enough to impact sales. “We show in this case that the firm loses about 1–2% of revenue,” explains Anderson. “That may not sound like a lot of money, but when you’re a really big company, losing 1–2% of revenue because of [deceptive] reviews is a huge figure.”

Meet the Deceivers
In considering who these reviewers are, and why they may have left deceptive reviews, it is important to remember that writing reviews is inherently an “anomalous activity,” explains Anderson. “Not many people write reviews. So we’re already down into fewer than 5% of consumers actually taking the time to write reviews. And now we’re finding 5% of that 5% are doing these somewhat deceptive reviews.”

Even so, he explains, it is not the case that just a few bad apples are responsible for the deception. “It’s actually thousands of customers writing these negative reviews. And so the question then is, well who are these guys? And the big surprise to us was: they’re really good customers! In fact they are way better than the average customer.” These deceptive reviewers are such good customers, in fact, that they have on average each purchased over 100 items from the retailer.

The elephant in the room here is, of course, why? Why would a loyal customer write a review of a product she never even purchased? One explanation is that these customers are upset with the company and wielding negative reviews as punishment. Yet it is hard to see how they could be too upset, as these customers continued to purchase products from the retailer even after leaving their deceptive reviews. Nor is there good evidence that these customers were simply racking up reviews to establish social standing and credibility. These reviewers tended to have purchased plenty of products for which they did not write reviews.

Instead, the researchers argue, these reviewers are acting as self-appointed brand managers. “They’re such good customers that they now feel like they should be guiding the company,” says Anderson. “And when the company makes mistakes in the eyes of the consumer, they want to correct them.” Again, linguistic analyses offers some support for this view: the pool of deceptive reviews contained tellingly high numbers of phrases such as “carry more” or “go back to”—phrases suggestive of requests aimed at the firm, rather than advice for fellow customers.

Discouraging Deceptive Reviews
What then can a company do to discourage customers from leaving deceptive product reviews? Resist the urge to take legal action against customers who engage in this behavior, says Anderson—even if you can prove their reviews to be groundless and harmful. After all, these may be some of your best customers. Instead, think redirection. Seek these customers out and offer them other avenues for expressing displeasure, doling out advice, and otherwise feeling heard.

The broader issue, though, is one of representation. Because so few customers leave reviews in the first place, retailers are left with feedback from a population that they know is not representative of their customer base, even without deception thrown into the mix. “That’s a problem,” says Anderson, “and the companies know this.” The pair of researchers has since teamed up with experts in boosting engagement and turnout for political campaigns. The hope is that the two teams of experts can learn from one another: perhaps tricks for increasing voter participation can be used to increase reviewer participation, and vice versa.

Editor’s Note. Why do you think loyal customers might leave negative reviews about products they have never purchased? And what can be done about it? Tell us your thoughts below. Eric Anderson has kindly offered to address your comments in a subsequent post for the Insight Blog.