Studies show that online ratings are one of the most trusted sources of consumer confidence in e-commerce decisions. But recent research suggests that they are systematically biased and easily manipulated.
A few months ago, I stopped in for a quick bite to eat at Dojo, a restaurant in New York City’s Greenwich Village. I had an idea of what I thought of the place. Of course I did — I ate there and experienced it for myself. The food was okay. The service was okay. On average, it was average.
So I went to rate the restaurant on Yelp with a strong idea of the star rating I would give it. I logged in, navigated to the page and clicked the button to write the review. I saw that, immediately to the right of where I would “click to rate,” a Yelp user named Shar H. was waxing poetic about Dojo’s “fresh and amazing, sweet and tart ginger dressing” — right under her bright red five-star rating.
I couldn’t help but be moved. I had thought the place deserved a three, but Shar had a point: As she put it, “the prices here are amazing!”
Her review moved me. And I gave the place a four.
As it turns out, my behavior is not uncommon. In fact, this type of social influence is dramatically biasing online ratings — one of the most trusted sources of consumer confidence in e-commerce decisions.
Online companies are already struggling to deal with “cyber-shilling,” where people or companies are paid to post negative or positive reviews about products or services. But new research suggests that the problem could be much larger than we might expect. We find evidence that some of the people writing reviews on the website of a prominent private label apparel company have never purchased the products they are reviewing.
That firm’s web site generates hundreds of thousands of product reviews. We found that about 5 percent of those reviews were by customers for whom there was no record of actually purchasing the item. Notably, these reviews were significantly more negative than the remaining 95 percent of the reviews, which were posted by customers who were known to have purchased the item. We are also able to replicate the effect using book reviews at Amazon.com.
We use the Internet to research things all the time. Whether it’s a big purchase like a vacation or something smaller like a pair of running shoes, we often begin with a search for the topic online, eventually drilling down to find specific product reviews and details. But what happens when the pattern of websites we visit triggers personalized online ads that follow us around the Internet? If you’re later reading an unrelated article on a news site, will you really pay more attention to an ad for the specific hotel or pair of shoes you were looking at earlier?
The answer is: probably not. It turns out that when people are at the early stages of researching on the Internet — and haven’t likely developed strong product preferences — they respond better to generalized messages intended for a mass audience. Ads that are too specific aren’t going to convert