Online platforms are cracking down on fake reviews that threaten the credibility of their rating systems.
In 1995, Jeff Bezos, who had just launched an online bookselling platform called Amazon, decided to implement a system that would allow users to distinguish untrustworthy sellers from virtuous ones. The goal was to give people confidence to buy things online from strangers, something that was at the time still very new. Bezos also wanted a system that would allow users to give recommendations to one another, saying publicly why they did or did not like a specific book. The five-star ranking system was born 25 years later, it is widely used throughout the web, by e-commerce platforms (Amazon, but also Walmart, Best Buy) and crow-sourced review websites such as Yelp or TripAdvisor alike.
They have become a cornerstone of these services. 82% of U.S. adults say they at least sometimes read online customer ratings or reviews before purchasing items for the first time. In an era where trust in politics, businesses and news outlets is at an all-time low, online reviews are an exception. 85% of people even trust them as much as a personal recommendation. For businesses, ignoring online reviews isn't an option. Every one-star increase can lead to a 5 to 9% increase in revenue. Good ratings also translate into a better ranking in Google's algorithms.
Fake reviews are proliferating...
But this system may not be as reliable as conventional wisdom holds. When Hillary Clinton published her memoir in September 2017, hundreds of reviews (some positive, some negative) appeared on Amazon, only a few hours after the book was released. The company had to take down a bunch of them, who had obviously been written by people who hadn't read the book and were trying to push the ratings up or down for partisan reasons. This problem doesn't only affect Amazon. In January 2018, the Yelp ratings of the Boca Raton Resort, a Florida hotel, quickly dropped after a customer, who happened to be a famous YouTuber and was angry about his treatment, asked his community to punish them.
People post images of clothing, electronic devices or household items, offering free products and a commission of usually 5 or 10 dollars against a five-star review.
These might sound like very specific and unique cases. But some studies have found out that fake reviews are indeed quite prevalent. In April 2018, two journalists from the Washington Post pointed out that many sellers were flooding Amazon with fake reviews. For some popular product categories, such as Bluetooth headphones and speakers, they even suggested that the vast majority of reviews appeared to violate Amazon's prohibition on paid reviews. This year, researchers at Which?, a UK-based consumer association, doubled down on their conclusions. After analysing the listings of hundreds of popular tech products in 14 online categories including headphones, dashcams, fitness trackers and smartwatches, they found out that many reviews were highly suspicious.
Both studies suggest that many of these potentially fake reviews originate from Facebook, where, in some private groups, brands are asking shoppers to give them good ratings in exchange for money or free products. A quick Facebook research seems to confirm these findings. In only a few minutes, one can join a bunch of groups with names as eloquent as "Amazon - Produits gratuits FR reviews" or "Amazon review deals coupons & increase sales helpline". There, people post images of clothing, electronic devices or household items, offering free products and a commission of usually 5 or 10 dollars against a five-star review.
...but so are tools to detect them
In 2015, as he was looking at the reviews of a health supplement on Amazon, Tommy Noonan quickly realized that there was something wrong. "This product had 580 reviews, and I dedicated a full day to reading each and every one. To my surprise, there were dozens of reviewers who admitted that they never took the product, yet still gave it 5-stars. Most of the reviewers were not verified. Many had only reviewed products from that brand. As I spent hours tallying up all the numbers, I realized that I could write a computer program to collect and calculate all the info, and I could point that program to any product on Amazon and run an analysis." That's how he came up with ReviewsMeta, a free web tool that helps internet users spot fake Amazon reviews.
According to Tommy Noonan, as competition is heating up on e-commerce platforms, new, unknown brands are tempted to take shortcuts in order to quickly attract customers. "Amazon has transformed into a hyper-competitive marketplace where some sellers will do whatever it takes to sell more products. There's so much product sold on Amazon that if you rank well for popular keywords, you can make a ton of money." His tool performs 12 different tests to tell whether a review should be trusted or not. After analyzing a quarter-billion Amazon reviews, his company estimates that between 7 and 9% of them appear to be suspicious.
For example, some brands used to offer a 99% discount in exchange for a five-star review
This figure is higher than Amazon's own estimations, as the company has claimed that less than 1% of its reviews are ingenuine. It doesn't mean that the company isn't dedicated to fighting these fake reviews. After banning incentivized reviews in 2016, Amazon has taken different actions to make its rating system more reliable. This includes targeting unverified reviews, for which Amazon has no evidence that the reviewer ever bought the item. "In the past, there was no limit on those, but since the last two years, for new products, Amazon has reduced the number of unverified reviews that can be published. After a certain amount (usually, about five of them), they will just stop appearing," explains Fahim Naim, a former category manager at Amazon who knows run an e-commerce consultancy business called eShopportunity. They also targeted some specific tricks that sellers used to bypass the incentivized reviews ban. For example, some brands used to offer a 99% discount in exchange for a five-star review. Now, if you get more than a 30% discount, the review will show up as unverified, or won't even show up at all. Verified reviews also have a much bigger impact on ratings than unverified reviews. Since 2015, the company has finally sued more than 1,000 sellers who were allegedly involved in creating fake reviews for their products.
Can the Blockchain eliminate fake reviews?
Yelp has also taken actions to crack down on dubious ratings, using an automated recommendation software to flag them. "Yelp's recommendation software is engineered to highlight the most useful and reliable reviews by continuously evaluating dozens of signals, combating the spread of misinformation. Reviews that may be solicited, fake, biased or are written by users we just don't know enough about, may be moved to the not recommended section. It is a dynamic process in which recommended reviews can change over time as our software learns more information. Unlike many other sites, our stance is quality over quantity when it comes to reviews. As a result, we only recommend about 71% of the reviews that are submitted," a Yelp spokesperson says.
As dishonest sellers keep finding new tricks, these companies also have to keep their detection methods up to date, turning this into a constant arms race. But some third party companies are hoping to completely eliminate the problem, by relying on groundbreaking technology. Monetha, a Lithuanian company, wants to use the Ethereum Blockchain to eliminate any possibility of cheating the system. "We offer a reputation platform for P2P commerce, where both the buyer and seller rate each other after the transaction. We're giving rating that is tied to a payment every single time, which means in order for you to leave a rating, you have to pay. For example, I can't leave a restaurant five stars without ever eating there," explains Justas Pikelis, Monetha's co-founder.
Legislators are also starting to look carefully at this issue. Last February, the American Federal Trade Commission (FTC) settled its first case over fake paid reviews on Amazon. The company Cure Encapsulations was considered guilty of paying customers to post fake reviews to boost the ratings of their weight-loss supplement and will have to pay $50,000 to the FTC. "People rely on reviews when they're shopping online," Andrew Smith, Director of the FTC's Bureau of Consumer Protection, said in a statement. "When a company buys fake reviews to inflate its Amazon ratings, it hurts both shoppers and companies that play by the rules."
Image : Céline Sallette as La Hyène, a professional "fake reviewer" in Vernon Subutex (a series produced by Canal +)