Online reviews: what you read isn’t always what you get
While big companies try to prevent them, fake user reviews are still a problem
Amazon “uses a number of mechanisms to detect and remove the small fraction of reviews that violate our guidelines”. Photograph: Rick Wilking/Reuters
The Irish love to shop online: an estimated €7.5 billion will be spent on internet purchases by the end of 2016, according to Virgin Media’s Digital Insights Report, and this will rise to €14.1 billion by 2021.
We also like to research before we buy: the Deloitte Ireland Consumer Review Online says that one-third of Irish consumers use online reviews for guidance before making a major purchase such as that of a holiday or television. But how do we know that those reviews are real? As they say, on the internet, nobody knows you’re a dog.
There are businesses out there looking to deceive the unsuspecting online consumer (or bring down the competition) with a barrage of fictitious feedback from phantom customers. Tactics range from writing fake glowing reviews of their own products or getting friends and family to love-bomb the review section to hiring people to write them, even using botnets to generate automated reviews.
The existence of fake reviews is problematic for ecommerce platforms such as eBay, Amazon or Etsy whose services include providing authentic customer feedback or those whose business model hinges on hosting a community of reviewers: the likes of Yelp and TripAdvisor.
The first line of defence for these organisations is the employment of staff and technology to weed out these kinds of reviews. From the point of view of honest traders who feel reviews are blatantly biased against them, there are, in most cases, opportunities to dispute a review while consumer protection exists in the form of European guidelines and legislation.
It is possible to build your business around fake review detection.
“Five years ago fake reviews practically didn’t exist,” says Ming Ooi, chief strategy officer of fake review-busting website Fakespot. com. “We didn’t realise how bad the situation was until our website just took off without any promotion or outreach on our part. That’s when we knew [fake reviews] were a big problem.”
Knock-off goodsOoi describes how easy it can be for a seller to “game” the system with a two-pronged strategy that includes fake reviews.
His observances of some ecommerce platforms were that some companies began flooding the market with low-end or knock-off goods using cut-and-paste product descriptions and images taken directly from an original product. This was done in conjunction with supplying their own “reviews” to pump up their ratings.
From the consumer perspective they are looking at an enticingly cheap pair of Manolo Blahniks with scores of enthusiastic user reviews.
Those with no time to write masses of fake reviews can easily obtain them online within hours. They can be bought from freelance “for-hire” sites such as Fiverr. A quick search yields oblique (and sometimes direct) offers to write positive online reviews for about €5.
I sent out a few queries of my own, asking if it was possible to commission a “positive product review” for an unseen and untested product. It is very possible although many now state they will not write and post reviews to well-known platforms such as Amazon and Yelp because it is getting increasingly difficult to game these systems.
Amazon has started cracking down on sellers trying to deceive its customers. Earlier this year the ecommerce platform sued three sellers who were found to have created fake accounts in order to post false customer reviews beneath their products.
In fact, since early 2015, Amazon has sued more than 1,000 people who posted fake reviews in exchange for payment and recently brought in a policy that an Amazon customer, in order to post a review, must have spent at least $50 since joining.
Legal action“Our goal is to make reviews as useful as possible for customers,” said a spokesperson for Amazon.
“We use a number of mechanisms to detect and remove the small fraction of reviews that violate our guidelines, and we terminate accounts. We’ve filed lawsuits against a number of individuals and businesses who were abusing the system.”
And while Amazon, as a marketplace hosting sellers, has taken legal action, it is also possible under European legislation for consumers to pursue deceptive vendors.
First of all, there are guidelines that clearly state fake reviews are unethical. The International Consumer Protection and Enforcement Network (ICPEN) unambiguously states: “Never pretend to be a consumer and write fake reviews.”
More importantly, there are legal implications when traders write, commission or post fictitious user reviews. Martina Nee, press and communications officer at the European Consumer Centre (ECC) Ireland, points to European legislation, the unfair commercial practices directive, that explicitly prohibits traders from “falsely representing oneself as consumer”.
“In other words, it is actually against the law for sellers to mislead consumers,” says Nee.
Having said that, it can be difficult for consumers to tell whether a review is real or fake. ECC Ireland’s advice is “always to do as much research as possible and not to just rely on reviews alone”.
“Checks include visiting the seller’s website to ensure they have a ‘contact us’ section because they are obligated under EU law to include contact details. If they don’t, this should ring alarm bells,” Nee says.
“Unfortunately, if a consumer is the victim of fraud, it can be very difficult for them to get any redress.”
This is why it is important that ecommerce platforms and user review sites have measures in place to catch a fake review before it ever gets published.
“Unfortunately, there are people out there who are willing to cross legal and ethical lines in an effort to mislead consumers,” says Marco Bilello, PR director for Yelp. “To counter this, Yelp goes to great lengths to protect consumers from fake reviews, through a mix of technology and human touch.”
Suspicious contentYelp has automated software that filters these reviews, going through an average of 108 million reviews from around the world on a daily basis.
Bilello says the website also relies on the community of users who are encouraged to flag any suspicious content they see for the user support team to evaluate. This process also includes detecting and not recommending reviews in the form of unhelpful rants and raves.
As with Amazon, Yelp is also combatting these paid-for reviews that can be bought for pennies and used to bulk up a bad product or service. “We take a very aggressive stance against ‘shill’ or biased review content,” says Bilello.
Yelp won’t tell us what this secret sauce is (“we don’t disclose precise details about the recommendation software to prevent people from gaming the system”) but “it involves many individual signals as well as pattern recognition and machine learning over time in order to highlight the most useful and reliable content for consumers”.
Businesses caught infringing are slapped with a “consumer alert” label for 90 days. Some of these are painfully obvious, like a whole bunch of reviews coming from the same IP address.
Genuine customersTripAdvisor’s business model is providing reviews for travellers and as with Yelp it has contended with fake user reviews but has systems in place to prevent this as best as possible.
“We dedicate an awful lot of time and resources to making sure that the content on our site meets our guidelines and that it is genuinely useful to travellers,” says TripAdvisor spokesperson Jams Kay.
“This encompasses everything from the moderation aspect – making sure the language is family-friendly – to making sure the content of reviews is relevant to travellers, ie it is about first-hand experience, right through to making sure that reviews are not what we would consider biased; that they are from genuine customers and travellers, not someone linked to the establishment itself or to a competitor establishment nearby.”
Data analytics plays a large part in this. As TripAdvisor has been gathering hundreds of millions of reviews for more than 16 years, it now has a pretty good idea of what a normal review pattern looks like around any given property.
Meanwhile FakeSpot, which reports having analysed 984,345 Amazon products and 3,984 Yelp listings to date, gives users a chance to do some investigating if they think a review looks a bit suspect.
The online tool searches for every publicly available review of a specific product, looking for various patterns; it is a giveaway if a large number of reviews has very similar phrasing or happens within the same time period.
“Once the algorithm does this, it goes one level deeper and analyses every person who has left a review, checking out their review history. What you find is that fake reviewers use the same language almost verbatim or even a straightforward cut-and-paste job,” says Ooi.
Despite the war on fake reviews, they are a bit like Pokémon: it is almost impossible to catch them all. It remains prudent to abide by the long-held principle of caveat emptor or buyer beware, even in this digital age of online shopping and those damn falsa scripto.