Online consumers, when looking through product reviews, are understandably concerned about the authenticity of reviews. But it is getting harder for the less reputable merchants to get away with acquiring fake reviews. One reason is that consumers have resources like ReviewMeta at their disposal. ReviewMeta does a very good job at applying statistical analysis to detect unnatural reviews. We tested ReviewMeta on a few products on Amazon. By concentrating on some more unusual products, after a few attempts, we found a “face-slimming breathable chin strap” which failed. It failed two tests – ‘Reviewer Ease’ and ‘Reviewer Participation’. The Reviewer Ease failure basically means that the product received a relatively high number of high ratings from reviewers who consistently and readily award products with a high rating, compared to the number of reviews from reviewers who generally give a broader range of ratings. The Reviewer Participation failure is essentially based on whether there is an imbalance in reviews from people who either only review a few products or people who only review a relatively large number of products. For example, if all the reviews were from people who don’t normally write reviews (and there is a relatively high number of reviews), then this could be an indication that a brand may have incentivised customers or fans to review the product. ReviewMeta perform 12 different tests, altogether. They are all tests that can only be performed programatically and, mostly, require collecting and analysing data beyond that directly available on the product page.
ReviewMeta currently only analyses reviews on Amazon and Bodybuilding.com (though expects to support more stores soon).
Another site performing analsysis to detect unnatural patterns is fakespot.com (which, unlike ReviewMeta, can check Tripadvisor).
Of course, it is not uncommon for the business entry on a review site itself to be fake, complete with fake business reviews. Performing a reverse image search on a listing’s image or a reviewer’s profile image is a technique at your disposal for finding out where else an image has been used. It might even reveal that a reviewer has grabbed the image of some one else’s public domain photo as their own. There mere fact that a reviewer’s photo is not of a person at all also leads to suspicions on the authenticity of the review.
Services providing fake reviews run the risk of being sued by Amazon who filed their first lawsuit of this kind in April 2015. In October 2015 it filed another lawsuit targeting over a thousand registered sellers on the website Fiverr. In 2016, Amazon adjusted its Community Guidelines to prohibit reviews solicited by marketers in exchange for discounts, freebies or other rewards (incentivised reviews), which up to then had been permitted by the service.
Amazon and other organisations, including Google, have also been developing algorithms to target fake reviews. In addition, Amazon has a feature whereby a “verified purchase” flag is added to reviews by customers who have purchased the product from Amazon. Note that the lack of a flag does not mean a review is fake, as the customer may have purchased the item elsewhere, but customers can still use this information to help them make purchasing decisions (e.g. what if a product has many reviews and none of them have the verified purchase flag?). Moreover, Amazon will add extra weight to overall review scores where reviews are from verified purchases. Other factors taken into account when calculating the review score include age of review and helpfulness votes by customers.
The review platform Yelp also has at least two systems in place to deter fake reviews; one being the means for customers to flag suspicious reviews so they can investigate further and the other is their posting of customer alerts when questionable reviews are uncovered.
In April 2017, the Airbnb accommodation letting and booking service documented the multiple ways they are protecting hosts and guests from fraudsters, including the means to flag suspicious listings on their mobile devices.
The above is just a selection of ways which make it hard for fraudsters and dishonest merchants to leave false or misleading reviews, and there is no doubt that even more sophisticated algorithms will be developed combined with increased manpower to tackle these issues. The likes of Amazon, Booking.com. Airbnb and Tripadvisor place enormous importance on the trustworthiness of reviews.