Online user reviews have become critical to e-commerce success, as 78% of shoppers trust others’ reviews more than the opinion of their own friends. And according to a 2017 Power Reviews study, 82% of shoppers specifically seek out negative reviews because consumers perceive ratings closer to a perfect 5.0 as “too good to be true.” In other words, all reviews are helpful, whether good, bad, or ugly. There’s just one problem, however: not all reviews are real.
Skin-care brand Sunday Riley recently was caught running a fake review program. Internal emails were leaked regarding a step-by-step process for how employees should hype good reviews and dislike negative ones. And, a weight-loss supplement retailer selling on Amazon was fined $12.8 million in connection to false claims and fake reviews.
Reviews in online shopping became ubiquitous thanks to Amazon, which has featured product reviews since 1995. But how much can consumers trust these reviews in light of recent bad behavior?
Gartner analyzed over 23,000 reviews on Amazon from across the Top 100 Best Selling products in Electronics, Health & Household, and Home & Kitchen between April and October of 2019 to test if reviewers had actually purchased the products, discernible through the ‘Verified Purchase’ tag that Amazon posts.
We cannot say conclusively whether a review is fake or not. But it’s clear based on our analysis that some categories seem more ambiguous than others: in Electronics, for example, 66% of all reviews come from non-verified reviewers. Compare this to Health & Household, where only 13% of reviews are unverified.
The issue of fake reviews may subside over time as Amazon and other e-tailers develop more accurate verification tools. In the meantime, for consumers, the best immediate way to screen out potentially fake reviews is to look for comments from verified purchasers, especially in categories like Electronics. Brands, meanwhile, must monitor reviews posted by unverified purchasers and work with Amazon to help clean these up as much as possible.