Can you trust online product ratings? Our investigation suggests even Amazon’s top-ranked, “consistently helpful” reviewers can provide ratings that are at best unhelpful and, at worst, display the hallmarks of fakery.
With millions of products jostling for our attention and cash, 55% of web shoppers turn to online reviews to help them make their choice, according to a 2017 KPMG report.
Retailers are acutely aware of the power of online reviews, with some creating glowing reviews of their own products or negative ones of their competitors’ (practices known as “astroturfing”).
In 2015, Craig Douglas, a partner in Christchurch building company Clearwater Homes, received a warning from the Commerce Commission for several testimonials on the Clearwater Homes website from supposedly happy customers.
The reviews were fake, written by Douglas and one of his friends, and had to be removed.
Concerningly, some websites may be in on the astroturfing game. This year, the commission filed charges against holiday accommodation site, Bachcare, alleging it misled consumers by manipulating online reviews.
The commission claims Bachcare removed negative comments from some reviews and refused to publish any reviews with a star rating lower than 3.5 stars.
Businesses in the market for fake feedback can pick it up pretty cheaply – we found freelancers on Fiverr.com offering to write basic reviews for NZ$8 each.
Major websites such as Amazon, Google, and TripAdvisor maintain they have robust procedures in place for detecting and removing fake reviews – from moderation checks to processes where users can flag suspicious reviews.
However, consumer investigations suggest suspicious critiques are still slipping through the cracks.
What we did
Amazon gives a little badge to top reviewers who provide “consistently helpful, high-quality reviews”. This badge appears next to their username above each review.
We scrutinized the 10 top-ranked reviewers holding these badges on Amazon, calculating the number of reviews over a 12-month time frame, the highest number of reviews posted in a single day, the average star rating, the number of times they gave a one-star review, repeated purchases, and overused stock phrases.
During the year, three reviewers repeated stock phrases in five or more different reviews, which we found suspicious. Survey results
Of the more than 1800 reviews left by the top-ranked reviewers, the overwhelming majority were positive.
All top-ranked reviewers gave an average rating of between four and five stars – indicating they’re easy to please or may have an incentive to rate a product highly.
Only three gave one or more one-star reviews during the year, though these accounted for just 0.05% of all their ratings.
One profile gave nothing but five-star reviews, though this was noted in the reviewer’s profile: “I’m a positive-only reviewer … This is my five-star account.”
Across the year, the most prolific reviewer (of the 10) posted 535 book reviews – which equates to a book-and-a-half every day.
The user often published reviews in clusters of three or more. On October 21, 2018, they posted 10 book reviews. Six reviews were published on December 14, 2018, and February 10, 2019.
Many of Amazon’s top 10 had “high-volume” posting days. Another user managed to crank out 14 five-star book reviews one day last October.
Our investigation also found profiles repeatedly relying on stock phrases to bulk out their reviews, which we found suspicious.
We found these phrases popping up repeatedly:
“This was my first book by the author, and it won’t be my last.”
“I can’t wait to see what [insert author’s name] brings us next.”
“Definitely a [book/series] I would recommend, and I look forward to reading more …”.
We reported the cases of repeated phrases to Amazon and asked it to investigate. The company said it “took appropriate action” but didn’t clarify what this was. The reviews we flagged remained on the site.
In a statement, Amazon said “automated technology and teams of trained human investigators” are in place to catch abusive and fake reviews before they appear on the website.
“Last year, we prevented more than 13 million attempts to leave an inauthentic review, and we took action against more than five million bad actor accounts attempting to manipulate reviews,” a spokesperson said.
By awarding a badge to certain users, Amazon lends credibility to these top-ranked reviewers’ seal of approval – despite the fact these reviewers hand out praise suspiciously often.
We’d like to see Amazon apply more scrutiny to anyone with these badges, including regular scans of their reviews for stock phrases.
We’d also like to see websites such as Amazon, which invite reviews, include an aggregate score with reviewers’ profile info. So if they hand out four or five stars like they’re going out of fashion, other shoppers will be able to spot this right away.
In assessing reviews, we kept an eye out for:
Skew. Real-life reviewers are typically measured – you like a few products, find another shoddy, rank most as average – so when you add up all your reviews, the score falls somewhere in the middle.
Not so with paid reviewers, whose overall score is likely to be suspiciously high or low because the company is paying for positive or negative feedback.
Even if they’re not fake, a reviewer who’s easy or impossible to please is little help in making a purchasing decision.
Frequency. Your average shopper isn’t going to spend vast chunks of valuable time-off posting reviews on Amazon.
High review counts or clusters of reviews posted on the same day are suspicious. At the same time, it’s questionable for someone to post a single five-star review years ago and nothing since.
Repetition. An ordinary punter is unlikely to buy similar gadgets over and over, particularly if they’ve given each product five stars.
Similarly, if reviews are written organically, it’s unlikely the reviewer would end up using the same phrasing.
There’s little incentive for independent reviewers, who are free to say as much or as little as they’d like, to “pad out” feedback. However, a paid reviewer might use stock descriptions – to meet a set word count and save time.
The UK organization, Which?, investigated the reviews on travel website TripAdvisor.
Analyzing the nearly 250,000 reviews on 100 top-ranked hotels, 15 hotels had the “blatant hallmarks of fake reviews”, according to Which? Consumer rights expert Adam French.
When TripAdvisor received a copy of the results, it said 14 of the 15 hotels had been busted with fake positive reviews in the past year.
Which? noted a suspicious pattern on many hotels: one negative rating and a quick succession of very positive reviews.
“A flood of five-star reviews after some bad reviews could indicate that a concerted ‘push’ for positive reviews has been coordinated,” Mr. French said.