How consumer reviews deceive you

MANILA, Philippines — Much retailing now involves consumer reviews.

Yet most consumer-review systems are fatally misleading.

No purchase verification. On TripAdvisor, *anyone with an email can review any establishment* without verification. TripAdvisor claims it can detect bulk fabrication, but it has no safeguards against a few people opening 20 email accounts with which to plant 20 false reviews, enough to destroy or boost most restaurants.

“Readers’ Choice” awards are no better. A well-known magazine once named my hotel one of “Asia’s 25 Best Resorts,” a distinction easily worth hundreds of thousands. A jealous competitor only needed to invest $100 for a few subscriptions through which to submit false reviews the following year. At the rarefied Best 25 level, very few negative reviews were needed to nudge us off the list.

Reviewer qualification. For a useful review about a restaurant in London, would you believe someone whose only previous review was of a Jollibee, or an experienced diner with reviews in many cities and countries? Yet every review system practices this foolish form of democracy, giving equal votes to everyone without considering past experience or diligence in reviewing.

The Michelin irrelevance pitfall. Michelin operates an “expert rating” system. Anonymous reviewers judge restaurant food (not service, not ambience). Except they don’t. I have patronized literally hundreds of Michelin-starred restaurants around the world, from French Laundry to Maaemo to Odette (see https://plantationbay.com/dining/restaurants), and think it’s at best Hit-or-Miss. Michelin reviewers seem obsessed with “prettiness” and “strangeness,” but ignore the qualities that diners really care about – taste, memorability, repeatability. “Did I enjoy eating this? One week after, can I still remember it? Do I want to eat it again in this lifetime?” Michelin doesn’t seem to care, so long as it looks “creative.”

Michelin has wandered into rating hotels now. I recently stayed at a Two-Michelin-Key hotel (a very rare accolade) in the US and it had one mediocre restaurant and a view, no other facilities to speak of, and a badly-furnished room. I wrote to a Three-Michelin-Key property with a question about their rooms and no one even bothered to answer me. Does Michelin fact-check their reviewers? Do the reviewers simply crib their ratings from TripAdvisor and Booking.com?

“Gaming” Amazon. Selling a $100 item? Choose Amazon’s “Individual Plan” – commission $0.99 per item. Now get 10 confederates to “buy” one. (With “merchant/vendor fulfillment,” no need for actual shipment.) 10 five-star verified reviews for just $10.

“Gaming” Booking.com. A few hundred reviews are enough to give a hotel lots of credibility. Suppose rooms cost $100. A hotel lines up 300 ghost travelers to book one-night stays on Booking.com/Expedia. They don’t need to actually go there; the hotel just records that they did. At 17 percent commission, the shady hotel pays $5,100 for 300 glowing reviews, which can easily draw several hundred thousand dollars’ worth of real bookings, AND influence “expert reviewer” systems by a psycho-persuasion technique called Anchoring. (Look it up). Talk about “Return on Investment.” One hotel I observed was open barely a few months but quickly amassed over a thousand mostly five-star reviews. (A quality hotel is lucky to get 50 legitimate reviews a month).

Brand-name bias. Most people won’t give a chain hotel a super-low rating, because they may need to use that chain again. People also don’t badmouth celebrity-backed establishments, fearing to contradict popular opinion. Such businesses are thus largely shielded from the lowest grade ranges. No such constraints protect stand-alone properties or small brands. Brand-name bias is real and it dis-favors smaller, less-famous businesses.

“Average” is nonsense. An “average” encompasses widely-varying views. Take Coca-Cola. Many people love it, but others might give it 1/10 on health or political grounds. Speculatively, Coke’s worldwide satisfaction average might be just 7/10, suggesting Coke is a mediocre brand. But obviously it isn’t, and 7/10 is a poor predictor of how any specific person will feel about having a Coke.

The “Karen” bias. Calculating an “average” gives disproportionate influence to nasty narcissists. Imagine a hotel with 10 reviews. Eight of those are 9/10. Two are 1/10 – one guest was fined for smoking, and another tried but failed to get a discount over a fabricated complaint. Both “get even” by posting 1/10 reviews. This hotel has 80 percent highly satisfied customers; if YOU are not a jerk, YOU are likely to be highly satisfied, too. But the “average” is 7.4, which is low enough that many travelers will remove it from consideration. The two nasty Karens were able to out-vote the eight happy customers.

Correcting this would be simple: instead of the “average score,” report the “median” (midway point between the upper and bottom halves), and the “mode” (most frequent individual score). In the above example, the hotel’s median and mode would both be 9.0.

In summary, most ratings systems are unwittingly deceptive, and are easy to exploit for fraud and extortion.

What to do?

Whether for an Amazon or a Booking.com, require that an account be open for at least (say) six months, with at least (say) six different actual orders and reviews, before any reviews by that account become “live”; and further require a threshold continuing expenditure and review activity, otherwise the past reviews are deleted. We can quibble over the numbers, but the intent is to make it harder to fake verified reviews.

Ignore the top 10 percent and bottom 10 percent of scores as a quick way to compensate for Brand-name Bias. Or give users the option to select this as a filter.

Proactively investigate anomalies, and demand criminal penalties for ratings tampering and false reviews in whatever form. Lobby with legislatures accordingly.

For its own long-term reputation, Michelin needs to overhaul its criteria, and review the honesty and competence of its reviewers.

Emasculate internet blackmailers brandishing fabricated complaints and unreasonable demands, simply by reporting median and mode instead of average.

Manuel González is the owner of Plantation Bay Resort & Spa in Cebu.

For full credentials, please see www.plantationbay.com/cred.