It may be hard to remember now, but Amazon started out as a book-selling website. That’s what makes it so ironic that, more than 25 years later, it still hasn’t figured out how to fairly curate reviews of, well, books.

Maybe it should stop trying.

Full disclosure: Like any other author, I watch my reviews. For a while, a was very sensitive to them, basking in positive reviews and having a conniption fit over the bad ones. I’ve gotten a lot more good reviews than negatives, but I’ve stopped paying quite as much attention to them because the entire process is so poorly defined and managed.

My ego’s only marginally involved these days; I’m confident in my writing ability, and anyone who wants to disagree is free to do so. I’ll accept constructive criticism and put the haters where they belong.

The problem is that negative reviews can hurt sales, and this is especially true on lightly reviewed books. A whopping 82% of customers seek out negative reviews, and negative reviews can stop 4 in 10 people from using a business. While binge-watching the comedy show Schitt’s Creek on Netflix, I came across an episode titled “Motel Review,” in which motel owner Johnny Rose (Eugene Levy) complains that, after four positive reviews, one bad notice has tanked the business’ overall rating.

Most people who run small businesses have felt his pain.

Amazon’s flawed model

With Amazon, it’s even worse.

For some inexplicable reason, Amazon doesn’t weigh all reviews the same, using a “machine learned model” (whatever that is) rather than a raw data average. The company explains its policy as follows: “The machine learned model takes into account factors including: the age of a review, helpfulness votes by customers and whether the reviews are from verified purchases.”

Three factors are listed. Let’s take them one by one.

Verified purchases count for more. This is, ostensibly, to curb bogus reviews — which it may do to some extent. But make no mistake about it, the main reason behind this policy is not to improve the quality of reviews. It’s to encourage authors to sell their books via Amazon, rather than through Barnes & Noble, at book shows, etc. (If you doubt this, note that Amazon offers extra money through KDP Select page views... but only to authors who sell their books exclusively on Amazon.)

Helpfulness votes by customers are also considered, but this isn’t particularly fair, because the person who’s scared away from something by a negative review is more likely to say it was “helpful” than someone intrigued by a positive one. There are two reasons for this:

First, people are more likely to look at negative reviews in the first place, because they want a reason not to spend their money. This is particularly true in lean economic times. Selling a book, or any product is a four-step process: You have to reach your audience, intrigue that audience, persuade that audience to buy your product, and then deliver a quality experience. Not buying something involves just one step: deciding it’s not worth the money.

Second, people are much more likely to mark a negative review as “helpful” because they’re already on the site. It’s easy to see a scathing review, breathe a sigh of relief that you didn’t buy the product, and click that “helpful” button. But in order to find a positive review helpful, you’ll have to buy the product, leave the site, wait for the product to arrive, use the product, and then return to the site, find the review, and click “helpful.” That sounds exhausting just writing it.

Negative reviews affect things far more than positive ones do, and even more than you might expect. The National Strategic Group suggests you need 10-12 good reviews to offset a single bad one. (Because 1 in 5 bad reviews are posted out of revenge, and others write critical reviews to feel knowledgeable or “superior,” this makes getting a good average more of an uphill slog.)

The third criterion Amazon uses is perhaps the most ridiculous of all: how recent the review is. This makes sense for a service, the quality of which may have declined over time —a restaurant that has changed ownership, for example, or a motel that’s gone downhill. But it’s an absurd standard to apply to static products such as books. They may have been published in 1950, but they’re still the same in 2020.

Books aren’t T-shirts

But there’s another crucial difference between books and other products sold on Amazon: they’re creative products, which means they’re largely subject to viewer tastes.

Say you buy a cellphone that starts malfunctioning right out of the box, or an “assembly required” item with missing pieces, or a T-shirt that doesn’t fit because the sizing’s wrong. Those are all technical issues. You can’t really argue with them. But books are different. When a reader doesn’t like a book, it’s usually all about taste. The eye of the beholder. Yes, some books are better than others, but their shortcomings are far harder to quantify unless they arrive with a cracked binding or riddled with grammatical errors.

To make matters worse, Amazon now allows people to simply rate books (from 1 to 5 stars) without even leaving a review, and these ratings sometimes get weighted higher than actual reviews. Stars alone don’t tell you what’s good or bad about a product, so they can’t possibly be more “helpful” than a detailed critique. Plus, they’re left anonymously, which reduces accountability.

How to fix it

The result of all this is an unfair and misleading system that’s skewed against authors and publishers. Because people are more likely to leave and read negative reviews than positive ones, fairness would dictate that — if there’s any “weighting” to be done — it should actually favor good reviews.

The “helpful” button should be eliminated. People who read reviews can decide for themselves whether they’re helpful or not. That would also eliminate the practice of Amazon highlighting the most helpful positive and negative reviews, which often highlight the most scathing criticisms just because they stopped people from spending their money. While that may be “helpful” to their bank accounts, it says nothing about the quality of the book.

Amazon should also separate out quality and creative reviews. For example, a company shouldn’t get a bad quality review if an item arrives broken because it was damaged in shipping, yet Amazon lumps these reviews in with all the others.

And hey, Amazon, stop treating books like T-shirts and smartphones. You’re not just lumping apples in with oranges; you’re putting ketchup on ice cream.

When it comes to books, there should at the very least be two categories: One for creative content and the other for technical issues such as editing and the book’s condition. A customer could be asked to check either of these boxes upon beginning a review, and could be given the option of writing separate reviews on both topics.

Amazon should use recency ratings only for services that may decline over time, not for static products. And it should revisit the idea of allowing users to leave anonymous ratings without reviews attached. They don’t help buyers make an informed decision, and they can damage reputations without any accountability as to why.

All told, Amazon’s current review policy is a mess. It doesn’t take into account the psychology of reviewing, and it’s patently unfair to authors and publishers. It’s not even fair to customers, who may be scared off perfectly good products because negative reviews are overemphasized. The only people who really benefit are the reviewers.

And, of course, Amazon, which loves those “verified purchases.”