Reviewing Amazon.com

I received this note with a product I bought on Amazon the other day. Leaving aside the horrendous grammar, I got to thinking about the algorithms behind Amazon’s product rankings that render buyers’ reviews and ratings so crucial to sellers (this note asks me to give a favourable seller rating, but I’d like to focus on buyer reviews). 

Take Kitty Thomas for example. On her blog, she clearly explains why, as an author, Amazon reviews are important:

When an author gets at least 20 reviews for their book on the site, Amazon starts actively recommending that book to people. And the page gets seen more. And this isn’t just a benefit that dries up. The more reviews you have, the more you get recommended.

There is never a point in time where you can have “too many Amazon reviews” unless there is a point in time where you can have “too many sales” and I’ve yet to meet an author or publisher who thinks you can have too many sales.

This is why I ask for Amazon reviews. It is not to feed my vanity or to make me look cool. It’s because it helps sales. I want my book to be seen and I want my book to be bought. I worked hard on it.

I couldn’t find any data to back up her assertion that 20 reviews is the minimum for Amazon to start recommending a book, but in general reviews are important to sellers; we know that.I couldn’t find any data to back up her assertion that 20 reviews is the minimum for Amazon to start recommending a book, but in general reviews are important to sellers; we know that.

I then discovered this (old) article that explains the secret behind Amazon’s $2.7 billion dollar revenue each year (it’s the very simple ‘was this review helpful to you’ question, if you were wondering), which explains how Amazon has modified their reviews system to ensure that the most helpful reviews always appear first.

Amazon quietly bumps the three most helpful reviews to the top. It tries to balance positive and negative reviews, so shoppers get a balanced perspective. An interesting side effect is how these selected reviews get more votes. If they are controversial (in that not everyone agrees they were helpful), their ratio goes down, allowing the most helpful reviews to bubble up past them.

This makes it a self-managing system, letting the reviews people find the most helpful to maintain their standing at the top of the list. The result is an understated implementation that works great.

That explains a seller’s anxiety about reviews, and taking that further, neutral reviews specifically. A neutral review is not going to help the seller because it is unlikely to be voted helpful and increase sales of his product; it’s likely to bubble down to obscurity at the bottom instead.

On to another (related) topic. According to the article, only one in 1,300 people leave a review, and approximately one in 7,300 indicate whether reviews are helpful.

As someone who, like most people I know, buys stuff on Amazon reasonably often, it is clear that there is no motivation for shoppers to leave a review. If it’s a bad product then obviously people are motivated to leave bad reviews, because it becomes a personal thing – this feeling of being cheated. But if it’s only average (not exceptional), there’s no motivation to leave a comment. The catalyst, to me, is passion. Harry Potter and the Deathly Hallows for example, which the article mentions, now has 3,505 reviews of which 2,720 give it 5 stars. It doesn’t require a rocket scientist to guess that the people who’ve left those reviews are massive Harry Potter fans – the kind who probably join the Harry Potter Alliance.

This doesn’t help most products on the site because let’s face it, most products do not have a personality, they don’t drum up extraordinary amounts of passion. So how can Amazon get more people to leave reviews for those kind of products, which no doubt people do buy? A camera bag for example, or a guitar?

Socializing the service is one mechanism that I’m surprised that Amazon hasn’t explored yet, for this purpose. I don’t mean creating profiles for users and so on. I mean giving users a view of which of their friends also uses Amazon: surely at least 50 of my friends use the service, if not more – plugging into Twitter or Facebook’s API is good enough to figure that out. I’d leave reviews if I knew my friends would find them interesting or useful. That can be taken further, and similar to Groupon or Dell Swarm, Amazon could give discounts to groups of friends who buy the same product over time based on mutual reviews.

There’s definitely an opportunity here – if Amazon doesn’t listen, perhaps eBay or Barnes & Noble will!

Image by soumit via Flickr , courtesy a Creative Commons license