By Jeremy Greenfield, Editorial Director, Digital Book World, @JDGsaid
Would you buy a book if it was like other books you bought and you knew others who bought what you liked to buy bought that book, too?
The answer to that convoluted question is one of the open secrets of Amazon’s success in the e-tail business: Once you’ve sold products to consumers at low prices and shipped them at little cost, suggest they buy another product – but not just any product, a product selected just for them by a complicated algorithm based on what they’ve searched, bought and otherwise shown interest in.
Consumers love it.
“It works really well because I’m always amazed at the books they suggest,” said Neesa Sweet, 63, a consultant from Highland Park, IL.
“I especially like it when I am looking for gift ideas,” said Jeryl Marcus, 47, a stay-at-home-mom with two daughters.
While the exact formula for Amazon’s recommendations engine is a closely guarded secret (you can read the official patent from 1998 here), most consumers assume it’s written solely to help them find more, better things to buy.
When asked how it works, consumers told Digital Book World some variation of “it’s an algorithm that tracks my past behavior and suggests things I might like.”
“Recommendations are generated based on my past purchases and even my past search queries,” said Griff Hanning, 28, a Colorado Springs, Colo.-based owner of a handyman company and founder of start-up HeyDont.com, a user-generated content site for stories of what not to do.
“I think the recommendations are generated simply by cross-referencing the authors and genres of the current purchase. For some recommendations it appears past purchases are referenced as well,” said Mike Pascale, 48, a freelance writer and storyboard professional from Modesto, Calif.
The fidelity of Amazon’s recommendation engine became a topic of conversation in the publishing world last week. PaidContent first reported suggestive remarks that Larry Kirshbaum, the head of Amazon Publishing, made during a public forum at the State University of New York’s Stony Brook Southampton campus. His remarks implied that the Amazon recommendation engine would favor books published by Amazon Publishing. Amazon Publishing is the New York-based book-publishing arm of the online bookseller.
He said: “I know there have been criticisms of Amazon in the press and from some sources, but the agents have by and large been extremely excited about the fact that we can publish in a different way. They love the idea of the database, that we can actually reach readers directly, that we can use — of course all of you who are, I’m sure, Amazon customers know the suggestible idea — if you like X, you’ll like Y. We’ll be the Y. We’re going to show some really innovative marketing ideas to sell books in ways that they’re not being sold right now.”
When Digital Book World asked an Amazon spokesperson if this meant that Amazon’s recommendation engine would be rigged to favor Amazon Publishing books, the official reply was, “no.” Amazon declined to provide further comment on the matter.
According to former Amazon engineers we spoke with, it’s unlikely that Amazon’s recommendation engine is set up to do anything but recommend products that users are most likely to buy – although, there is rumored to be discussion within the company about whether it can be used for marketing purposes.
“There was quite a bit of pressure from marketing back in the 2000 to 2002 period when I was there to bias the recommendations toward more profitable items, but it never happened,” said Greg Linden, a senior manager and principal engineer at Amazon until 2002 when he left to pursue his Master’s in Business Administration at Stanford. Linden was alone among ex-Amazon employees we spoke with that agreed to talk about the company publicly.
According to Linden, online testing confirmed that biasing the recommendation engine in that way would result in less revenue for the company.
“[It's] Certainly possible that things could be different now – it is a decade later,” he said. But “anything that doesn’t maximize the usefulness to customers of the recommendations will hurt sales, and it is hard to make up for that huge cost any other way.”
Other Amazon observers we spoke with suggested that it was as important to Amazon for it to maintain the integrity of its recommendation system as it is to Google to maintain the integrity of its search results.
The Rigging Consumers Are Really Worried About
Consumers aren’t so much worried about whether Amazon rigs its own system; they’re worried about a different kind of foul play – by authors.
“I used to trust book recommendations until I saw how authors and self-help gurus were using the ratings system to boost their sales,” said Nicole Guillaume, 32, the owner of a dog-training company in Corona, Calif.
According to Guillaume, a well-known self-help expert emailed her asking for “help” on Amazon in the form of a book review on the day her book came out. Hundreds of people immediately responded with five-star reviews, despite the unlikelihood of them actually owning the book since it just came out that day. A few weeks later, another author tried the same thing with Guillaume and had similar, positive results.
“I am disgusted with such ploys,” said Guillaume. “I understand that these experts and authors want to increase their sales, but the rating system was created so that real people could create real reviews. Having people create reviews of books they’ve never read totally defeats the purpose.”
A spate of positive reviews on Amazon also spooks Holly Wolf, 51, of Fleetwood, Pa., who is chief marketing officer for Conestoga Bank.
“I generally use Amazon for business books. I look for the number of reviews. If they are all positive, I’m suspicious,” she said.
You have to take the book reviews “with a grain of salt,” said Jay Buerck, 29, from St. Louis and the chief operating officer for an online reputation management company.
“If the book only has a few reviews and all of the reviews are five stars and overly positive testimonials about how great the book is, I tend to discount the reviews,” he said.
Unlike Guillaume, Buerck is a little more generous toward authors he suspects of manipulating the reviews system.
“While I don’t think that ‘sock puppetry’ is being employed by these authors or publishers, it’s hard not to think that their friends were huddled over their computer typing away this glowing review of the book,” he said. “A good mix of honest reviews is the best way to go when it comes to online reviews. It’s tough to please everyone all the time.”
Write to Jeremy Greenfield
Amazon image via Shutterstock