This post is the second part in a series aimed at helping readers to more effectively identify books they will enjoy—that is, to invest their reading time wisely, and to reap all of the dividends accruing therefrom.
Say what you want about Amazon: the company has more data on American book buyers than any other entity in the world.
The New Yorker recently surveyed the rise of the corporate giant from its humble origins in the book business. Founder and CEO Jeff Bezos—no idiot, that guy—believed that by compiling a massive database on consumer preferences, he could foresee market trends faster than any other industry players.
Not twenty years after its founding, Amazon obtained a patent on what it’s calling “anticipatory shipping.” That’s right. Jeff Bezos thinks he now has enough data to know what you want even before you do.
Seems like with all of that data floating around, the company ought to be able to give a pretty good book recommendation, right?
Well, yes and no.
Let’s say that I read and loved Marilynne Robinson’s Gilead. (Which I did. It’s one of my favorite novels of all time—it didn’t get the Pulitzer by accident—and it comes with my highest recommendation. But then again, you don’t know me and I don’t know you, so you’ll have to take that recommendation with a grain of salt.)
Let’s say that I read and loved Gilead and now I want to know what to read next. If I pull up that book on Amazon, I’m given a list. “Customers Who Bought This Item Also Bought”:
- Tinkers by Paul Harding (Read it but didn’t like it.)
- Housekeeping by Marilynne Robinson (Read it—phenomenal.)
- Home by Marilynne Robinson (Read it and liked it, but not quite as much as Gilead or Housekeeping.)
- The Way to Rainy Mountain by N. Scott Momaday (Read it, but still not quite sure what to make of it—and not at all sure what it has in common with Gilead.)
- The Passion According to G.H. by Clarice Lispector (Never heard of it.)
- A Visit from the Goon Squad by Jennifer Egan (Haven’t read it yet, but it’s on my list.)
…and then another 16 pages of recommendations.
Those recommendations are not ranked or numbered in any way, nor do they come with any information about why other customers bought both books.
More pertinently, they also come with no guarantee of whether other customers liked the other books. Or whether they even liked Gilead, for that matter.
All Amazon is willing to tell us is that members of the set of customers who bought Gilead also bought the 17 pages of other books listed in the section explicitly designated for that purpose.
The list includes both David Foster Wallace and Cormac McCarthy, two writers about as different as different can be—and neither, to my mind, especially close to the esteemed Ms. Robinson.
I don’t know about you, but that list doesn’t strike me as an especially credible source of recommendations. I’m looking for something more than a tabulated list of sales data.
It should easily be possible for Amazon to make such recommendations. Netflix, after all, compiles taste preferences based on movies you’ve watched and liked. It then recommends new movies based on that data—and, critically, tells you why it’s recommending them.
It also learns as you go. I have in the past relied on Netflix to recommend films I’ve never heard of, and I have enjoyed many of its selections.
In contrast, I have never chosen my next book by relying on Amazon’s similar user data.
It’s not just Netflix that offers better recommendations, either: several music recommendation engines have gotten very sophisticated in recent years.
Take Pandora, for example. When I told Pandora that I like the Zac Brown Band, it played a series of tracks tagged with descriptions akin to the following:
“Based on what you’ve told us so far, we’re playing this track because it features mellow rock instrumentation, folk influences, a subtle use of vocal harmony, acoustic sonority, and repetitive melodic phrasing.”
It also gave me the option to immediately thumbs-up or thumbs-down any song, and then it adapted to that feedback while continuing to justify its selections.
That beats the heck out of Amazon’s recommendations. Can you imagine getting something like this?
“Based on what you’ve told us so far, we’re recommending this book because it features gorgeously lush and evocative language, philosophical influences, a quiet and stately use of plotting, a deep sense of place, and profound emotional resonance.”
Amazon has its uses, but until it takes some cues from services like Netflix and Pandora, making book recommendations will never rank high on that list.
Fortunately, there are some services that attempt to apply the Netflix/Pandora model to book recommendations. Check back for a review of those services in the next installment in this ongoing series for discerning readers pressed for time.
Michael Noltemeyer is a third-year MFA candidate at the University of New Mexico. He is the Nonfiction Editor for Blue Mesa Review.