Let’s start with the idea of a recommendation engine. It’s an informal term to describe the complex algorithms that parse masses of data to make suggestions of the “if you like that, you’ll like this” variety.
Online services crave good recommendation engines. Back in 2006, Netflix launched a contest with a $1-million prize to the team that could improve its algorithm by at least 10%. It gave programmers up to five years to reach the goal, and awarded the prize after just three, in 2009. Now we know that Netflix has created upwards of 76,897 unique micro movie genres to feed into its algorithm to automatically create personalized recommendations.
Book buyers are likely to know another recommendation engine: Amazon’s. Each book’s sales page, even before the complete book description or price, has a section labeled “customers who bought this also bought” with five other books. People who purchased Teju Cole’s newly published “Every Day Is for the Thief” from Amazon also bought Dinaw Mengestu’s “All Our Names” and Lydia Davis’ “Can’t and Won’t,” for example.
Of course, Amazon’s recommendation engine works across its site, for blenders, pneumatic drills, diapers. But let’s stick to books.
The book recommendation engine is meant to stand in for the human salesperson behind the counter in a bookstore. Tell a bookseller what you like to read, or what the person you’re shopping for does, and they can steer you to a book that suits. Same goes for librarians.
In general, human recommendation engines have, John Henry-like, stayed ahead of the robot algorithm recommendation engines so far. But being more nuanced isn’t everything.
Book buyers who use Amazon have become very accustomed to seeing the “people who bought this also bought” option when browsing online. And so, enter Bookish.
Bookish, acquired by Zola Books in January, is an independent book recommendation engine, except it was never entirely independent. Bookish was founded by three major publishers with a big announcement and then a fizzling, much-delayed launch. It claimed editorial independence from its funders but never found a clear foothold online, struggling to reach readers.
Since being acquired by Zola, an independent book discovery site and online store, Bookish has been employed as the recommendation engine for the New York Public Library, as announced Monday.
“The deal was in discussion before Zola Books acquired Bookish,” Zola’s Chief Marketing Officer Lynda Radosevich told The Times in an email. Bookish’s recommendations have been added to a service called BiblioCommons, which is or will be implemented by public libraries that include Seattle, Santa Monica, Chicago, Omaha, Boston and more.
Does it work? Visitors to this section of the New York Public Library’s website see book covers go by in a scroll that looks not unlike Amazon’s more static “people who bought this also bought” row. Clicking through to any single title pulls up “Bookish Recommends” along the right-hand side.
Does it work? Andrew Solomon’s “Far from the Tree: Parents, Children and the Search for Identity,” brings up Jared Diamond’s “The World Until Yesterday,” which doesn’t share the topic but is another serious nonfiction book. The next is David Nasaw’s “The Patriarch,” a biography of Joseph P. Kennedy; it could be said to be a bit like Solomon’s book, in that it deals with parenting, obliquely, although it is more focused on the man. The next is the memoir, “Dwarf” by Tiffanie DiDonato, which is lighter than the other books but does connect with Solomon’s in a deeper way, in that it’s the story of someone who was different from her parents, which is the specific subject of “Far from the Tree.”
By comparison, Amazon’s recommendation engine falls a little flat. Its customers who bought “Far from the Tree” also bought Solomon’s previous book, “The Noonday Demon,” in both paperback and hardcover, as well as his novel, “A Stone Boat.” Anyone online could click on Solomon’s name and see what other books he’d written for themselves. Closer to the mark is “Monkey Mind: A Memoir of Anxiety,” by Daniel Smith, but it’s less connected to Solomon’s new book than to “The Noonday Demon,” which is about depression.
As far as this example goes, readers will be well-served by the recommendations Bookish provides on library websites. Zola says it has 780,000 titles in its database and it has figured out how to classify them so the connections make sense.
But Zola is in the business of selling books. So why provide tools to library borrowers? In a model not being employed by the NYPL, Radosevich explained, “clicks on recommended books point back to book pages on Zola’s Bookish.com and partners earn a revenue share for any sales.”