Analyzing a book’s search terms can result in useful—and sometimes surprising—insights. Did you know, for instance, that some search terms used for romance books are “alien menage romance,” “awkward romance” and “dinosaur romance?”
That’s the kind of information Kadaxis collects and analyzes, and for two reasons: to help readers find books and decrease publishers’ risks when investing in books.
“We found that, for some genres of commercial fiction, we could detect a weak signal, but it was almost impossible to accurately identify outliers,” says Founder and CEO Chris Sim. “We concluded that external factors such as marketing budget, the editing process and seasonality are all material contributors to a book’s success—data we didn’t have access to.”
Kadaxis was founded out of Sim’s desire to test theories he’d developed while employed at a New York-based publishing startup. He calls it his grand attempt at predicting the marketability of a book from its text.
“We spoke with a lot of authors, publishers and other industry people and kept hearing the same challenges relating to discoverability and metadata,” says Sim. “That led to analyses of how publishers and authors were creating and using data and the reverse engineering of how book search engines functioned.”
Sim decided to use machine learning technology to analyze a large sum of books and reviews. His technology recommends BISAC categories and keywords for books, a standard used by many companies to categorize books based on topical content. The result was Kadaxis’s two products: BookDiscovery.co, which lets readers search for books using phrases that are natural to them (e.g, “sad story”), and AuthorCheckpoint.com, which recommends BISAC categories when an author uploads her manuscript and analyzes marketability.
To Sim, the key to great book discoverability is replicating word-of-mouth recommendations. Computer-based recommendations must understand the books they’re recommending, a reader’s needs at a given time, and the reader herself—similar to the way a human would.
“Most of the work we’re doing now is understanding the different nuances of books through data analysis and attempting to determine which elements are most compelling to readers,” Sim says.
From that, Sim hopes to improve the ratio of readers to books and help authors reach larger audiences.
“The hardest part of writing a book should be writing it, not finding someone to read it,” Sim explains. “Some people think that applying algorithms to art detracts from it, but I believe technology can help scale the appreciation of it.”
To get all the ebook and digital publishing news you need every day in your inbox at 8:00 AM, sign up for the DBW Daily today!