Beyond the online bookselling chimera

Nathan Maharaj

Nathan Maharaj is the Senior Director of Merchandising for Kobo, where he manages a globe-spanning team of booksellers. He’s written here about the online bookselling chimera to give you a glimpse of what you can expect from his Tech Forum 2017 keynote talk, Bionic Bookselling.

Over the 20-year migration of so much of bookselling onto the internet, whether it’s selling physical books online or my line of work, which is selling ebooks to customers on their mobile devices or e-readers, the industry in general has struggled to find a mode for conveying the expert personal touch of a bookseller. The solution that seems to have been settled on long ago is a chimeric offering consisting of some form of recommendation service employing collaborative filtering alongside human-curated assortments of new releases and value-oriented promotions. Machines selling books the way machines sell books over here; humans selling books the way humans sell books over there.

There’s merit to each approach. The efficiency and elegance of a good recommendations system is a thing of beauty. Machines are able to observe millions of humans performing millions of actions and find otherwise undetectable patterns and opportunities. Somewhere out there is a thriller reader who never reads romance, but he might really enjoy a book from a new romantic suspense imprint. Why? Not because a robot read it and thought it was pretty good, and not because of some kind of keyword matching that saw a strong correlation of “love” and “gun.” Rather, it’s because a machine observed that a significant portion of the people who engage with these books over here (in this case the thrillers our reader loves to stay up late with) also have a tendency to engage with those books over there (action-driven romance stories). Aided by machine-driven recommendations, a reader can find themselves happily crossing categories without the thought of categorization (let alone notions of editorial heresy) ever crossing their minds.

Recommendation systems work. They drive sales. Customers engage with them. If they have one drawback, it’s that they take a lot for granted, such as the customer’s attention and willingness to purchase. Automated systems struggle to convincingly convey urgency, or to meaningfully differentiate one customer-directed message from the next. Why these books? Why now? These are complications far beyond offering a reader a title to fill a likely gap in their library. Automating these aspects of bookselling is the equivalent of passing the bookselling equivalent of the Turing Test.

Selecting as-yet unproven titles, grouping them and other books into lists that readers want to consume whole, absorbing and negotiating promotional timing and pricing, crafting messaging and positioning around offers — these are all enormously complex things that a human bookseller does very well.

And yet, no matter how broad-minded and smart a promotion on romance books may be, it’ll be of little appeal to someone who seldom reads anything but literary fiction. And if you have a small but active group of readers interested in big ideas in business and economics (and you really love selling these kinds of books like I do), how do you engage them meaningfully without alienating everybody else or spending a totally disproportionate amount of effort on them? And will we ever consistently please the person who loves reading hard-boiled detective stories and popular science books about dinosaurs? These are hard constraints for a human to break.

What we’re learning at Kobo is that the future of bookselling is going to include a blend of human curation and automation driven by Big Data methodologies. Machines aren’t the most dynamic retailers. The skills of a bookseller don’t scale. Rather than focusing on minimizing these weaknesses, we’re delegating work between machines and humans to mutually amplify their strengths. Increasingly, we’re spending human effort on identifying books from every category that are the best of their kind, creating contexts to grab attention and drive sales, and employing machines to deliver the right books to the right readers. We’re spending less time choosing whether to curate or automate any given initiative, and instead focusing on how to thoughtfully curate automated outputs and how to automate distribution and presentation of curated assortments.

Rather than leaving machines alone to process a web of reader-to-book-to-reader relationships in isolation, we’re inserting the curiosity and judgement of human subject matter experts — booksellers! — to help solve one of the oldest problems in online bookselling and engage readers in ways we hadn’t thought possible until now.

If you want more of Nathan’s insight follow him on Twitter and LinkedIn. And, if your curiosity is piqued and your thoughts are zooming around in your head, register now for Tech Forum 2017! We like your kind!