Audible’s Senior Vice President of Data Science, Haftan Eckholdt, visited CDS last Friday to discuss how data science plays a vital role in maximizing the Audible customer experience.
Audible is the world’s leading retailer and producer of digital audio books. Key to finding, engaging, and retaining Audible’s subscription-based customers is identifying the texts and audio books that are most valuable to their clientele.
Since value is typically marked by what is purchased, generating a best-seller audio book list is relatively straightforward as its based on major trends in Audible’s transactional data. But, as Eckholdt explained, the ‘best-seller’ group accounts for perhaps only half of Audible’s customer base. How can we use data science to identify and understand the diverse and highly particular literary tastes of those outside of the trend followers?
Working with a repository of over 250 thousand English manuscripts and sound files, Audible is using data science / natural language processing to develop topic, style, and sound taxonomies. For example, what linguistic markers can best identify writing style so that the platform can recommend audio books that are written most similarly to the customer’s favorite writer? Do customers who share the same age, education level, gender, or marital status prefer the same plot lines? Should they? And, which accents and vocal ranges are most appealing to their customers in the US, UK, Germany, and elsewhere?
Buying a book is a highly subjective decision making process (rather than, let’s say, buying a kitchen skillet), and answering these crucial qualitative questions will help Audible match customers to audio books based on individual literary preferences rather than mainstream tastes or demography.
By using data science to illuminate the secret structures behind words and speech, this fascinating data-driven interrogation of literary style also promises to yield enormously valuable suggestions about how we perceive and assign value to literature and sound. For those looking to apply their data science skills to questions of literary value, the time is ripe to apply for a coveted spot at Audible.
by Cherrie Kwok