Culture industries increasingly use our data to sell us their products. It’s time to use their data to study them. To that end, we created the Post45 Data Collective, an open access site that peer reviews and publishes literary and cultural data. This is a partnership between the Data Collective and Public Books, a series called Hacking the Culture Industries, brings you data-driven essays that change how we understand audiobooks, bestselling books, streaming music, video games, influential literary institutions such as the New York Times and the New Yorker, and more. Together, they show a new way of understanding how culture is made, and how we can make it better.
—Laura McGrath and Dan Sinykin
For an increasing number of people, reading means listening to streamed audio files through a smartphone. The audiobook has a long history,1 of course, but what is new is its commercial impact: For the first time, audiobooks can no longer be seen as a niche market. Now, the audio medium competes with print books and ebooks for the attention of book readers in a large and diverse range of national book markets. Most people in the book trade believe that the audiobook share will continue to grow in the coming years. According to the Association of American Publishers (AAP), 8.1 percent of the revenues of the total US book trade in 2021 came from audiobooks.2 This figure can be compared to ebooks (11.6 percent), but also to change over time: in fact, it is audiobooks—in contrast to all other book formats—that have shown a rapid and steady increase over the past ten years.
The audiobook boom is altering the book business and reading culture. It provides opportunities for people to read more and in new ways, but it also affects how “reading” can be understood. In highlighting the complexities of popular fiction reading, Janice Radway once famously objected to the metaphor of consumption when equated with reading. Reading is not a passive thing, she claimed.3 I agree, and a multitude of readership scholars have convincingly proved this to be true: reading is active, participatory, social.
But thanks to audiobooks, we might need to update this reasoning. In fact, some of the audiobook practices surfacing indeed seem to be exactly this: passive. One can easily pose the argument that the rise of audiobooks is a sign of an ongoing crisis of our book culture, where people no longer actively engage in books but lend them half an ear as a mere distraction while doing something else. People are reading while doing the dishes, driving, working out, sleeping, etc. Can such practices really be regarded as reading? In any case, passivity must be a problem for literature, right?
In one sense, it is true. But it is also not true, since print-book sales are not dropping when audiobook streams are skyrocketing. Perhaps audiobooks are not primarily competing with print books and ebooks, but with podcasts and other audio media? If this is so, audiobooks could be regarded not as a threat to our book culture but, rather, as a defender. Well, I don’t believe that to be the case either. But I do believe that audiobooks are about to fundamentally change our reading habits.
In fact, what appears to be happening is rather that people are expanding how they make use of books. Or, if you will, expanding what reading is, and what it can be.
But how are audiobooks changing how we read? And how does one investigate such changes?
I map contemporary audiobook reading practices, simultaneously in detail and at scale, and do so by drawing from a unique dataset of book consumption data from Storytel:4 a global book-streaming platform and second to Amazon’s Audible, one of the largest in this market. The dataset I employ covers the consumption habits for all Storytel users in Sweden for 504 titles in the commercial top segment.5 These private reading patterns are tracked per individual user and per hour during a one-year period (from May 1, 2020, to April 30, 2021). In terms of details, this means that we can follow the patterns of each individual audiobook user for these particular 504 books: which of the titles they listen to; which titles they complete and which ones they abandon; when they listen, and when they stop listening. In terms of scale, in total ~74.6 million data points are analyzed, distributed over ~432,000 individual readers. The main difference between Sweden and the US is that audiobooks are even more popular in Sweden, with a market share of around 25 percent, mostly because of the impact of the all-you-can-eat model of all major Swedish subscription-based book-streaming services.
Let’s start by looking at temporal aspects. Audiobooks are read during all hours of the day. This is a truism, obviously, but it is worth mentioning, since (on the Storytel platform) audiobook reading diverges from ebook reading in this respect. Audiobook reading peaks in the afternoon, has higher consumption rates during mornings than evenings, and remains steady at night. Ebooks, on the contrary, have a distinct peak of consumption in the evening hours, especially between 8 and 10 p.m., when almost a fifth of the total consumption takes place (see Figure 1). Ebook consumption furthermore rises on the weekends, while the tendency for audiobooks is the opposite: Mondays to Fridays are when people listen the most.
Ebooks, then, are read mostly during traditional leisure times (evenings, weekends). But audiobooks are read at any time, not least during traditional working hours. Let’s assume that the temporal patterns for ebook reading are a proxy for print reading (if we believe that reading with your eyes is a fairly similar practice even when reading in different formats). In that case, audiobook reading clearly diverges from other kinds of reading.
This pattern is anticipated and is explained by the flexibility of the audiobook medium. That’s another way of saying, again, that people can and do read audiobooks in more situations than they might read in other formats. That said, I find the drop in audiobook listening around dinnertime especially telling, as it highlights the unsocial part of book reading. Apparently, it is fine to combine audiobooks with work, studies, and other daytime activities, but when people eat dinner, they need to plug out for a couple of hours to spend some time with their family and friends.
Interestingly, regarding audiobooks, all kinds of literature in the dataset share these temporal reading patterns. I had anticipated finding differences between, for instance, genres, but there were none. This holds true even on the level of individual books, be they high-pace thrillers or Nobel Prize winners. For instance, there are no individual books in the selection that are primarily consumed at night, or only in the daytime. If we had access to data on all books, and also more peripheral titles, we would likely see larger discrepancies. Even so, the result indicates that what people read doesn’t seem to immensely affect when they read, generally speaking.
One explanation for the stability of the listening patterns has to do with the length of novels. Audiobooks regularly exceed 10 hours of running time, sometimes 20 hours. You generally need several days to complete a full-length novel, which seems to make many readers fit audiobooks into their everyday routines. Another reason has to do with audiobook reading being private, which allows personal habits to come forward. You seldom listen to audiobooks in social situations or together with others (as shown by the dinnertime drop). And everyone’s personal habits taken together get us the generic curve in Figure 1.
Personal habits: there you have it. It is not books or genres that first and foremost steer reader behavior but individual preferences of use. Many readers seem to have integrated distinct audiobook-reading practices in their lives and to stick to them. What varies are the practices between individual readers and groups of readers. Most notably, heavy readers are often night readers. Or phrased differently: night readers are often heavy readers. When I adjusted various thresholds regarding minimum consumption for users, the pattern was clear: the higher the consumption, the more users were biased toward night reading. A probable explanation is that night readers have fewer disturbances than daytime readers. If you use audiobooks when going to sleep, it is easy to just keep listening for hours, and if you fall asleep, the Storytel app tracks three hours of reading before it stops. If you are listening to audiobooks while working night shifts, you can likely go on for a long time.
Thus, the choice of book medium matters, but so does the type of audiobook reader. But can we delve deeper into reader types and the different ways audiobooks are used?
One approach is to group audiobook readers by two parameters: number of unique books consumed and number of total minutes streamed. The scatter plot in Figure 2 shows all individual users as blue dots and how they distribute along these two axes. The red dotted line shows the result of a linear regression analysis, and its positive curve makes sense: in general, the more books you read, the more time you spend reading on the platform. But what the scatter plot also tells us is that the users’ reading practices are indeed scattered. If the linear relationship were perfect—that is, if all users completed all books they started listening to and never listened to the same book more than one time—the blue dots would line up neatly along the red dotted line. But here there are lots of readers far away from the regression line.
Based on these metrics, and by going through the habits of lots of individual user cases, I’d like to focus on three types of audiobook readers. I call them repeaters, swappers, and superusers. In the scatter plot, these kinds of readers are outliers, but remember that we only have access to reader data on 504 commercially strong titles, which account for a fraction of the titles on the Storytel platform. The outliers in this figure are, therefore, the users whom we have the most data on; they are the readers who best fit our selection criteria—that is, heavy consumers of commercially strong titles.
Repeaters are audiobook readers who listen to the same books over and over again. They are heavy readers who listen a lot, often during the night. Since they have particular tastes, they are less affected by things such as platform design, categorizations, and recommendations; these users already know what they like.
Take for instance Reader A,6 who reads almost every night, and exclusively in the late evenings and at night. Their sessions regularly start between 10 and 11 p.m. and then go on for three hours or more. The recurring 180-minute sessions suggest reading audiobooks while going to sleep or when having sleeping trouble. Curiously, Reader A reads only books in the Little Beach Street Bakery series by the British romantic fiction writer Jenny Colgan; in fact, the first book in the series, The Little Beach Street Bakery (2014), make up over 95 percent of the user’s total consumption. This particular book seems to be repeatedly used as a sleeping pill, taken more or less every night. Another repeater, Reader B, reads even more, and reads only Stieg Larsson’s Millennium Trilogy. The three books are repeated in order, over and over again, during days and nights, for over five hours per day on average. For this user, then, the book series seems to be used as Muzak: a kind of life soundtrack constantly plugged into their ears.
Audiobooks are about to fundamentally change our reading habits.
Swappers are readers who try out lots of books, only to abandon most of them early on and move on to others. They are frequent readers, but not as heavy as the repeaters. They have few preferences regarding authors and narrators, and could be characterized as open, curious, and impatient, highly affected by platform design, categorizations, and recommendations.
A swapper like Reader C reads quite a lot, around one hour per day on average, with consumption peaks in early mornings and around noon. More notable than the number of hours spent on audiobook reading is the high number of unique books and authors: they listened to over 90 unique books in the selection by more than 60 different authors over the year studied. The typical pattern is to rapidly try out a number of books before sticking to one. For instance, in November 2020, the user tried out three Swedish crime novels, a British romantic comedy, and a middlebrow literary novel before settling with Delia Owens’s best-selling Where the Crawdads Sing (2018), which they completed. Then, Reader C tried another nine novels and abandoned them before completing the next one.
Superusers, finally, are binge readers who listen to books all the time. They are very heavy readers who read books by lots of authors (though they also repeat favorites), and they are highly biased toward popular fiction. We can characterize superusers as semi-open readers: they have their preferences regarding genres and authors, but since they read so much, they are keen on trying out new books by unfamiliar authors.
Reader D reads almost all the time, both during the day and at night. When they finish a book, a new one is instantly started, and very few books are not completed. The user also binge-reads entire book series. In particular, series by Swedish crime writers, such as Lars Kepler and Camilla Läckberg, were reread several times during the year studied.
These kinds of readers show habits fostered by the Storytel platform, either due to the audiobook medium (repeaters), or the subscription-based consumption model (swappers), or a combination of both (superusers). Although the example users showcased here are extreme, they point at reading practices more broadly. They also highlight that reading in the age of streamed audio is something quite far from what we generally mean when we talk about reading.
But regardless of how we choose to define reading, the money tells its own story. That’s because Storytel and similar actors pay publishers not per book but per number of streamed minutes on their platforms. So as long as the audiobook player is playing, all kinds of audiobook consumption are, in economic terms, treated as “reading” and are equally important for publishers and authors. The activity grade of the user doesn’t matter. It means that Reader A above bought The Little Beach Street Bakery just over a hundred times between May 2020 and April 2021. Or rather, that Reader A generated revenues for the publisher just over a hundred times the revenue of a single streaming of the book.
This is a crucial difference from when books are sold as entities, and it has certain consequences. No one would buy a hundred copies of the same book, but with streamed audiobooks, this is actually happening. As I have noted elsewhere, payment per streamed minute will make publishers and authors much more interested in books that are completed by readers, since they will generate more revenues back to the publishers and authors. It is reasonable to believe that a similar logic will apply for books that are often repeated.
Books have been sold as entities for hundreds of years. This binary business model—either you sell a complete book, or you don’t—is now being challenged by a new one, where readers can choose to quit anywhere within a narrative and, albeit indirectly, only pay for the parts read. Or, as the example of Reader A shows, they will pay (again, indirectly) for the same book over and over again. What is happening with streaming audiobooks can be characterized as a triple revolution, one that is taking place simultaneously on the levels of the medium, the distribution, and the business model.
Will it change our book culture? Definitely. As I have tried to show, streamed-audiobook platforms make us read in new ways, and the business model results in publishers and authors getting paid in new ways, which will affect what is written and published in the first place. The change is already happening.
- For further insights on the history of audiobooks, see Matthew Rubery, The Untold Story of the Talking Book (Harvard University Press, 2016). ↩
- The figures for 2021 come from Porter Anderson, “AAP StatShot: The United States’ Publishing Industry Gained 12.2 Percent in 2021,” Publishing Perspectives, January 26, 2022. It should also be noted that the Audio Publishers Association (APA) reports higher figures for audiobook revenues than AAP. The most thorough account of the audiobook format in the US from a publishing studies’ perspective is found in the chapter “The New Orality” in John B. Thompson’s Book Wars: The Digital Revolution in Publishing (Polity Press, 2021). ↩
- Janice Radway, ”Reading Is Not Eating: Mass-Produced Literature and the Theoretical, Methodological, and Political Consequences of a Metaphor,” Book Research Quarterly, no. 2 (1986). ↩
- This essay is based on the research project “Patterns of Popularity: Towards a Holistic Understanding of Contemporary Bestselling Fiction,” led by Karl Berglund and funded by the Swedish Research Council 2020–2023 (SRC 2019-02829). The project involves a collaboration between Uppsala University and Storytel, where the latter provide book-consumption data from their platform to the project members for scientific investigation. None of the researchers involved have commercial interests in or other relationships of any kind with Storytel. All users are anonymized in the source data and unknown to the project group. ↩
- The “commercial top segment” of fiction in Sweden is operationalized with the help of three sources: 1) all fiction bestsellers in print in Sweden from 2004 to 2020 (334 titles); 2) the most popular fiction titles on the Storytel platform that weren’t already bestsellers (65 titles); and 3) all “Storytel Originals” from 2016 to 2020: books that were produced by Storytel directly for the audiobook format and which are increasingly popular on the platform. In total, this gives us 504 titles, where most are genre fiction—primarily crime fiction and romance—but the dataset also contains a fair amount of prestigious fiction by award-winning authors, as well as middlebrow literary fiction. ↩
- Due to ethical concerns, the examples of individual readers showcased in this essay are not real readers. Instead, they share characteristics with several real readers in the dataset, without exactly matching one specific user. ↩