Spotify Teardown: Inside the Black Box of Streaming Music

By Maria Eriksson, Rasmus Fleischer, Anna Johansson, Pelle Snickars and Patrick Vonderau Cambridge, MA: MIT Press, 2019. 288 pages. $19.95 Paperback (ISBN: 9780262038904)

Spotify Teardown: Inside the Black Box of Streaming Music is an ambitious, wide-ranging, and mis-titled book. Studying firms like Spotify is notoriously challenging; it is practically impossible for researchers (especially those with critical orientations) to gain access to the code or the offices that we might envision as the “insides” of such organizations. The present book is no exception: in spite of a large, prestigious grant from the Swedish government, the authors found their early entreaties to Spotify rebuffed and ignored (p. 1, 182). Undaunted, they embraced what they call an “attitude of outsideness” (p. 190), reframing their lack of access as a virtue, avoiding “a formal procedure of gatekeeper introduction” that, they claim, “would have limited and biased the research” (p. 2). The result is a methodological commitment to exteriority, which pays off in the form of delightfully inventive methods that reimagine the nature of the thing we call Spotify. Thus, if I suggest that this book should be subtitled “Outside the Black Box,” it is no insult: its greatest contribution may be demonstrating how important the “outside” is—both to Spotify’s functioning and to what we understand it to be.

As the book’s structure, acknowledgements, and quintuple authorship make clear, it is the outcome of a rangy set of projects that try to apprehend Spotify from many different positions. Each of the book’s chapters and research vignettes can be understood as a study of one aspect of Spotify’s exterior: one studies an online Swedish political campaign, aimed at keeping the company’s job growth in the country; another captures the network traffic that the client software sends out into a tangle of advertising servers; another analyzes a corpus of the company’s job postings, examining them for clues about corporate values and future directions.

These efforts are most interesting in aggregate, as potential research directions that others might pursue further; together, they usefully refigure Spotify as a “changing, interspersed, multidirectional” entity (p. 15), not simply a piece of software with a stable function or a company with an enduring mission. Individually, these projects are uneven. One of the more interesting interventions describes the authors’ attempt to put their own music on Spotify: they discover that third-party aggregator services play an important role in shaping what makes it into the catalog and what metadata comes with it. Others, like the analysis of job postings that finds “the most commonly used adjectives were better, strong, free, proud, new, global, creative, and excellent” (p. 200), are anticlimactic.

The most promising and successful of the book’s approaches is the telling of Spotify’s history through its rounds of venture funding. Reframing the story of Spotify through its finances rather than its features allows the authors to show how the cultural framing of “music streaming” has changed as Spotify has sought the funding necessary to run for over a decade without making a profit. The limits of this method become evident, however, as the book strains to find causal relationships between funding and platform change: new features are described as “certainly not Spotify’s own idea” (p.51) or “apparently, a main condition of the investment deal” (p. 55), unnecessarily and speculatively reducing the complexity of organizational decision-making to singular economic causes. Nonetheless, the history traced here will be of interest to readers who want to learn how companies like Spotify are embedded in financial markets.

The book begins and ends with the story of a legal threat: in response to the authors’ use of automated “bot” accounts for studying patterns in algorithmic recommendation, Spotify sent them a cease and desist notice, while also notifying their funder—a threat to academic research that was rightfully met with public outrage at the time. Ironically, this large response was occasioned by one of the book’s least informative methods. These bots worked quite literally outside the black box, varying inputs to a hidden system in order to analyze its outputs, but as the authors note, their experiments failed to produce sensible results: they found “entirely random patterns of recommendations” (p. 145), and sometimes, strangely, no recommendations at all. These results are presented without the statistics and level of detail that would be necessary for readers to produce their own interpretations, so we can only conclude that the recommender works in mysterious ways (or perhaps, not at all).

Given Spotify’s silence and sudden legal antagonism, it is understandable that the book takes an adversarial stance; however, this sometimes becomes an analytic problem, as a surprisingly tendentious style obscures its findings, leads to analytical non sequiturs, and mystifies the reasons why Spotify might do what it does. For instance, the multi-dimensionial refiguration of “Spotify” I praised earlier becomes less of a description than an indictment when the authors criticize Spotify’s “opportunistic, rather than innovative, strategy and the inconsistent ways in which the company positioned itself” (p. 41). Thus, common startup behavior is pathologized rather than critiqued, placing the authors in the strange position of policing the border around “genuine” innovation. Instead of identifying contemporary software companies as intrinsically heterogeneous and multiple, the book interprets change as duplicity and variety as incoherence. At one point, the book disputes founder Daniel Ek’s musical bonafides, as though his childhood musicality matters to its argument: “While it is often reported that Ek played the guitar as a child, there is hardly any evidence that music defined his social life as he grew up” (p. 41). Corporate behavior is either nefarious or mysterious, with very few alternatives.

The authors’ commitment to outsideness is thus Spotify Teardown’s greatest strength and weakness. We need more books with this aspiration: to throw everything we have at tech giants, inventing new methods to draw out their many aspects and entanglements. But we also need more reflection on these methods’ limits and the assumptions embedded in them. An adversarial stance is not enough. Rather than taking the outside for granted, as a position of integrity for researchers and a source of impurity for corporations, we might hope for methods and theories that break down the inside/outside dichotomy. As many of the arguments in the book demonstrate, we might be better off abandoning the figure of the black box altogether.

Nick Seaver, Department of Anthropology, Tufts University