Exclusive: What "Spotify Teardown" Means For Artists, Fans and Journalists

By: Cherie Hu

Published: 2021-01-11

Happy Sunday everyone!

Before I kick off my first Patreon-exclusive essay, I wanted to share the fantastic news that we’re now well past the $700/month mark on my Patreon page, which means I get to launch a Discord channel for Shapers + above—yes, that means you!

I’m going to be out of town (speaking at the University of Oregon) for all of this week, so will open up the Discord channel in early March once I can dedicate more of my undivided attention to the launch. I also recorded the first episode of the Water & Music podcast yesterday (with fellow writer/Patreon-er David Turner), which will be out very shortly!

Below is the first in an ongoing series of Patreon-exclusive essays, under the title Method v. Madness, which will offer deep-dives into more academic and theoretical writings on the future of music and entertainment. The aim is to tie the concepts in these essays to trends and actionable insights for people in and around music, in a way that is accessible and evergreen. You can find future installments in this series under the tags “exclusive essays” and “method v madness.”

Today’s inaugural essay is about a book (hence its lengthiness… apologies in advance), but most future installments will focus on shorter papers. I will try my best to pull from papers that are available online as free PDFs, so you can explore the resources yourselves as well.

Thank you all so much again for supporting and joining me on this journey! 🙂

—–

What Is Spotify Teardown?

Released last week via MIT Press, Spotify Teardown is a roughly 200-page, historical look at Spotify’s evolution as a company and as a product—and an investigation into the overlooked commercial decisions underlying Spotify’s alleged role as a “disruptor.”

The authors—Maria Eriksson, Rasmus Fleischer, Anna Johansson, Pelle Snickars and Patrick Vonderau—are all Swedish researchers and professors of media, economics and ethnology (I also interviewed Vonderau for my piece in Variety last year about music-streaming fraud). While the book is largely accessible to a wider audience, its tone is academic, drawing from key papers on cultural theory, media & platform studies, digital anthropology, psychology, politics and other fields.

One can boil down Spotify Teardown into two key questions:

The answers to these two questions are constantly changing—and Spotify, just like any other major/public company with a robust communications department, doesn’t always want these shifts to be 100% transparent.

One of the most fascinating and revealing turns of phrase in this book is how the authors describe audio files on streaming services as having “social lives“ (p. 78). No, they don’t mean their audio files are hanging out with each other and chatting at their local bar after work. Rather, they’re talking about how audio files are constantly interfacing with other companies and datasets outside of Spotify’s immediate boundaries, often for the sake of organizing music and categorizing and targeting users.

This is part of why a more multifaceted, systems-oriented approach to researching Spotify as a company is so important; as the authors write, “audio files cannot be studied as isolated or autonomous entities” (p. 83), but rather should be treated as one of many nodes in a complex and ever-shifting network of interlocking platforms, datasets and business models.

An intervention, not an exposé—and what it means for journalists

Unlike what the book’s title may suggest, what Spotify Teardown will not give you is an exposé of the inner workings of Spotify as an organization.

In other words, no anonymous gossip on internal departments or leaked product previews, and no never-before-seen perspectives on what happened when the company went public, launched in the U.S. or achieved other key milestones.

That may come across to you as a disappointment—but it actually serves the authors’ meta-thesis about the challenges of engaging tech companies meaningfully with knowledge creation, especially that of a more critical nature.

In fact, the authors tried multiple times to reach out to Spotify through official channels to interview employees one-on-one, and “to proactively engage Spotify in a conversation about its culture” (p. 186). But they faced a lot of resistance from the company due to the research project’s more critical nature, and realized that they would be largely on their own to unpack the service’s infrastructure as outsiders. Hence why the book uses the word “teardown,” instead of merely “behind the scenes.”

In fact, in May 2017, Spotify sent a cease-and-desist letter to the Swedish Research Council to pull its $1+ million in funding from the research project behind Spotify Teardown—allegedly because the authors allegedly broke the company’s terms of service in their investigations.

While they didn’t collect any personal data or invade anyone’s privacy, the authors did stage concrete “interventions” and “breaching experiments” on Spotify’s infrastructure, including but not limited to:

“The only way to go to the heart of what these new platforms and actors really are is to break their terms of service,” co-author Pelle Snickars told me over Skype. “From an academic perspective, this is nothing new. All studies involving algorithmic auditing have to do this.”

As Snickars and his co-authors wrote in their book’s opening chapter, this approach was arguably to their added benefit: “to follow a formal procedure of gatekeeper introduction or even to seek official endorsement by Spotify would have limited and biased this research” (p. 2).

This immediately made me reflect on my traditional process as a freelance writer covering these types of companies. I’d like to think of my approach to researching and writing articles as entrepreneurial, analytical and thorough, and I have a solid network of friends and sources in the music industry to corroborate or challenge my findings.

But if I ever want to write a piece about Spotify, there’s still an expectation that I go through their official channels, i.e. PR reps, to organize interviews with a carefully-curated selection of the company’s execs. Most of them have likely been trained in what not to share, and how to frame the little that can be exposed.

Since finishing this book, many questions have been swirling in my head: What might a more explicitly interventionist journalistic paradigm look like in music and tech? How might writers who are covering their beat as outsiders look past what the authors refer to as “company spin and ‘industrial self-theorizing'” (p. 2)—e.g. rewriting press releases, or pulling only from academic and market research commissioned by the company itself—and take matters of unpacking the actual backend of a product or business into their own hands? How might journalists collaborate more with academia to bring these kinds of projects to a wider public?

This isn’t far from reality at all, and in fact lies at the heart of investigative journalism. Many writers are also embracing the paradigm of immersive journalism—which celebrates rather than eschews authors’ firsthand experiences in coloring news reports and cultural commentary, making readers feel more like direct witnesses of a story (VR is a popular medium for this approach).

Music writers like Liz Pelly have conducted their own DIY experiments on Spotify; Jane Wong, a software developer renowned for reverse-engineering sites like Facebook, Instagram, Google, Pinterest and many others herself for glitches and early/internal features, is a frequent source for sites like TechCrunch. I think the publication and support of books like Spotify Teardown will encourage even more interventionist approaches to covering tech companies in the future.

My top three takeaways

Another important point is that not all of the ideas explored in Spotify Teardown—including but not limited to platform centralization, algorithmic bias/subjectivity and the rise of mood-based curation—are inherently “new.” If you work in music, you have likely dealt with some form of these issues throughout your entire career.

But this lack of “newness” is ultimately more illuminating than disappointing. It reinforces how Spotify is simply a new actor capitalizing on old frameworks and sources of power across media, technology and politics—debunking the myth that Spotify is a “lone innovator” that can be understood best only in its isolation.

Below are three of my top takeaways from Spotify Teardown, which I tie to more recent Spotify news that happened after the book’s copy was finalized—particularly Spotify’s foray into podcasting.

1. Spotify is a prescriptive mediator, not a neutral “intermediary”

Historically, Spotify’s PR strategy has relied on the concept of helping artists “bypass intermediaries” such as record labels to distribute their music. Yet over time, more onlookers have argued that in cannibalizing other intermediaries and in centralizing its power and influence, especially vis-à-vis curation and recommendation, Spotify would simply become another intermediary in itself.

Spotify Teardown takes this argument one step further. The authors argue that Spotify is not just an intermediary for music’s distribution; it is “a mediator … that actively reproduces the meaning of the songs it is supposed to distribute” (p. 171). In other words, Spotify is not a neutral company, but rather imbues the content on its platform with its own interpretations and motives, often beyond the content owners’ control.

To support their point, the authors devote many pages to how Spotify packages music on its homepage, in the pursuit of soundtracking its users’ lives. In particular, the authors argue, Spotify disproportionately promotes music as a functional device for “self-governance through mood control … and the management of psychological capital” (p. 123).

The features the authors mention as key examples include Spotify’s vast range of mood- and activity-driven playlists, the platform’s now-defunct Running feature and its discounted bundle with meditation app Headspace. More recently, Spotify also collaborated with an astrologer on a collection of horoscope playlists designed, in the company’s own words, to help users “reflect and find direction in their lives.”

In short, unlike what many musicians have suggested, the authors argue that music is not Spotify’s product. Instead, Spotify’s selling point is a self-serve emotional oasis: “music streaming framed as a deeply personal and intimate—even happiness-inducing—practice.”

This means that Spotify may end up “privileg[ing] certain ways of attending to music” (p. 136) over others, which has indeed trickled into a sense of privileging certain songwriting styles over others.

2. Spotify’s advertising market remains a black box for artists and consumers

According to its Q4’18 earnings report, Spotify makes over 88% of its revenue from premium subscriptions, but 56% of its users still don’t pay to use the service. I would posit that the majority of music-streaming consumers globally still do not pay for access to their service of choice.

From a research standpoint, this raises an interesting question about Spotify’s true status as an alleged “two-sided marketplace” for artists and fans. As the authors wrote, if “the majority of listeners still neither buy nor sell [music] … where is the market, and what is actually being sold?” (pp. 157–159).

The obvious answer: aaaaads.

While this is nothing new, the way Spotify’s ad-buy “market” actually looks remains quite opaque to consumers and, more importantly, to the musicians supplying the platform with their content.

To try to bring more transparency to this landscape, the authors of Spotify Teardown used an open-source tool called Wireshark to “eavesdrop” on the data traffic triggered when they pressed play on free and paid Spotify accounts.

As it turns out, a single stream triggered not just a play count for an artist, but also data traffic to multiple programmatic ad networks, including AppNexus, AudienceScience, MediaMath, Rubicon Project and Turn (now Amobee). The research team also intercepted data “packets” being sent to wider Internet backbone networks like the AOL Transit Data Network, content delivery networks (CDNs) like Akamai and Amazon CloudFront and cloud platforms like Google (to which Spotify is paying hundreds of millions of dollars over the next three years).

This network is largely invisible to the end user—raising questions about which elements of a user’s personal information are being transmitted from Spotify across these various pathways, and whether there is an easy way for said users to opt out. In a recent interview with Rolling Stone, Spotify Teardown co-authors Snickars and Eriksson suggested that Spotify should receive the same scrutiny as Facebook with respect to privacy practices and handling user data.

There are also potentially complex political implications. CDNs in particular—which deliver higher-speed internet access to whichever customers, usually concentrated in major urban centers, are willing to pay the premium fee—complicate debates around net neutrality, a principle that Spotify itself has supported.

In addition, the authors question what the benefits of advertising really are for “musicians—or cultural producers more generally—whose content is the incentive attracting both sides” (p. 155). While one obvious benefit is monetization, this comes at the expense of the artists’ abilities to control how their music is presented and contextualized (particularly as Spotify begins to sell algorithmic channels like Discover Weekly as ad products).

One potential future is that Spotify’s ad market may become more transparent as podcasts gain more prominence on its platform.

Unless they’re partnering with dynamic ad services through their networks, podcast hosts still exercise greater autonomy over which brands they choose to advertise in their episodes (which often materialize as permanent, “live-read” ads). As I argued for Billboard, Spotify’s acquisition of Anchor—a podcast distribution platform with an innovative direct-to-host ad marketplace—could lead to a genuine flourishing of a more transparent “two-sided marketplace” of independent producers on one end, and brands and advertisers on the other. Brands advertising “direct-to-artist,” and the artist being in the driver’s seat with respect to these deals, could be a genuine game-changer.

3. Spotify’s relatively closed, centralized present is in tension with its largely open past—and may come into conflict with its podcasting ambitions

While Spotify Teardown‘s copywas finalized and published prior to Spotify venturing more aggressively into the podcast world, many of the authors’ arguments hold even more strongly now that Spotify is evolving further as a publicly-traded media company.

Spotify was a vastly different company in 2009 compared to now in 2019, and one of the biggest unseen differences between then and now is that the service no longer uses a peer-to-peer infrastructure. In its early years, Spotify leveraged P2P to help scale the platform without the need to invest in its own servers or bandwidth, making the service much faster than its rivals in the music-streaming sector.

Then in spring 2014, TorrentFreak reported that Spotify was going to shut down its  P2P network altogether, resorting instead to its own central servers. The Spotify Teardown authors called the transition “a major infrastructural event”—but it was covered very little in the press, and wasn’t even written about on the company blog (and I had never come across the change while researching Spotify as a budding journalist).

The massive shift from P2P to central was also nearly unnoticeable by everyday users, and the Spotify Teardown authors argue that corporate storytelling played a significant role: namely, “the move toward centralization did not fit the company’s preferred image” (p. 63) as a democratic, equalizing music service.

Later in 2014, Spotify also killed its third-party app development platform, which had launched three years prior in an effort to open up some of Spotify’s data and resources to software developers to build their own music apps. The likes of Rolling Stone, Last.fm and the first version of Soundrop had built popular apps atop Spotify’s platform.

These two events also run in parallel Spotify’s gradual and now complete de-prioritization of third-party playlist curators, in favor of their owned-and-operated playlist properties. Like the P2P-to-central transition, this shift was never really communicated to subscribers because it wasn’t consistent with the company’s image.

In short, while it’s still true that technically any artist can distribute their music onto Spotify for little to no cost, the company’s ecosystem is much less open than it used to be.

But what does this mean for podcasting—a fundamentally open format?

Content regulation has always been difficult, if not discouraged, in the podcast world. Music licensing for podcasts presents a perfect example: because licensing agreements still comprise such a wild-west territory with little standardization of terms, the costs and paperwork for music and content licensing ultimately fall on individual hosts, not on platforms like Spotify.

In addition, because podcast curation and recommendation are such massive, unsolved engineering problems at scale, these functions remain largely distributed among many smaller platforms and individual human beings—i.e. peak “peer-to-peer.”

There has already been a lot of debate in the media about whether Spotify’s ownership of Gimlet and Anchor will bring an unprecedented level of one-size-fits-all centralization to the way podcasts are produced, curated, consumed and monetized. Some have written that Spotify may not invest as heavily in products that take advantage of its vast podcast catalog, in the same way the company has channels like Discover Weekly for music. As Eric Silver of the Bello Collective suggested, “Will curation even matter or exist when there is an in-house product?”

In an ideal world for the podcast industry—where Spotify builds and supports a thriving ecosystem of podcasts on its platform, including viable tools for hosts to build audiences on their own terms and engage with and solicit feedback from listeners—the service may end up looking a lot more like its past.