Pariah Carey Part Three: Mysterious Marketing and Trying to Game the Algo

Catch up on the Pariah Carey series: Part One and Part Two

Across four parts, Pariah Carey – whose etymology you can find here – covers the recent music marketing data bootcamp co-hosted by Water & Music Academy and Music Tomorrow. The month-long series featured eight nutrient-dense sessions on music data, filled with case studies and actionable frameworks to guide folks through the terrifying gauntlet of the digital marketing lifecycle.

Rather than simply recap the bootcamp, though, I’m putting those sessions’ wisdom to the test, and offering up my own nascent music journey as a sacrificial pig. The hope is that my experience can be useful for others looking to take their first steps.

In part one, I reviewed data practices and benchmark setting. Then I built a foundation, creating my artist story and developing ‘SMART’ (Specific, Measurable, Achievable, Relevant and Time-bound) goals (I’m tracking them here) for sharing that story and building community.

In part two, I recapped approaches to developing a marketing plan and managing community relationships. I also explored tools and insights to help structure my content framework (which I’ve made publicly viewable here). Today, I’m sharing insights from sessions five and six, which covered marketing automation and the mysteries of algorithmic recommendation systems.

Understand your story
Set goals
Choose the tools (and content pillars) that help you tell that story
Build a content strategy – across music and social – to achieve those goals
Observe how people respond to your content
Track and adapt

SESSIONS FIVE AND SIX: Marketing automation + Discoverability: Introduction to recommendation systems

Remember, as we begin, that there are 100,000 songs uploaded to streaming platforms every day. Reckoning with the unceasing flood of music – and the consequent marketing efforts required to “cut through the noise” – has been one of the bootcamp’s central themes. Good music is no longer enough, alas, but the “right” data can help an artist rise above.

Recognizing and making actionable the “right” data, though, is tough. That’s where marketing automation tools can help – the right data can “reduce noise and uncertainty” and save artists time, said presenter Alex Brees, a former trading analyst who now applies his acumen to guiding artist journeys.

Brees was motivated to transition after watching independent artist friends who – despite their talent – struggled to succeed. He founded un:hurd, an app that “turns your data into a series of actions that allow you to complete a data-driven marketing campaign in just a few clicks.” “For us,” he said, “it’s giving artists back that time [to make music] – we take care of the marketing side.”

The un:hurd app pulls data from an artist’s streaming and social profiles, creating “competitive analyses” with artists who have comparable personas. Those insights, coupled with demographics data – like where your listeners are located, or the behavior and identity of communities from similar artists – inform how (and to whom) artists promote themselves, and they power customized data-driven marketing campaigns.

Success is still contingent on the content itself, though. “Seventy to 80 percent of the impact of an ad comes down to the quality of the content,” Brees said.

Added fellow presenter Kristin Grant: “The onus is more on the quality of content than ever before. The tools just have to catch up to support and close the gap.”

Grant is the chief executive officer of Wescott Multimedia, which focuses on finding spikes in data – i.e. moments in which an artist sees an uptick in engagement, for whatever reason – and automates campaigns to take advantage of that attention.

Wescott’s product ingests artist-specific data, measures and detects spikes, generates and deploys ads – complete with contextualized in-house ad copy – and then optimizes ad spend accordingly.

The impact of “spending money when people are paying attention,” Grant says, is a 25 percent increase in revenue. “Timing is everything,” she continued. “There’s a 3-5-day window that you have to react to the spike.” Automating the process optimizes the efficacy of that response, and as with un:hurd, enables artists to reap the benefits of their data without sacrificing the quality required to “cut through.”

One reason the bar for success is so high is the irrational – but very real – expectation that artists should make quality music and quality non-music content. As long as the streaming paradigm persists (aka while 99 percent of artists can’t earn a living from streaming), that’s just not possible. So when forced to operate within a system of algorithms that perpetuates the system and said expectations, saving time is – sadly – about the most an artist can ask for.

But should we have to operate within that system? Broadly, that’s a question being asked by many web3 builders. And they’re not alone.

In late 2020, the UK Parliament’s Digital Culture Media & Sport Select Committee announced an inquiry into the economics of music streaming. They explored how recommender systems “could reflect biases that may subsequently reduce new music discovery, homogenize taste and disempower self-releasing artists,” explained David Hesmondhalgh in his presentation, citing the inquiry. “Written evidence submitted by many creators was ‘critical of the opacity of algorithmic curation, and called for greater oversight.’”

Hesmondhalgh – a Professor of Media, Music and Culture, University of Leeds – stressed the words ‘biases’ and ‘opacity,’ which were key focal points of the subsequent reporting done by the Center of Data Ethics and Innovation, the branch commissioned by the inquiry.

In February 2023, the Center shared their findings. Of the musicians that were surveyed:

≈90 percent were concerned about systemic biases that prioritize some artists over others
≈70 percent were concerned about biases across certain demographics
≈85 percent were concerned about biases relative to genre
≈80 percent felt if they were treated unfairly, wouldn’t know who they should contact

Demonstrated here are clear concerns of “popularity bias,” a phenomenon in which the already popular become more popular. And if we look at statistics like 98.6 percent of Spotify artists make $36 per quarter, and 1 percent of artists generate 90 percent of all streams, or studies on gender biases conducted by Liz Pelly and Maria Eriksson and Sofia Johansson, it’s hard to argue that traditional power structures aren’t embedded within our algorithms (we humans are filled with implicit biases, after all, and we make the algorithms).

That said, we have “very little insight into what is happening in actual streaming platform practice” and therefore can’t confirm the existence of popularity bias, Hesmondhalgh explained. “Music recommendation is cloaked behind a veil of confusion.”

But even if we can’t confirm the actual existence of popularity bias, an assumed popularity bias exists, and that’s important. As Hesmondhalgh elaborated, that assumption prompts people to create schemes and tools to trigger the algorithms that ostensibly increase popularity – whether it’s “real” or not.

Today, “gaming” the algorithm is standard practice, but in this contest, only the game creators know all the rules – and they can change them when they no longer work in their favor.

“It still matters that music streaming platforms are driven by profit,” Hesmondhalgh said, “and may therefore have an incentive to keep pushing the work of the most popular artists – or at least not to take active measures to counter that tendency.”

“Systems by which culture is circulated have always been complex,” he added, referencing TV and radio as comparisons. But things are how they are until they’re not. “Might it be possible,” he posed, “to develop public service recommendation systems to counter that profit driven tendency?”

As we await the revolution, let’s turn back to the assumption-based streaming reality we live in today – within which tools like un:hurd and Wescott have been built to give artists a leg up.

These tools in particular, though, need to ingest data that I – as an artist – have not yet created, so for the fresh slate musicians out there (i.e. the folks for whom Pariah Carey is designed), they may still be premature – which is important to recognize. As Brees said, “20 to 30 percent of artists are trying to promote their music when they’re not ready.” They don’t have a Spotify bio, for instance, or their tracks haven’t been produced and mastered.

Confirming that preparatory work is done is a good exercise in the interim. And as we build and test our content strategies, and release music onto streaming services, remember there are tools like these designed to cut through the noise. They’ll be here when we’re ready.

In the next edition of Pariah Carey, we’ll be recapping the final two sessions of the bootcamp, which cover boosting digital discoverability and new digital frontiers.

Until then, here's to you Mariah 🥂

lead image: Kristin Grant