Under the Hood: How Revelator Uses Data to Accelerate the Streaming Economy

Under the Hood: How Revelator Uses Real-Time Data to Accelerate the Streaming Economy

This article has been made possible (and free for all of our readers) with the support of Revelator, blockchain-based digital asset management, data analytics and distribution platform.

Hi, and welcome to the first article in our irregular “Under the Hood” series, dedicated to breaking down the data, tech, and vision behind some of the music industry’s most innovative data solutions. Today, we’ll take a deep dive into the Revelator’s Original Works platform and, specifically, the RPS (revenue per stream) estimation model that allows Revelator to provide artists with daily advances against their future streaming cashflows. The RPS estimations essentially enable Revelator to bypass the established music industry processes for royalty allocation and pay the artists up to 6 months ahead of other distributors — yet, it is just a first step in the company’s plan to reform the streaming economy.

The Challenges of the Streaming Economy

According to the IFPI Global Music Report, in 2020, streaming made up over 62% of total global recording revenues. With the overall streaming revenue up 19,9%, it has also become the main driving power behind the recording industry’s 7,4% growth observed last year — and even though the development of streaming is expected to slow down as we move into the 2020s, it is likely to remain the primary source of growth for the recording business for the years to come. 

By 2021, the DSPs’ role as the primary mode of music consumption is pretty much set in stone. Even as alternative monetization models are gaining traction with the industry, it’s safe to assume that throughout the 2020s, streaming will remain by far the most popular way for consumers to access (and discover) music. Yet, as the streaming economy takes center stage, the question of fair artist (and songwriter) remuneration within the system remains largely unanswered.

The main challenge of the streaming economy is pretty straightforward: within the current system, most middle-class artists out there can’t extract enough value from streaming alone to allow them to build a sustainable career. That fact became blatantly obvious following the COVID crisis putting the live industry to a halt — with no live shows to rely on, most artists had to focus on streaming as their primary source of income, propelling the discussion of streaming payouts to the front and center of the music business, and giving birth to the #BrokenRecord and #FixStreaming movements.

The streaming economy issue won’t be solved in one day — in fact, it might not be solved at all, at least not in the sense of raising the average payouts. Sure, streaming platforms are now increasingly experimenting with new ways of royalty allocation: while falling short of actually reforming their payout system, Deezer has been experimenting with user-centric payouts since early 2017; Soundcloud followed through with the user-centric model earlier this year, branding it as fan-powered royalties; and, just recently, Tidal unveiled its new hybrid payout system, where on top of the usual pro-rata calculations, up to 10% of the subscription fee would go directly to the subscriber’s most-streamed artist. 

However, there’s an argument to be made that emerging payout models won’t change much in terms of overall value extracted from the music catalog. Sure, they might be marginally better for niche artists with loyal audiences, but alternative calculation models don’t address the central problem of the streaming economy — the fact that virtually all music in existence is still priced at just $9,99 per month. And the chances are, this cost/value proposition is here to stay.

Yet, that doesn’t mean that nothing can be done to build a more sustainable streaming industry. The truth is that streaming platforms are just the first step of a very complex value chain running from the consumer to the artist — and there’s still a lot of room for improvement throughout that system that could allow artists to have more control over their rights and, ultimately, extract more value from streaming consumption.

Which brings me to the actual topic of today’s article: earlier this month, we at Music Tomorrow had a chance to take a look under the hood of one of the innovative products looking to do just that — Original Works, which is somewhat of a spin-off to the Revelator’s core distribution, asset management and data analytics offering. Original Works is a product developed with an ambition to create a financial layer for the music industry that would open up a new chapter in music-related investments. This product vision is supported by the three primary pillars of the Original Works:

  • Providing quicker access to streaming cashflows 
  • Improving the velocity and liquidity of music assets 
  • Offering automated estimations of catalog/asset valuation and communicating it transparently — both to the copyright holders and potential investors

However, let’s not get ahead of ourselves. The first step towards that vision was introducing the royalty advance system, which currently allows Revelator to make daily advance payments to the right holders on the platform using real-time streaming consumption data. With that system in place, Revelator can process streaming royalties — that can typically take up to three months to complete their journey from DSPs and into the artist’s pocket — in near-real-time. Today, we’ll break down how Revelator achieved that and look at some of the most significant tech & data challenges the team faced along the way.

How are Streaming Royalties Paid Out to Artists?

However, before we get into it, we need to take stock of the royalty distribution processes currently employed by most streaming platforms. Commonly, DSPs report official sales (i.e., royalties due) on a monthly or trimestrial basis in what is known as Digital Sales Reports (DSRs in the DDEX format). Once these official statements are released, distributors in charge of the catalog can invoice the streaming service based on the licensing deal in place. These official sales reports are usually issued within a month from the end of the reported period — which means that it can take up to two months for the distributors to get the data they need to request the payment from the streaming service. 
Furthermore, once the distributor invoices the DSP for the royalties due, the service has up to 30 days to foot the bill — so the whole invoicing process can also take quite a bit of time. In practice, the current standard for streaming royalty distribution means that the process of streaming royalty distribution would often take three months — stretching up to over half a year in some cases. The royalty processing delays on the recording side of the industry are pretty mild compared to the generally accepted revenue allocation timeframes in the publishing business, where royalties typically take more than a year to process (if you’re interested to learn more challenges of the publishing royalties processing, check out Julie’s guest post on Water & Music). Yet, these three to six months of delay still matter — especially since artists rely on these cash flows to not only pay their bills but also support further career opportunities.

Master royalty flow in the streaming industry
Master royalty flow in the streaming industry

At the same time, in parallel to the royalty allocation system outlined above, most DSPs also provide daily consumption reports to the distributors through what is commonly known as “daily sales” API. These daily reports are used to source the distributor’s analytics dashboards — yet, they are not bindable to pay out the actual royalties. DSPs need to account for things like churned and new subscribers, promotional offers, and other adjustments to the overall royalty pool to calculate the royalties due under the pro-rata model — and so the precise value of the consumption reported through daily API cannot be established before the end of the reporting period. 

How are Streaming Payouts Calculated?

Now, the pro-rata royalty attribution model has been covered countless times, and there’s a ton of materials available online (including this article by yours truly) that can give you a detailed overview of how the streaming royalties are calculated. However, to make sure that everyone is on the same page, let’s take a quick detour and run through the basics of the pro-rata royalty calculation model. 

At the first step of the calculation, the streaming service would divide its user base into groups based on the factors like the user’s current subscription plan and country. This step is necessary as different users within the same streaming ecosystem contribute widely different sums to the system. For example, while Spotify’s premium users in the UK would pay over $13/month for their subscription, their counterparts in India (a country with one of the cheapest Spotify premium plans) would generate just $1,7/month. So naturally, the streams generated by these users have to be considered separately. 

In the current landscape of the streaming market, with its near-global presence and countless subscription options, from family to individual plans to ad-supported and bundled offers, streaming services have to account for hundreds of different revenue pools, one for each of the possible country/plan combinations. Then, for each group defined, the DSP would calculate the royalty pool by applying a global payout rate (which is a subject of negotiation between DSPs and major labels and Merlin, representing the most independent catalog) to the overall revenue generated by a specific user group. These figures are rarely publicly disclosed, but an educated guess would put it at around 70% of the streaming service’s revenue (further split 13/57% between publishing/master sides of the copyright). The resulting figure is the total sum of the copyright holders for the streaming consumption within a specific user group.

To divide that pie between the artists on the platform, DSPs would also calculate the “share of content” for each asset — a share of streams attributed to a specific song within all the streams within a particular revenue/content pool. To make it a bit more clearer, let’s use an example. 

Imagine that a streaming service X makes $100,000 every month — for simplicity’s sake, let’s also imagine it only operates on a single market, with a single subscription plan. Now, if there was a total of 10 million streams within a given month, and your music accounted for 100,000 of these, assuming the global payout rate on the master side at 57%, your total revenue would make $100,000 * 57% * 100,000/10,000,000 (your “share of content”) = $570. Your songwriter would get another $130 — although it would be paid out through a different value chain on the publishing side of the industry.

Now, with that out of the way, let’s dig down into Original Works and showcase how exactly Revelator could provide daily advances against streaming royalties via its Artist Wallet, essentially beating the established music industry processes by up to 6 months.

How Revelator Uses Real-time Streaming Data to Create a Better Way for Processing Streaming Royalties

Like any other digital music distributor out there, Revelator had access to two distinct sets of data provided by DSPs: the “daily sales” API-sourced consumption data, and historical official sales reports, characterizing both consumption and the consequent payouts. On the surface, the wealth of historical revenue data should allow Revelator to quickly calculate a sort of “average pay per stream” rate. Then, it’s a simple question of applying that rate to the daily streams figures provided through consumption APIs — easy as one-two-three, right?

Well, not really. You see, despite the music industry’s obsession with an “average payout per stream”, there’s no such thing as an “average stream” in today’s world. As we’ve outlined above, under the pro-rata model, the streaming revenue is aggregated across dozens or even hundreds of independent revenue/content pools — and even within these user groups, there wouldn’t be a stable per-stream rate, as the factors ultimately dictating these average payouts remain in constant flux. For example, imagine if Spotify unveils a new retention feature that raises the average time users spend on the platform (and average streams generated by a single user). If the total subscription revenue stays the same, such a feature will diminish the average steam value — and that’s just one of the countless phenomena that could influence the per-stream payout rates. 

So, despite what readily available “royalty calculators” would have you believe, providing estimates for streaming revenue based on consumption data is not as easy as multiplying the number of streams by a constant. Instead, it’s a massive big data task that requires a holistic approach supported by an infrastructure that would allow you to process enormous amounts of data. 

Thankfully, when setting out on this journey, Revelator already had the last component in place. Having already built a comprehensive data platform for digital distributors & independent labels as the company’s core offering, Revelator already had an automated data pipeline unifying all the data generated across various DSPs into a single, centralized data warehouse — allowing for BI access to streaming data in complete detail. Moreover, with its infrastructure stretching far beyond aggregated data, typical for most distributors focused on internal dashboarding, Revelator was uniquely positioned to build its daily streaming royalty processing model.

With the data infrastructure already dealt with, it was time to get the data scientists on board to experiment with the RPS (revenue per stream) estimation models. The question was simple: say we receive an ad-supported stream from Spotify in Spain through the daily sales API. Looking at the historical streaming revenue, how much will that stream be worth on the official sales report we’ll get in a month from now?

Without getting into too much detail into the particularities of the models used — I mean, some things do have to remain private — Revelator’s team shortly arrived at an intriguing conclusion. They found that there was no single data model that would outperform all the rest across the board. Instead, as it often is when dealing with predictions in complex contexts, different models were better suited for different estimation environments. For instance, some models showed better accuracy when working with a more limited set of data (when estimating RPS for a local market lacking historical revenue data, for example); others performed best for estimating RPS for a specific income type, specific market, or a specific streaming platform. Furthermore, the team shortly found out that revenue per stream is never stable, fluctuating across markets, DSPs and time.

So, instead of using a stable, unified estimation model, Revelator’s team opted for a more dynamic and agile approach, with multiple models simultaneously applied to create independent RPS estimations. Those estimates are rated based on historical context-specific performance data and their reliability — with an algorithm in place to automatically select the model best suited for the task at hand. Such an approach allowed Revelator to both provide accurate estimations and create a self-improving system that evolves with each prediction made as more and more performance data is fed into the system.

Original Works RPS estimation model
Original Works RPS estimation model

The latter reliability metric is designed to not only assess the confidence of the final prediction but also determine the maximum daily advanceable amount provided to the artist as a share of the final revenue estimation. Simply put, when the RPS algorithm is confident of the estimate made, the artist can get up to 80% of the total estimated revenue as their daily advance. On the other hand, when the estimation is made in a more challenging environment (i.e., when estimating a value of a stream on an emerging market/platform based on a limited context-specific dataset), the reliability metric would be significantly lower — which means the maximum share of advance would also go down to account for the potential gaps in the dataset and the consequent risks.

However, the reliability metric is not the only mechanism to deal with the “cold start” problem. Certain connections are made within the dataset to provide more input into the system. For example, when working with limited local market data, like trying to estimate the value of a stream on Spotify in Finland, the RPS model would also consider the data on the adjacent Baltic markets like Estonia, Lithuania, and Latvia under the assumption that the revenue per stream is correlated across the region. However, whenever such extrapolations are employed, the reliability metrics are affected accordingly.

By introducing the daily advances through Original Works, Revelator has essentially gone into the micro-loan business — and so naturally, fraud prevention had to be a point of consideration for the team. Fraudulent streaming activity is, unfortunately, not such a rare occasion on the current music market. To restrict the potential ways to abuse the system, the team also introduced additional buffers for the maximum advanceable amount for some of the “unproven” artists on the platform. In addition, when calculating the final advanceable sum, the algorithms would also consider the artist’s previous releases to verify if the artist already had a release with enough historical data within the Revelator’s system — or an extensive catalog on streaming platforms. That way, the company can ensure that the streams reported through daily sales API are likely to hold up to scrutiny and end up on the official sales reports at the end of the month.

With all the components outlined above in place, Revelator’s RPS system is able to provide accurate estimates for revenue per stream across multiple platforms, numerous income types, and over hundreds of local markets. These estimates are updated every day and applied to the daily consumption data reported by the DSPs to offer artists on the platform with near real-time streaming advances — a first step towards turning master copyrights into a transparent, highly liquid financial asset.

So, what’s the Revelator’s vision for the future of Original Works? Well, first and foremost, the self-improving RPS system is set to become more and more precise as more artists join the daily advance program, feeding the system with data on historical accuracy of estimations and filling some of the gaps present in the present datasets. However, the RPS system we’ve focused on today is just a single component in a more complex system in the making. The potential here is developing a comprehensive fintech layer for music that would allow independent artists and their teams to treat their copyrights as an investment asset rather than a source of a passive income.

Original Works today presents a solid foundation for that vision. Imagine adding a predictive analytics layer, leveraging historical consumption data to predict the song’s trajectory in the next 3/6/12 months, and marrying that data with the output of the RPS system to automatically provide estimations for the value of the given song. In today’s recording market, such financial insights are only available for the major catalog by virtue of the major label’s extensive analytics departments. Automating that valuation process and putting the insights in the hands of the industry could open up a massive host of opportunities for investment in the independent catalog — and ultimately, allow artists across the board to leverage their streaming copyrights to secure further funding, grow their careers and find a new way to monetize their music within the streaming economy.