The Life of a Stream

Henry Marsden
12 minutes ago
6 min read

Someone in Hamburg presses play. Three and a half minutes later, the song ends, and somewhere between eighteen months and never, a fraction of a cent arrives in a songwriter's account.

Photo by - Landsmann - — *Photo by* *- Landsmann -*

That gap- between the play and the payment- is where this entire industry actually lives, and it is almost completely invisible, including to plenty of people who work in music. It's also one of the things that gets me out of bed in the morning. As a newly minted publisher a decade ago I was constantly fielding questions from my writers (essentially just my friends- the only people who would actually sign with me back then) about when they'd be paid for the millions of streams they could see on Spotify.

This week let's follow one stream all the way through, and look at every place it can go wrong. Fair warning: this is where the complexity becomes obvious rather than theoretical- which is rather the point.

One Play, Three Chains

A single stream doesn't generate one payment. It generates several, along entirely separate paths.

On the recording side, the service pays whoever delivered the track- the label or distributor- who then pays the artist. That path is comparatively simple, for a multitude of reasons. One of the reasons is recording royalties are ‘pushed’; song royalties have to be ‘pulled’. With songs, the licensing partners must first identify that their work was streamed at all, and then claim what they are owed. Nobody simply hands it to them.

On the song side- the composition- one play triggers two distinct rights: the mechanical right (the reproduction of the work) and the performing right (its communication to the public). A stream is legally considered to do both at once.

But two rights does not mean two payments. The performing right itself splits into the writer's share and the publisher's share, which can travel separately- different entities, different deductions, different speeds. A single stream therefore flows down three distinct royalty chains for publishing alone: the writer's performing share, the publisher's performing share, and the mechanical.

Three chains, three sets of intermediaries and- crucially- three separate matching processes that can succeed or fail independently. Before our stream has travelled anywhere it has already split three ways, and everything after that happens in parallel, through different infrastructure.

Never Break the Chain

Which chains your stream travels down is not fixed. It depends on three things: the writer's society, which determines the class of repertoire the work belongs to; the publisher, which determines the route it takes; and the market where the stream happened, which determines the rates and the split.

That last one is not a rounding error. Of the total song royalty on a UK stream, roughly half is apportioned to the mechanical right and half to the performing right (itself split evenly between the writer's and publisher's shares). The same song streamed in France is apportioned 75% mechanical, 12.5% / 12.5% performing.

It also means the money rarely goes where you would assume. Digital mechanical licensing in Europe has largely detached from the local-society model- DSPs increasingly license repertoire through multi-territorial hubs (ICE, MINT, Universal with SACEM, Sony's SOLAR, AMRA) rather than territory by territory. So our Hamburg stream doesn't simply go to the German society and stop. Why that is, and why a song effectively carries a nationality of its own, is worth an edition to itself- which is next week's. For now, the consequence is what matters.

One play. Three chains. Potentially three entirely different organisations processing the same three minutes of music, on different timelines, using different data.

Not All Matches Are Equal… but some are more equal than others

Which brings us to the point I think is the most fascinating, and most frustrating. A song can be immaculately registered in one place and that registration be worth precisely nothing, because the digital licensing flow never touches it.

Matching is not a single, universal event. There is no global ledger where a recording is connected to a composition once and everyone downstream inherits it. It happens separately, and distinctly, at every hub.

The same recording can be perfectly matched at ICE and completely unmatched at SACEM- linked to the right work in one system and the wrong work in another. Each organisation runs its own process, against its own database, with its own version of your registrations.

That means your catalog does not have a data quality score. It has one per hub- and those hubs are wildly unequal in what they're worth to you.

A pop catalog with heavy US streaming has most of its mechanical exposure sitting in the MLC's matching. A Canadian Francophone singer-songwriter may have far more riding on ICE (if signed to Sony) or SACEM (if signed to Universal)- neither of which may be their own society! The right question is never “is our data clean?” It is: “where does our money actually flow, and how good is our matching there?”

Fixing a thousand registrations at a society handling 2% of your income feels productive, but changes almost nothing. Fixing twenty at the hub processing 40% is transformative. Different effort, wildly different return- and you can only tell which is which if you've mapped your revenue to its actual routes first.

… and Then You Wait

Assuming everything works. Our stream still has to survive the calendar.

The DSP reports usage for the month, typically some weeks after it ends. The hub or society ingests, processes and matches it- adding weeks or months. Distributions run on fixed cycles: often quarterly, sometimes semi-annually, occasionally annually. Then the publisher receives, processes and allocates- or sub-publisher/administrator first for foreign usage.

Stack those and the honest answer for a domestic stream is six months to a year. A UK writer signed to a major, streamed in France, runs twelve to eighteen months. The same writer streamed in Brazil: eighteen months to two years.

Consider what that lag does to everyone's judgement. A catalog buyer reading last year's statements is looking at usage from 2+ years ago. A publisher chasing a matching problem finds it a year after the money went astray. And unmatched income doesn't wait patiently- it is distributed pro-rata, using matched income as the proxy for who receives what. So an unmatched work misses out twice: once on what it earned, and again on its share of everybody else's.

Every handoff is a point where a work can be unmatched, mismatched, partially claimed or held in suspense- and the chains aren't even the same shape. The first link matters most, because that is where a track report full of ISRCs has to become a claim against a work. Get that wrong and nothing downstream can rescue it.

The Uncomfortable Summary

A single stream of a single song involves three royalty chains (four including master rights), multiple hubs, a chain of sub-publishers, and up to two years of latency before a songwriter sees a fraction of a cent.

This is not anyone's incompetence, but is baked in as infrastructure. It's what happens when frameworks designed for sheet music and radio are extended, patched and internationalised for borderless consumption, each layer making sense when it was added.

It does however explain why “just fix the metadata” is such an unsatisfying instruction, and why data work here isn't really about tidiness. It's about knowing which routes your money takes, and making sure your song is correctly identified at each point where it changes hands.

A stream is a single, instantaneous event for the listener. For the song, it's a two-year journey through a dozen institutions, any one of which can lose it. For a sense of how far from solved this is: the UK’s MMF proposes is that songwriters should be paid no later than two months after a stream. Set that against six months to two years.

If you've ever tried to explain this to someone outside the industry- how far did you get before their eyes glazed over? And which junction in that chain do you think leaks the most?