It sounds like the system is working as intended for the record companies. And also like this might be a situation where a class action could actually be useful.
Clearly there needs to be a single legal, authoritative, and publicly accessible source for any music product's (public) metadata. Before it gets aired by anyone. A good model for this might be DOI ('Digital object identifier'). Anyone- music companies, artists, stores and streamers - would get a number for each published title where they could then find all the related metadata. Including reliable sources that can be queried for (private) legal details.
It would seem like the rights orgs like ASCAP, BMI, Harry Fox, etc etc should have been all over this long ago. But maybe it'd mean their extinction. Or it would would complicate the back-room deals.
There sort of is. But at the same time there absolutely is not.
There have been various attempts to do this and they’ve all fallen flat because of vested interests and because the music industry is pretty bad at technology.
There’s also a school of thought that says that the reason it’s never worked is because it would shine an uncomfortable light on some things - for example publishers (accidentally or knowingly) collecting revenues on song copyrights they no longer represent, or CMOs distributing money badly.
I’m not, no. Musicbrainz is amazing and I’ve known them for a long time but it doesn’t address this problem - which is to do with authoritative first party metadata that define rights ownership.
What I’m referring to is the structural identifiers like ISRC and ISWC.
A big problem with music metadata is that there are lots of silos in the music industry: the big split is between people who control and administer sound recording copyrights and people who control and administer song copyrights or publishing rights.
Another commenter below talks about the need for identifiers.
There are currently two primary identifiers. ISRC on the recording side and ISWC on the song side. There are actually a whole host of other identifiers such as IPI (the “interested party identifier”).
Stakeholders generally have access to the information relating to these identifiers - and indeed there is a certain amount of public info for these as well.
Here’s the problem though.
The ISRC is the identifier that relates to a particular sound recording - a track - and the ISWC is the identifier that relates to the song itself - the music and words which are encompasses within the recording.
One of the biggest issues just now is that not all DSPs (by which I mean streaming, download and other consumption platforms - such as TikTok, YouTube, Meta, Peloton) require an ISWC on ingest of a sound recording. So you can have a recording being used and generating revenue where there is no corresponding data about who the underlying songwriters are - meaning those songwriters are not able to get an accurate picture of the exploration of their copyright.
One of the other issues - which is part of the reason for this - is the length of time it takes to register the “publishing” in the song and get an ISWC issued. When you distribute a song to digital platforms your distributor or label will issue the ISRC. When you write a song to get an ISWC you need to register it with the issuing body for ISWC via a participating publisher - who upon acceptance will then issue you the ISWC. This can take weeks or months.
One of the other issues though is about capturing metadata.
There are a whole host of rights that need to be captured in the metadata to ensure the right people get paid at the right time for usage of their rights.
This can be the writers - or co-writers - of a song but it can also be the performances of musicians captured in a sound recording. Writers, producers, performing artists, session musicians, record labels, publishers - these all need to be linked via metadata.
There are also whole load of ways that revenue can be generated for usage and at least the same number of ways that those revenues flow through the system.
For example:
Three songwriters create a song. One is based in the USA. One in Germany and one in the UK. They use a significant element from another song that was written by two French songwriters.
This song is recorded by a UK based artist signed to a US based record label and produced by a Spanish producer. A sample is included from a song written and recorded in the 1960s and as part of the agreement for use of the sample the writers of the original song get a writer credit - and a few points of the songwriting royalties - and the holder of the copyright in the recording (a long defunct label whose copyrights have been acquired by an investor) also get a split on the recording. It is released to streaming platforms and other DSPs. It’s a global success.
It is licensed for use on TikTok as a TikTok Sound and to Meta for the Instagram Sticker.
Synchronisation deals are done to use the music in a film - released in cinemas - as well as a terrestrial broadcast show on the BBC in the UK and a Netflix produced show.
It’s performed at festivals - and covered by other artists in live performances.
It’s played on radio in various territories including the UK, Belgium and the USA.
In this scenario there are something like 30 different types of revenue stream that I can think of off the top of my head, and because of the complexity of the rights landscape for some of the performers and songwriters some of those revenues could take a year or more to reach the relevant stakeholders and be accounted.
Music revenues are INSANELY complicated. Good metadata is essential. But capturing, verifying, ingesting, tracking usage and properly predicting, quantifying and monitoring the resulting revenue streams is a mixture of accountancy, art, guesswork and witchcraft.
The way revenue moves from DSPs to rightholders is Byzantine in its complexity.
For example:
Sound recordings revenues in streaming are fairly straight forward. The DSP pays out to the sound recording copyright owner. But on the publishing side there are two different revenue streams each time this track is played on a DSP: one for “mechanicals” and one for “performance”. In our example above the way these royalties reach the songwriters will depend on all sorts of things - including which CMO (“Collective Management Organisation”) they are members of - which will normally differ based on where the writer is resident - but also whether their publisher (the business partner of a songwriter - sort of analogous to a record label for a performer) has a direct licence with the steaming platform in question.
A stream of the recording of the song in Brazil would often result in the streaming platform paying out the publishing royalties to the local Brazilian CMO which then sends its correspondent societies in other territories data that allows them to claim the royalties they are owed. It’s a little bit like peering between ISPs but far slower and far less efficient.
When a DSP like Spotify can’t attribute the underlying song in a recording to a writer or writers it just pays that money out to the publisher CMO in the territory in which the stream took place. Unless another society representing that writer claims the money it’s eventually just distributed to the revenue pool for local writer members of that society.
The key thing is this: if you stream a piece of music the way the money reaches the songwriters, recording artists, other performers on the track, and business partners like labels and publishers is far from straight forward. A single stream does not result in a defined amount of money - because streaming revenues are determined by the amount of money generated from users in a particular revenue window, and then that total revenue pool is divvied up based on “market share” of the music consumed. This means that for a recording artist 1MM streams of a particular track in June can generate an entirely different amount of money compared to 1MM streams of exactly the same track in May.
For labels and recording artists this is one thing. For songwriters 1MM streams in May and 1MM streams in June can be an even bigger difference - because (for the sake of argument) if the 1MM streams all took place in the songwriter’s home territory in May but all the streams were in a different territory in June then even if everything else was the same their revenue from those two periods would be entirely different - and the foreign revenue could take a year or more to reach them.
Even with good metadata there are all sorts of ways things can go wrong.
Thank you for the detailed insights. I am a mastering engineer and I wish to provide my clients with as much revenue potential as possible. At the moment I fill in RIFF INFO, ID3 and EBU 3325 aXML chunk with as much information as I can get from my clients (which is sometimes very little...)
Really curious about your solution, can you please share more?
Hey saaaaaam . I'd love to pick your brains a bit on this. I'm working on a web3 protocol ("Legato") relating to many of these issues. Any way we can connect? (I can't find a way to DM you on HN). roy at osherove .dot [com]
I've been contributing to Musicbrainz for years and I highly recommend it. I use the "Beets" client to automatically tag my library (and occasionally "Picard" to manually clean up mismatches) and it works really well.
Musicbrainz is amazing but the problem is that musicbrainz is about attaching metadata after a track and underlying song has been released.
The structural problem is about capturing and attaching metadata before that recording of a song is ever heard, and ensuring the metadata flows through the digital value chain, and then out with the revenues trust result, through the royalty chains.
On top of this I think there’s classical music, which none of then large streaming services get right, hence Primephonic and IDAGIO. Apple bought Primephonic so I wonder if they they intend to release an improved metatdata model with Apple Music.
See, this whole metadata problem being used as a means to credit somebody’s royalties on a platform that wants to appear both on moral high ground and judicious to the artists and public is exactly the kind of problem that a blockchain would solve, technically speaking.
Isn't this what the legal system is for? If the situation were reversed, I can guarantee they'd sue him for the difference.