Music Publishing and the missing £500 million
Recent research by The Ivors Academy suggested the data-streaming gap to be worth £500 million annually. That’s a LOT of money slipping down the back of the industry’s sofa- tangibly pinching the pockets of creators.
I’ve spent the last 2 years building a platform to tackle this at scale- drawing creatives into a framework they are reliant upon yet understandably struggle to integrate with. Creatr Club finds bad data which causes missing money for creators.
Whose problem is this? If we don’t know whose problem it is, where do we go to address it? I’m 18 months in- having spent 100s of hours with creatives and industry colleagues, in 1-on-1s, in-depth qualitative conversations and aggregating research with 1,000s of participants. The net result? It’s complicated (as ever with the music industry).
Will music publishing ever be able to pull £500m out of a hat and give it to the creatives who earned it? And whose hat should we be looking in to find this ‘missing’ value?
Heads up, you’ll need to know the following:
- DSP = Digital Service Provider (digital stores for music- e.g. Spotify, Apple Music, YouTube, Amazon, Deezer)
- CMO = Collective Management Organisation (handles blanket licensing on behalf of rightsholders- e.g. PRS, ASCAP, SACEM, BMI). Synonymous with ‘societies’
- Rightsholders = owners/controllers of song or recording copyrights. Synonymous with publishers/labels, but can mean unpublished writers
The problem(s)
I’ve written about the metadata problem previously- poor global data makes it a struggle to get songwriters paid. But why? Amongst the other issues relating to poor pay through to songwriters (the recordings-song split, the need for intermediaries, etc.) revenue issues purely arising from metadata generally fall into 3 camps:
- Lack of ISWC-ISRC matches
- Lack of identified creators (typically via their IPIs)
- Duplicate/conflicting data registrations- split/title/formatting/name discrepancies
1) ISWC-ISRC matches
When a recording is streamed, the underlying song copyright is also ‘streamed’ (debatable whether that is ‘performed’ or ‘sold’…). The recording generates money but the song also generates money- both of which need to be licensed and paid for. The information on who owns the recording is readily available, but for various reasons the underlying song ownership is difficult to discern. With no link between them (that is, between their identifying codes- ISRC and ISWC), the recording can earn royalties while the song earns nothing for the same usage. I stream a track- the recording owner gets paid, but whoever wrote the song often doesn’t get paid. In a digital age, this is absurd. It’s also curiously legal- courtesy of blanket licensing. Lack of metadata standards, industry interoperability and no global repertoire database mean these perfectly legal uploads aren’t being paid for- their creators simply don’t get revenue despite their work being used.
2) Identifying creators
This one is obvious- if creators are not correctly assigned to a song registration, they cannot be paid when royalties are allocated to that registration. The nature of music creation is forever becoming more collaborative- more writers writing with more writers. The complexities of international revenue collection also necessitates more publishers and sub-publishing chains than ever. These are the 2 facets of ‘rights fragmentation’- more and more parties needing to know, agree upon, and crucially register the same set of data. If a registering entity (publisher or songwriter) doesn’t know the full names and IPI numbers of all the creators involved then there will exist song registrations missing individually identified creators. At some point there will be allocatable revenue that cannot be paid because a writer is simply not known.
3) Duplicate and conflicting registrations
The same rights fragmentation causes duplicate work registrations- often with discrepant data between them. My favourite example of this is a song credited equally 3 ways- 33.33% each. Publishers rush to capture the outstanding 0.01% by registering their writer as controlling 33.34% rather than 33.33%. Often all of the 3 publishers do this- the song is registered with 100.02%, is over-claimed, and hence cannot be paid out to¹. Registering parties also often introduce spelling and grammar errors in names or titles. This makes matching and merging an impossible task further down the administrative chain, where there is little contact with the original (and knowledgeable) parties.
Does this £500m actually exist? Some CMOs have argued that black-boxes are a misnomer- that using analogies (claiming or paying out pro-rata by market-share) is an unavoidable fact of collective licensing. Moving to perfectly accurate (let alone efficient) reporting and payment for each and every music usage is perhaps an unattainable dream, but there are significant gains to be made in trying. Yes, honest CMOs tend not to ‘hoard’ creator earnings for their own benefit (though some certainly do), but paying out anything on a market share calculation is synonymous with what a ‘black box’ is.
£500m in its entirety is unlikely to be sat stationary in the industry’s accounts. It is also perhaps in itself an inflated figure, and one that some CMOs naturally find offensive considering their accuracy efforts. Other CMOs are conspicuously quiet on the issue. The figure itself, however large, is still an undeniable signpost to the damage that poor data inflicts on genuine attempts to pay accurately. Damage tangibly felt by creators.
In headline terms, ‘The industry’ doesn’t know who to pay for many song uses. It’s not a technical problem, but a political one- despite being about data. Much of the nuance is lost in translation when debates on the economics of music streaming rage. The industry doesn’t integrate creatives well (though it needs to, because it needs them), and creatives can’t understand why they aren’t appropriately paid. Let’s peel apart the value chain to see how the issue is viewed and tackled at each stage.
Are DSPs to blame?
Spotify et al. don’t do themselves any favours with how little they are perceived to pay music creators. There is nuance and layers of complexity to the value delivered by DSPs, but ‘little’ is a relative term. ‘Little’ in comparison to the historic value of songs, the comparative value of songs vs. recordings and what is considered ‘fair’ (which brings its own relativity and nuance).
DSPs like Apple Music, YouTube, Spotify, etc. only deal in recording information, not song information. It would help songwriters and publishers if song metadata was required at the point of upload, so why not? Partly there is no incentive for DSPs- they report specifically on recording usage and rely on CMOs (who they pass songwriting revenue on to) to work out which song is represented by which recording. They have no direct reason to increase the complexity of their systems despite it clearly benefiting creators. This highlights a classic misconception- creators are the supplier and not the customer for a DSP. Effectively they get away with the minimum amount of processing possible- understandable with shareholder value to prioritise.
One reason DSPs tend not to get involved is that music distribution in a digital environment is rapid. The collation (and even allocation) of the required song metadata is not. Song information often simply isn’t available by the time a release is ready. ISRCs (the required unique international standard recording code) are decentralised, effectively being self-applied. The equivalent song codes (ISWCs) are the opposite. There is a stringent and centralised process for their allocation- including that each contributing composer is a PRO member. ISWC allocation thus lags behind the ISRC allocation, and hence the release of music- let alone because gathering the necessary composer IPIs is a struggle. Music is out in the wild, is being listened to, but with no attached song codes to help identify it.
2 clear evolutions to the system would be faster allocation of ISWCs (work already underway by CISAC), and a requirement for DSPs to ingest song data at the point of upload. The latter was recently suggested by the MMF, and is being investigated by the IPO working group on metadata². This would also clearly drive the first- DSPs requiring ISWCs would force the publishing world towards faster allocations. The shifting sands of song splits make things trickier, compounded by the ever increasing collaborative nature of songwriting. These make collating song metadata a challenge- one that is the remit of CMOs.
DSPs pay CMOs- so it’s their problem?
So CMOs get recording data from DSPs, which they use to calculate song revenue against. Logically CMOs should then have some sort of global database noting which recordings represent which songs. They have one. Each. Sort of. And their view certainly isn’t global (or considered authoritative).
The framework of nationally based licensing societies (CMOs) is a function of history- at odds with our modern, digitally interconnected world. Compositions travelled little across borders historically, so energies could sensibly be focussed on licensing the local market- relying on reciprocal agreements with foreign societies to cover what happened abroad. The model hasn’t evolved much, though the world in which it operates has- dramatically. Global consumption through the proliferation of the internet is causing significant friction against this national licensing framework. Copyright law is local, yet music is truly global. The music publishing industry (and the songwriters reliant upon it) is still slave to the interconnectivity of 350+ local CMOs- many of whom are using outdated tech and manual processes. These CMO compatriots struggle to share data- for example ISRC-ISWC matches aren’t communally ‘pooled’ (despite attempts at projects pertaining to this). Rightsholders therefore must maintain accurate data in each individual CMO database to ensure accurate global allocation of song royalties- no mean feat.
There is general consensus that everyone in the value chain would benefit from a single, global reference point for song metadata- CMOs have even previously tried unsuccessfully to create one. Alongside the fatal barriers of “who will pay for this” and “who will control the data”, a centralised dataset represents an existential crisis for societies. CMO’s mandate is blanket licensing- which can only be built upon a foundation of copyright data. If that data is out of their control, could the core activity of licensing also slip through their fingers? Major publishers believe so- having long seen CMOs as a nuisance which prevent their copyrights being licensed directly for digital. Removing CMOs would remove the ‘checks and balances’ they provide- e.g. passing 50% of performance revenue directly to songwriters. In theory writers can compare their CMO and publisher statements to check publisher deductions/pay through, though few songwriters do this in practice. CMOs also can’t be wholesale written off as they typically licence local markets extremely well. They’ve built relationship with every pub, club, café, shop, gig venue etc. who use music- providing a scalable way for most music licensees and licensors to interact. As we’ve seen however, this relies on a firm and accurate copyright data foundation. Centralising this data effort would produce a cost saving for all (which really equates to more money going into the pockets of songwriters), if it were politically feasible.
One evolution in this regard is the emergence of ‘hubs’- like Mint and ICE (the joint venture between PRS, STIM and GEMA). Joining back offices (and copyright databases) of PROs is a good step towards building efficiency for the industry- by firstly aligning datasets and secondly reducing duplication of work (particularly song-recording matching). ICE Cube has long been touted as the upcoming gold standard in compiling and reconciling an accurate industry data set. However the much delayed project is yet to be realised despite its promise of reforming innovation- though it has recently become a priority for PRS’s ‘big data’ centric CEO. An unintended consequence of back-office consolidation is that publishers are getting even less line-of-sight to the databases whose copyright picture is used to allocate their royalties. This is more acute for songwriters, who have little transparency to the industry as it is despite creating the very IP that everyone is monetising. Integrating publishers and writers into the matching process would alleviate these concerns, reduce discrepancies and improve efficiency via transparency.
It is a challenge for societies to invest in long term solutions like ICE Cube- increased accuracy comes with a large medium-term cost. This, as ever, effectively comes out of the pockets of songwriters and their rightsholders. Are publishers willing to forgo stronger profits in the short term to build accuracy in the longer term? It is already a struggle to ‘keep up’ with the trillions of lines of data that need processing now, let alone creating a more accurate system for the future. This is one of the reasons black boxes (or market share distributions) have remained. Improved data infrastructure is a 4–5 year investment for societies at a minimum, which is at odds with publishing companies looking to maximise their quarterly profits. Increasing accuracy can only redistribute the existing pot of money, and not increase how much is fed into the top of the funnel. Those with the greatest resource to improve accuracy (majors) are also disincentivized from doing so because they take the largest part of a pro-rata distribution. The entire industry says it’s keen for improved accuracy of royalty allocation- but misaligned incentives have made it tricky to realise beyond lip service.
CMO and hub’s datasets are built upon the information they are provided- song data from publishers/songwriters and recording information from… well… publishers/songwriters. Collaboration between the recordings and songs side of the industry would make a huge difference to this process. As with DSPs there is little incentive for record labels to improve the scope of label-copy- the release data sheet that could include song information as well as the current recording information. If recording and song data was correlated before any system ingestion it would save huge ‘after-the-fact’ matching expense further down the line- as Mark Douglas (CIO, PPL) pointed out in his MBW Column (issue 17, pg. 86) calling for data management at source- a data supply chain.
Matching songs-recordings (ISWCs to ISRCs) currently should be a top priority for CMOs- and something that should be publicly shared by each society for the benefit of all. An unenviable yet major task of publishing administrators is tracking down ISRCs to provide to CMOs alongside the associate song information- and preferably pre-release. Publishers have a direct financial interest in getting accurate metadata to CMOs in a timely manner, but there are hurdles in the process for them too.
So CMOs get their data from publishers…?
A pre-existing ISRC-ISWC match with a clean data registration represents a tangible cost saving for societies. This is really more money into the pockets of songwriters- both directly because of the match and in the lowered associated processing cost. CMOs rely on clean data ingestion from publishers- and as we’ve explored, this ingestion happens separately, on a territory-by-territory basis. Hit songs now have an average of 4 songwriters and 6 publishers per song- rights fragmentation by its very definition, and at the earliest phase of data creation, pre-ingestion.
Publishers and the songwriters they represent have the most to gain from clean and efficient data processing. A large part of publishing administrators time is spent chasing data- from record labels (e.g. ISRCs), from writers (for splits and co-writer info) and from other publishers (e.g. IPIs). The more information a publisher can gather, deconflict and then disseminate before release, the better chance they have of being accurately and promptly paid. It’s worth noting that the publishing industry’s current ‘promptly’ is 12+ months, particularly head-scratching compared to the 1–2 months on the recordings side. If songs can begin their ‘registration life’ with a clean bill of data-health they stand a much better chance of picking up royalties quickly and accurately.
Getting recording data ahead of a release is a significant challenge for publishers. An ISRC might not yet have been assigned (often only at upload for DIY artists), or the release date not finalised. As with DSPs, record labels have no incentive to share recording data- it simply represents a cost for them.
Aside from gathering data, another challenge comes when the subsequent collated data leaves the original publisher’s system, and meets with data sent from the co-controllers’ systems. With 6 publishers on average per work, each work registration is submitted 6 times- and often not with corroborating data. Has each controlling party sent the precise same set of data? The same writers, with the same IPIs, the same splits and the same attached recordings? The same spellings and grammar? The issue is exacerbated down the administration chain- as original publishers send data to sub-publishers (who may even send it on again to further sub-sub-publishers). These parties then all register the same work, but when discrepancies become apparent there is little line of sight back to the original parties involved. Issues go unresolved, payments go unpaid. The replication of registration work is a hot-bed for introducing errors. Once the horse has bolted it is tricky to triangulate what the revenue-blocking data entry might be, and where it sits in a foreign database. Errors are often only discovered when expected revenue doesn’t trickle through after 12–18 months!
CMO data originates from publishers- if publishers, sub-publishers and administrators had an easier time of gathering and sharing an accurate and collated data picture it would go a long way to making CMOs processing easier (= cheaper). Of significant benefit would be a platform where publishers (and the full chain of title representatives) could interact on the same set of global song registrations- with line of sight for songwriters and CMOs. Transparency would build both authority and hence efficiency for all- though this can only come with a foundation of trust.
Creators are always those greatest affected by inaccuracy- whatever their revenue collection framework. Both they and CMOs rely on publishers to disseminate accurate copyright data- including what is gathered from record labels. The glaring issue is that a lot of the required data comes from the creators themselves…
Erm so it’s a creator problem……?
There’s little argument that creators suffer from inaccurate data. Line-by-line reporting will always increase the amount of revenue fed through to creators, though this may be at the expense of the parties currently benefiting from inaccuracy (as discussed above). Creators are also the originators of much of the data- what is the song and who contributed to it in what way.
Songwriters trust their publishers to handle data for them. It is one of the most basic tasks of a publisher- administering the copyrights. Songs must be registered accurately for every penny generated in every territory for every use to come back to the writers, noting that publishers only get paid when this revenue is collected. This is of course reliant on gleaning clean data from creators.
An interesting dynamic is that despite creators suffering the most financially, they tend to engage with the problem the least:
- They have an ever growing list of tasks on their plate that crowd out time for their core remit- actually being creative (the latest being the push to become a content creator)
- Their natural creative gifting often doesn’t partner well with the need for an administrative mindset
- Music metadata is complex and overwhelming, compounded by the industry’s opaqueness
- The value ‘unlocked’ isn’t always worth the time input
- Data management is what they pay their publishers for
- Chasing missing revenue is not as rewarding as creating fresh music- income generation is often a proxy for a feeling of success/impact on listeners
In this regard education is not the silver bullet to getting creators to engage in the problem. Is education important? Yes. Should the resources be available for a creator to dig deeper/investigate? Certainly. But our research has shown that even if the data problem and the steps to resolve it are clearly laid out for music makers they still have an aversion to completing the necessary task- even with a clear and present financial incentive. Creators just don’t have the capacity/tendency to engage- and can be emotionally overwhelmed when forced to.
Where does this leave us? Needing a system they can trust- a system that just works.
Look at it this way. When you travel abroad you can happily use your debit card to make purchases, merrily tap-tap-tapping away in shops/cafes/restaurants. You get a banking app notification saying a transaction has occurred, and your available balance has decreased accordingly. Little do you (or the shop vendor) know the transaction takes days to reconcile through the international banking system. Neither you nor the vendor need to know your identifiers, or need to chase the bank/credit card provider/Bank of England to ensure the money flows freely between you both. It just works. Yes there is a cost, but it is negligible- and worth it for the ease of the transaction. Because it works, neither party needs to dig deeper or be educated on the mechanics. The shop can concentrate on selling, and you can concentrate on buying. There is implicit trust- education is therefore not a requirement.
The opposite is true for songwriters. Because the system clearly isn’t working for them they are being forced to learn how publishing administration works, often against their natural inclination. They need to know their identifiers, the identifiers of their peers, ISRCs, IPIs, ISWCs… the list goes on. Education only overwhelms them further and increases their to-do list- publishers and CMOs need songwriters to manage data, yet songwriters rely on their publishers and CMOs to manage data for them. It’s a symbiotic relationship- and one that requires smoother facilitation, and of course, trust. Any tool solely educating music creators or reliant upon them to change their behaviour is likely to fail because it hasn’t understood how they feel.
Creators want to spend time being creative, not managing data.
Stopping £500m going down the drain
Trust is needed. A system that facilitates trust is needed. Transparency and interoperability are a prerequisite for such a system- which requires industry collaboration.
Each tier of the value chain can contribute to eradicating unallocable revenue- though they each feel the pain in different ways. This is one of the stumbling blocks of the past- gathering the necessary parties together around common solutions is a struggle. Creators shouldn’t be burdened with the unenviable task of data management, yet it is what gets them paid. Publishers and CMOs could work together more closely (with the recordings industry in tandem) to invest in infrastructure that future-proofs the ecosystem. This would ensure more accurate returns for all in the long run, but comes at a cost to profits in the shorter term.
Can disparate parties come together to facilitate trust for creators? Or will a third party gather enough creators to hold meaningful conversations and build integrations on their behalf? This is often the case on the demand side, yet the supply side has remained relatively immune as its power has consolidated.
One thing is for certain- creators are missing out, and are unlikely to find £500m under the industry’s mattress without someone they trust helping to lift it for them.
[1] Some societies do have rules which allow payments to be made if the over/under-claim is not precisely 100.00%, yet very close (e.g. 100.02% claimed).
[2] There is currently a metadata working group convened by the IPO in the fallout of the DMCS inquiry into the economics of streaming. Many of the problems mentioned here are actively being discussed, with the direction of travel moving towards better data sharing and earlier allocation and capture of key identifiers (inc. ISWCs). DSPs could accept song information alongside a recording upload- a significant win for re-aligning incentives of record labels with that of their publishing compatriots.
