Fix your music data. Find your royalties.

The gap between the play and the payment is where the music industry lives, and it is almost completely invisible, including to plenty of people who work in music. This week let's follow one stream all the way through, and look at every place it can go wrong.

Dancing Queen is literally, T-000.000.001-0. So let's do something anyone reading this can replicate, and take a look at the databases- The MLC's public search, PRS's works search, ASCAP's and BMI's repertoire tools- and look up Dancing Queen to see how the ISWC has fared in its 30+ years of history.

The Leakage Ledger is a recurring edition where I’ll publish actual numbers from actual catalog work. Anonymised (obviously) and stripped of anything identifying- but real. One catalog, the findings, and what they were worth. No thesis to defend, just the pure maths. Tangible numbers- those that you can see, touch and feel, change that conversation. A publisher can't benchmark themselves against a sentiment, but they can absolutely benchmark themselves against a percentage.

Last year I wrote about real issues to be found in the MLC’s dataset- an exploration of what was really sitting inside the industry's most transparent copyright dataset. The headline then: the MLC reported around 90% of royalty value matched, while our analysis showed only 22% of ISRCs had been attributed to any work- some 170 million of 217 million recordings sitting unlinked, with the long tail holding more recoverable money than the industry's hits-first workflows were eve

Every catalog acquisition starts in the same place. A data room opens, and inside sits the familiar furniture: a stack of royalty statements, some contract summaries, perhaps a seller-commissioned earnings report with a decay curve that always seems to flatten out at exactly the right moment (funny that). And at the centre of it all, the schedule of works- the document that defines what is actually for sale.

Most catalog cleanup projects begin in a familiar place- are the right recordings attached to the right works. Are all the ISRCs present? Are they linked to the correct compositions? The difficulty is that recording-work matching relies on a much deeper assumption: that the list of works being matched to is itself clean.

US termination rights have been somewhat of an industry secret in times gone by- and understandably so given their creator friendly purpose. For any not familiar, this is the legal right of a songwriter to reclaim US rights they have previously transferred (say under a publishing deal) after 35 years. These sit quietly inside a catalog for years, sometimes decades, looking like a future technicality rather than a present commercial issue.

If publishers want to use AI meaningfully, they first need to know whether their own data is in a fit state to support it. Publishing data is not a neat database of facts waiting to be queried. It is a living, fragmented, territory-specific web of claims, registrations and human judgement. Anyone who has spent any length of time inside the operational machinery of publishing knows this instinctively- this is the world AI is entering.

AI is forcing every business to ask a difficult question: what work actually needs a human, and why? ‘Work’ is context, judgement, memory (personal and corporate), relationships, accountability, taste, intuition and the ability to discern why something matters in the first place. If those things are stripped out too quickly, the short-term efficiency gain can become a long-term organisational cost.

A significant portion of the last decade in the Music Industry has been spent discussing capital inflows. Billions has poured into catalogs, billions into infrastructure and billions into technology and platforms. This week the latest turn in that saga is being reported- that Sony Music Group is in advanced talks to acquire Recognition for $4Bn. This is particularly illustrative of a wider dynamic emerging across the market- how does the money leave?