The Metadata Iceberg

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? Are those links reflected at the key licensing hubs? This is often the visible layer of the problem, and is also often the layer that produces the clearest commercial urgency.

Photo by Derek Oyen

A recording with no work matched means unclaimed royalties. Likewise a song with missing ISRCs means income is being held, misallocated, or simply bucketed in with pro-rata market share distributions. When catalog owners talk about “cleanup”, this is often what they mean: getting recordings and works connected properly across the ecosystem.

That work matters enormously, and is one of the most direct ways to find missing value inside a catalog. The difficulty is that recording-work matching relies on a much deeper assumption: that the list of works being matched to is itself clean.

That is where the project usually starts to become more complicated (or in my world- more interesting).

What exactly is the ‘work’?

Before you can confidently say “this recording belongs to this work”, you need to know what “this work” actually is.

Is this work the same as that work? Is the slightly different title an alternative title, a typo, a translation, a version, a remix, cue, derivative, or a duplicate registration that is referring to the ‘same’ work? Does it share the same ISWC? Should it share the same ISWC? Are the writers identical, or merely similar? Are the shares consistent? Are the writer roles the same? Has the same composition been registered separately by different parties using partial or conflicting information?

This is the metadata iceberg- with the recording match being only the visible tip. Underneath it sits a larger set of questions about works, contributors, identifiers, titles and registrations that all needs disambiguation.

A catalog can appear messy at the recording layer while the real issue sits at the work layer. Equally, a catalog can appear clean at the work layer until external datasets are reconciled and reveal that the same repertoire exists in multiple slightly different forms across societies and administrators. It's in the differences that revenue leaks away.

The vision for the ISWC (International Standard Work Code) is to solve this. In principle, it is the identifier designed to uniquely represent the musical work- so any party can have confidence they are communicating about the same work. In practice, it only earns its value when the surrounding data is trusted. An ISWC is short hand for identifying a unique combination of title, writers and their roles (in fact, you cannot have the same set of writers in the same roles register a song with the same title). When that same work identifier appears against conflicting titles, writer sets or registrations, it becomes significantly harder to rely on operationally.

Identifiers, including ISWC, are a form of currency. Their value depends on confidence. Once confidence drops, the identifier still exists, but they become effectively worthless.

Don’t forget alternative titles

Alternative titles are often treated as a relatively harmless metadata detail. In reality, they are often one of the main ways ambiguity enters a catalog.

A song may have a primary title in the publisher’s system, a slightly different title at a PRO, a translated title in another territory, a shortened title in a statement, and a cue-style title in a production music context. Or it may have all of the above, but the primary title it is registered under is what another party considers an Alternative Title- noting that often only the sending parties primary title is what appears on an export.

The problem is that without a disciplined internal view of titles, every external report becomes harder to reconcile because of this. You are no longer asking whether an outside dataset agrees with your catalog- you are first asking whether your catalog has enough structure to interpret what the outside dataset is trying to tell you.

A repertoire export from a society, licensing hub or administrator can be very valuable in surfacing missing registrations, conflicting claims, duplicate works, unknown identifiers and unlinked recordings. But the report is only useful if it can be brought into a system that knows how to compare, prioritise and resolve those signals.

Otherwise, it becomes another spreadsheet (and the music industry already has quite enough spreadsheets!)

Contributors are the deeper layer

The deeper layer still is contributor disambiguation.

A work cannot be disambiguated properly unless the contributors are also disambiguated. Two works may appear to be different because one writer is associated under a different name, pseudonym, patronym or even assumed name. There may be multiple IPIs, and for valid reasons.

This is where catalog work becomes less like database cleaning and more like rights archaeology. A duplicate work registration may have been created because a third party only had partial information. A derivative may have been treated as a new work because the underlying relationship was not clear. A remix, adaptation, translation or medley may have been attached to repertoire in a way that makes sense commercially but creates confusion operationally (as anyone who’s ever worked with samples will know well!).

The challenge is that each layer depends on the layer beneath it.

Recording-work match depends on a clean work list. A clean work list depends on title and identifier disambiguation. Title and identifier disambiguation depends on contributor disambiguation. Contributor disambiguation depends on being able to resolve names, IPIs, roles and historic registration behaviour with enough clarity to make a decision with confidence.

Large-scale catalog cleanup is rarely solved by one export, one enrichment source, or one matching exercise.

More data means more noise

The best catalog work usually involves pulling together as many different signals as possible: internal repertoire, PRO data, MLC data, Publisher data, royalty statements, DSP data, recording metadata, ISRCs, ISWCs, IPIs, alternative titles, purchase agreements and schedules… the list goes one.

But more data only helps when it can be handled accurately, efficiently and systematically. Without that structure, every additional source simply creates more noise. More files to open. More exceptions to review. More slightly different versions of the same title.

At small catalog scale this can be held together by memory and manual review. An experienced catalog manager knows their repertoire, knows the writers, knows the oddities, and can often spot any issue quickly. At large catalog scale, that same expertise is still essential, but it needs leverage.

A catalog of 500 works can be cleaned through careful manual attention. A catalog of 50,000 works requires process. A catalog of 500,000 works requires infrastructure. The underlying judgement does not disappear as the catalog grows; the need to deploy that judgement repeatably and consistently becomes the central challenge.

This is the real work behind “clean data”. Everyone wants a clean catalog, though very few people mean the same thing when they say it, or appreciate what goes into it (or what the potential returns will be for committing to such an investment).

A clean catalog is useful because it makes the next action easier. It makes claims easier. It makes conflicts easier to identify. It makes royalty statements easier to reconcile. It makes catalog transfers less painful. It makes valuations more reliable. It makes income leakage easier to spot. It gives experienced teams a stronger foundation for decision-making.

The compounding effect

The larger the catalog, the more important this becomes.

At scale, metadata problems compound. A small ambiguity in writer identity can create duplicate works, duplicate works can distort recording matches, and poor recording matches create missing revenue. Missing revenue understates catalog value, and understated value tangibly affects acquisition decisions, reporting, cashflow and long-term business returns.

Recording matches may be the obvious place to begin, but they are rarely the whole problem- they are the tip of the iceberg.

Catalog cleanup is not just about finding errors. It is about building the conditions where errors can be found repeatedly, prioritised intelligently and resolved with confidence.

Repeatable processes create economies of scale. Systemised actions create better returns from judgement over time. Clean internal data makes external reconciliation possible, and properly deployed tooling allows experienced catalog managers to focus on the decisions that actually require experience, and that deliver economic benefit.

Catalog metadata work sits quietly underneath the commercial life of a catalog, but when it is weak, everything built on top of it becomes harder to trust.