Interpolation Nation
In the 1970s it took an average of 1.7 songwriters to write a hit, today it takes around 4.5. That's a 165% increase, and the number is still climbing- Beyoncé's ‘Heated' carries eighteen credited writers alone.
The statistic usually gets told as a story about how pop is made: bigger writing rooms, more producers, more collaborative sessions. All true. But a significant slice of that growth isn't about who was in the room at all, but about who was added afterwards. Often the song is borrowing from another song.
Since the 90s there has been a slowly increasing trend to sample and interpolate, and it is quietly doing something to publishing data that the industry's registration systems were always unlikely to be able to absorb.
Why not borrow a song?
A sample is a “reuse” of the original audio recording- and as it is often built upon and incorporated into another creative work, it needs clearance from both the label (master) and the publisher (composition). An interpolation typically re-records or “re-interprets” the melody or lyric, so the master doesn’t necessarily come into it.
Of course, that can half the negotiation- typically removing the most expensive and slowest counterparty, and putting the whole thing within reach of artists who could never have afforded an original master clearance. Being the cheaper door into the same nostalgic and recognisable real estate is a key reason why interpolation has gone from occasional homage to a standard writing-room strategy.
Examples abound- from global smash APT. borrowing from Hey Mickey, to Venga Boy’s Boom Boom Boom interpolating ABBA’s Lay All Your Love On Me.
The commercial logic is unarguable. A familiar melody arrives pre-loaded with recognition- exactly the mere exposure effect we covered last year, bought rather than built. For catalog owners it has become a genuine growth channel: a fifty-year-old copyright can find itself earning from a brand new hit, in a market that oftentimes didn’t even exist the first time round- such as Tik-Tok.
None of this is a particular problem- the problem is what it does to the paperwork. The operational reality of an interpolation is the track is written, recorded and often already scheduled before the clearance conversation concludes (or sometimes even begins!).
The Registration Problem
Retroactive split changes are, from a data perspective, close to a worst case scenario. They force amendments across every society and hub where the work has travelled to- and as we know, that is not one place- it is several, on different timelines, with different processes. Each amendment is a chance for the varying registrations to diverge, which eventually ends up blocking revenue.
Then there always comes the harder question, the one that keeps catalog people up at night: is the interpolating song a new work, or a version of the original? In practice it is a new work with additional writers attached- a new ISWC (as new writers/roles). But of course, it doesn't always get registered that way. Sometimes it is delivered as a version (same ISWC), sometimes as a derivative (also sometimes same ISWC!), sometimes as an entirely unconnected work with an overlapping writer set. Theoretically, 18 different parties could have registered Beyoncé's ‘Heated', with each having a defensible view of their claim, and as a result producing 18 incompatible records.
That is exactly the metadata iceberg in motion- ambiguity at the work layer that only becomes visible when money fails to arrive.
Inter-Inter-interpolations
Now for the fun part…. this often compounds, because interpolations can themselves be interpolated. Sampled tracks can be sampled.
Latto's ‘Big Energy' interpolates Mariah Carey's ‘Fantasy', which itself interpolates Tom Tom Club's ‘Genius of Love'. So a single contemporary record inherits a chain of derivation three generations deep, and with it every writer, publisher and society attached to each link. The credit list is not long because a lot of people wrote it., but because it is carrying its own family tree!
Every generation of that tree brings its own negotiations, clearances and splits, let alone its own territorial arrangements, repertoire class, and history of registration errors. An interpolation doesn't just add writers, but inherits the entire data quality of everything it borrowed from.
Interpolation exposure is now a real feature of catalog valuation, on both sides. If you own classic repertoire, interpolations are upside- but only the ones registered correctly are realisable upside. Everything else is a well-performing record generating income that you cannot collect. If you are buying a contemporary catalog, you may be buying works whose splits are still capable of moving, attached to underlying works you haven't examined (or that could be open to litigation if clearances haven’t been sought- including further up the generational ‘tree’).
These should be key diligence considerations. I’ve seen enough APA schedules with unknown registrations featuring hundreds of writers, and often with some key name- Eminem, JayZ, etc.- that signals sample use. Some catalogs can be heavily affected, and that impact needs to be considered (both ways) in any valuation.
Creativity > Administration
None of this is an argument against interpolation. It is one of the few genuinely creative-and-commercial, hand-in-glove, trends to have emerged facilitated by the adoption of technology. It puts money back into older catalogs, and gives new artists a route into familiar material. It also produces some very good records!
The problem is that the creative practice has scaled and the administrative practice hasn't. Splits are still agreed in emails, filed by hand, and amended after the fact across dozens of systems that were designed on the assumption that a song has a small, fixed set of writers and stays that way.
A song with eighteen writers is not eighteen times harder to administer than a song with two- it is many MANY times more challenging because you are actually managing the number of relationships between those writers, multiplied across every hub the work touches (and multiplied by what generation of sample/interpolation it is).
The industry has got better at clearing interpolations, but arguably not correspondingly so at registering them. And (of course I saved this for the last line), this doesn’t even touch the use of generative AI- the ultimate interpolation scaling machine. THAT, is certainly a topic for another day…
Comments