A data alchemist, according to ChatGPT, is someone who has "exceptional skill in transforming raw data into valuable insights and actionable information." If you could digest that word salad and put music to it, you would begin to conjure the likeness of Glenn McDonald.

McDonald was music streaming app Spotify's data alchemist until the end of 2023. Spotify is the world’s most popular music streamer and Glenn McDonald, described by Billboard magazine as being like a "song-recommendation algorithm in human form," helped shape the tastes of millions of music lovers. He is, he told Ray D’Arcy, a passionate believer in the power of music to unify people:

"I believe music and food are the two things that break through our fears of the other most effectively."

The idea is that by listening to other people’s music and by eating their food, it becomes clear that they’re not as strange as you might have initially thought they were. It’s a solid notion, unless the other people are big fans of the panpipe version of Ice, Ice Baby and are serving you airfried hedgehog meat.

The recorded music industry peaked, Glenn says, with the CD in the 1990s, when people bought music they already owned on vinyl just so they could have it on CD. Then Napster came along and things would never be the same again. And although the industry still hasn’t got back to its 90s peak, it’s now pretty close, thanks to people deciding to stream their music with the likes of Spotify, Apple Music and Amazon Music, rather than rip CDs onto hard drives. Glenn explained the reason the major labels bought into streaming in the first place:

"Each individual person would be spending less money than the heaviest spenders spent in the CD era, but the average would be much higher."

Streamers like Spotify amass an enormous amount of data from their users and that’s where Glenn’s data alchemy came in. His job was to figure out what to do with all that data:

"The main thing that I found that we could do is basically collate collective knowledge. Like, when you as a user listen to music, you don’t do it randomly. You don’t just listen to random songs from the universe, you have tastes and you listen to things that you like and in doing so, you’re forming knowledge. You’re expressing your taste and your expertise and what you know about music. And when you put 600 million people’s listening together, that is a lot of knowledge."

The end goal of the collation of all that knowledge was that the algorithm gives us recommendations for music it thinks we might like, based on what it’s observed us listening to over time. But the purely machine-driven recommendations could lack the human touch:

"Algorithms don’t think, really. And they don’t have opinions and they don’t have emotions and they don’t really understand this idea of liking. And this can be a problem because the fact that you listen to something a lot is a good sign that you probably like it, but not always."

Maybe that album was only playing in the background and you forgot about it. The algorithm can’t distinguish that. It’s only interested in the patterns it can see in order to build its recommendations. But even the whole notion of recommendations is something that only humans would have (naturally machines don’t have notions because they're not Irish):

"The emotions and the words we attach to them are human acts, so saying, 'This is a recommendation,’ that’s a human attaching a supposed meaning to a pattern that the math finds."

Streaming services are all about the "math" and many recording artists will argue that the payment they get from streaming is not good. When Ray asks whether artists get some money every time a track of theirs is streamed, Glenn’s answer is worthy of Donald Rumsfeld and his known unknowns in its verbal gymnastics:

"You could sort of think of it as yes, but technically no. So, the way it works is, all the money – there's a system called pro rata, it’s Latin, but what it means is all the money that gets paid to Spotify gets put into a big pool and then all the streams that people play get accounted for and the money gets divided up according to your share of streams."

That's very yes but technically no. Glenn has written a book about his time at Spotify – You Have Not Yet Heard Your Favourite Song – and it gets two thumbs up from Ray. Glenn's algorithms are being augmented by large language models (LLMs) and AI should improve recommendations because, well, it's AI, right? We'll see. You can hear Ray and Glenn's full conversation by clicking above.

You Have Not Yet Heard Your Favourite Song: How Streaming Changes Music by Glenn McDonald is published by Canbury Press.