The AI overlords are taking over!
And they're curating incredible playlists made just for you.
How Spotify cracked the enigma of music curation and built robots that suggest songs you actually like.
I'm kind of a Luddite. I prefer old, proven technology to the new stuff. I'm not exactly a late adopter, but you won't find me at the cutting edge either.
That's why when Spotify launched Discover Weekly in 2015, I was skeptical. I had been a Pandora Radio user for the good part of a decade. It was fine for car rides and background music, but less than ideal for discovering new music.
The best discovery method at the time was still recommendations from friends. I didn't believe Spotify had anything new or different to offer, and their previous attempts at mimicking Pandora's radio channels were less than successful.
Fast forward to today and you'll find me regularly listening to my Spotify Discover Weekly playlist, adding new songs to my "Liked Songs" playlist. I still enjoy finding new songs through friends, but Discover Weekly is one of the largest contributors to my evolving taste in music.
What makes Spotify Discover Weekly so successful?
It's all in the alghoritms, my friends.
Spotify has discovered the secret of music recommendations and has trained their robots to serve up personalized mixtapes that feel like they were curated by a friend.
And that's because they kind of are.
Discover Weekly creator, Matthew Ogle, says the alghorithm works like this:
The real core of it is looking at the relationships between songs based on what other users are playlisting around the songs that you've been listening to and essentially finding the missing ones – the ones you haven't heard yet.
Spotify knows what songs you like based on what's on your own playlist. It then cross-references those to what other users who like those songs also listen to.
That sounds exactly like how I was turned onto the Mars Volta back in 2005. A friend of a friend and I got talking about song of our favorite bands — At The Drive-In and Led Zeppelin both came up.
"So there's this band called The Mars Volta that's led by the singer of At The Drive-In," she said. "And they really channel that 1970s sound that Led Zepplin is all about."
Spotify does this over and over again until it has 30 songs — a great mixtape length — that it hand delivers to you each week.
But here's the trick.
The alghorithm also sprinkles in a little bit of familiarity. Sometimes it's a song from a band you already know well, or it could be a song you favorited six months ago.
Ogle says that's because of a happy accident early on in the products' development.
The code meant to filter out stuff users' already listened to had a few bugs. It accidentally added some of those songs to the playlists. Ogle and his team found when they fixed the bug, people were actually less engaged.
So they did what any sensible product or marketing team would do. They added the bug back in.
Something for everyone.
Before apps like Pandora and Spotify, everyone was listening to the same thing.
Charts like Billboard's Top 100 and Rolling Stone's Top 100 Songs have boxed music listeners in for over half a century.
Like network television, these charts served up what everybody's listening to, and seem to say, "hey, why don't you listen to it too?" They're a bit like that middle-aged uncle who's desperately trying to stay relevant.
But music is all about rebellion. Listeners want that thing that no one else has heard. They want to be the first to discover a band before they were popular.
Listeners don't want what everyone else is listening too.
Everybody wants to be somebody special.
There are so many tips and tactics out there for how to personalize your marketing. But most of them are half-assed and based on the assumption that your audience is stupid.
Sure, you could add a [FirstName] tag into every email, but most subscribers still know that this is sent in bulk. It's insulting really.
What makes Spotify's tactic so powerful is that it's completely transparent.
Yes, this is built by an alghorithm. Yes, we get it wrong sometimes. Yes, we use your data to build playlists for you and others. But it's mutually benefitting. So...
When Will Glasser, Jon Kraft, and Tim Westergren started the company that would become Pandora Media, their goal was to create a personalized radio station for each user. Spotify took that one step further with Discover Weekly.
They wanted to build a mixtape for every user every single week. What's more personalized than that?
And it's contributed to Spotify's reign as the number one music app, with over 2 billion hours of music streamed by listeners since 2015.
What can we learn from Spotify?
Spotify does a few things well with its Discover Weekly product.
They're completely transparent about what the product is (automate music discovery), they deliver something personalized that doesn't insult users' intelligence (a mixtape built by AI), and they provide a service that solves a huge problem (how to find new music).
The question is...
How can you do the same for your product?