My playbook to hack the Spotify algorithm
How to discover new music on Spotify, a list of pro-tips and hacks.
Spotify is by a significant margin my longest running personal subscription, having been a paying subscriber uninterrupted for over 13 years. My original love for Spotify was driven not only by the fact that it had most of the staple songs from my iTunes library, but especially that it had so much weird and obscure things. One of my early playlists was just cool covers of well-known songs. You can go into random genre deep dives, from modern classical, to classic funk, to math rock. You can listen to the entire discography of Elvis Presley, and if you’re planning a trip to France you can pull up a playlist of the best French Chansons in 5 seconds. Music streaming is one of the coolest inventions in media ever, on par with Edison’s Phonograph.
But I have a big and long-running beef with the Spotify recommendation algorithm, and as Spotify has increasingly de-emphasised human made playlists and social discovery in favour of algorithmic suggestions, I’ve had to come up with new ways to beat the algorithm into submission. There are tips and tricks here that I regularly tell people in person, and one of the big perks of starting this Substack has been that in many cases I can send people a link to one of my posts on a subject instead of repeating myself 😂
Caveat: are you an explorer?
My taste profile is really, really broad and I am novelty-seeking. I want to hear new stuff and I want to be surprised. Obviously I also use Spotify as a background soundtrack to my life, but even in that capacity I’m looking for a quite high degree of variety. Therefore, my main gripe with the Spotify algorithm is it’s bias to repetition.
These Daily Mix playlists are one of the most pointless features of Spotify for me. When I first saw them my first thought was: “Oh cool! Those playlists have some familiar classics but also new songs that have a similar vibe, right? ……Right?”
But invariably they consist of songs I already know, and even worse, songs I have heard a large number of times in the recent past. Yes, I know the Rolling Stones. When I want to hear them, I’ll find them myself. Show me the good stuff that I don’t know already!
This is how I’m wired. If you’re not wired like that, if you have happily used the same running playlist 3x per week for the past 5 years, you probably don’t have the same problem I do. So in that case, here is what you may still get out of this post: You could do the opposite of what I do and perhaps still improve your Spotify experience, you could stop reading, or you could read on simply out of interest.
Where to discover new music, within the constraints of your own bubble
These are some good first steps. You probably know these already:
Discover weekly
Release radar
Genre specific playlists (via ‘Home’ and then scroll down until you find things like “Throwback”, “Focus”, etc)
Spotify’s built-in lists of “Fresh Finds” (there are several for different genres), which as far as I can tell do adjust to your own taste profile
New playlist + recommended method:
Create a new playlist and toss in a couple of songs in the style that you want to find more of
Listen to the ‘Recommended’ songs that will appear at the bottom underneath the playlist and add the good ones
Spotify’s new AI based options
The Daylist:
A relatively new Spotify feature is the Daylist. It will have subtitles like “Funk groovy Tuesday evening” and it will change several times throughout the day. When this feature just came out it was the best thing since Discover Weekly. It was so f’cking fresh! Unfortunately, Spotify being Spotify, after using it regularly it has started to try and back me into endless repetition again. The fact that I listened to Modern Blues Rock once on Thursday afternoon does not fucking mean I want to listen to Modern Blues Rock every fucking Thursday afternoon until the day I die!
But, having said that, the Daylist is still often good. Both in terms of the ‘genres’ it comes up with, and in terms of the music choices within them, it continues to be delightfully surprising.
The Daylist is obviously an AI feature that is underpinned by LLMs, and it gives you insight into its prompt as the subtitle of the playlist. Some of the delightful ones people have shared online are: dirty rock happy indie thursday evening, Folklore soul crushing afternoon, and Feelgood coastal grandmother.
In terms of hours played, I think the Daylist is one of my personal top used features on Spotify right now. If you’re sleeping on it, you have to give it a try.
AI Playlist (only on the mobile app)
Spotify now also has basically the inverse of the Daylist, or the ability to leverage the Daylist algorithm any way you want (at least that’s what I suspect). You can give it a prompt and it will create a playlist. For example, I can do: “Fusion Jazz funky and new” → I will get some of that. This is a great complement to the daylist because it has happened to me that I find a great Daylist but it refreshes before I listen to all the songs. If I remember the ‘prompt’ I can get a new list with similar music via this AI Playlist method.
As for this feature’s ability to explore new grounds, it’s a mixed bag. It still seems to bias quite strongly to things you already have in playlists or have recently heard. It also comes with the familiar weaknesses of current LLM based systems. When I asked for “Similar vibe as The Verve but modern and new to me”, the AI reported it had created a playlist of songs similar to The Verve that I had not heard yet, but all the songs in the playlist were actually literally by The Verve. When I followed up with “not BY the verve, just similar vibe..” it did some of that ✨Ai MaGiC ✨ and reported the songs were now not by the Verve anymore….but they all still were. Oh well!
Strong-arming your algorithm to adjust genres
Spotify’s algorithms have a strong recency bias; what you listen to recently disproportionately affects what it will recommend. You can use this to your advantage. My algo can sometimes get locked into my background music preferences if I’ve had a week with lots of focused work (ambient, lounge, lo-fi house, synthwave). That means I’ll end up having a shortage of other genres that I’m into, like rap and jazz. An effective way to straighten this out is to just leave your spotify running some rap music playlists and radio overnight. The next day, like magic, a lot of the dynamic playlists will have started adjusting to my ‘new’ taste. Now I can discover some new rap music (or whatever it is I want to get more of).
Excluding your kids’ stuff
If you have kids songs messing up your algorithm, you are just not taking this seriously enough. There are good ways to prevent this.
Exclude from your taste profile
On every playlist, you can press the 3 dots, and exclude the playlist from your taste profile. That will ensure it’s not taken into account for your Discover Weekly and any other recommendations.
So instead of searching ‘Baby shark’ and letting Spotify run with it, take 5 minutes to create a couple of playlists with your kids’ favorite kids songs, and then exclude them from the algo.
Spotify Family
If your Spotify account through some bogus narrative or false dichotomy is in danger of becoming the ‘household account’ linked to Sonos or whatever: DO NOT FALL FOR THIS. Just get Spotify family and create a dedicated household account for free to use with your home system. Ringfence that shit!
Problem solved. No excuses on this one.
Getting out of your bubble, within Spotify
Alright, if you made it this far you must be dedicated to discovering new (to you) music. We’re getting into higher effort territory here. It’s clear that the only way to fully break out of Spotify’s burning desire to back us into a corner, is to break out of our own algorithmic filter bubble completely.
Getting other people to share their Discover weekly with you
One well known, effective, fun and even social way to do this is to ask other people to share the link to THEIR Discover weekly. That will give you multiple Discover weekly’s in your side bar to listen to every week.
This works really well. Except of course that your friends are also being pushed into their own respective corners, and you may get tired of their filterbubbles after a while as well. But still, this works.
Making ‘Blends’ with your friends
If you press the big ‘+’ (New playlist) button and choose the option to create a Blend you’ll get a link that you can send to one or more friends. Once they accept, Spotify will create a playlist that blends your taste profiles. It’ll also show who originated each song: was it in both your playlists, or just one or the other. By going through and focusing on the songs that were only in your friend’s playlists, you can discover some new bangers. But I personally have been quite disappointed in Blends. I made a blend with one friend specifically because of his taste in Hip Hop and Rap. But of course he also listens to other stuff. Because Spotify tries to find the commonalities in your taste profiles, it ends up excluding all that good Hip Hop and give me things that are very close to what I already listen to. Sad.
Searching and following playlists made by users, not by Spotify
This is another sure-fire way to break out of your own algorithmically designed mansion where the same damn music is playing behind every single door you open. Some things to keep an eye out for:
Skip passed all the playlists that are “By Spotify” (duh)
Assess real users briefly: Is this a real person, or is there some commercial intent? There are a lot of record labels that create playlists. Many are not so great. Some are
Check if the playlist is being actively updated (sort by date added)
Getting out of your bubble, outside Spotify
Find playlists curated by artists
A bit of a next-level version of the previous one. A nice example is this one curated by members of the Bombay Bicyle Club. Another great one is by Pepijn Lanen, a Dutch rapper. Unfortunately, these playlists tend to predictably suffer from two issues: (1) the artist is trying too hard to display an obscure, high-brow taste, sacrificing listenability, and (2) the artists often abandon the project. But it is totally worthwhile to google if some of the artists you like have playlists of other music they like. You might get lucky!
Check if your favorite bar of coffee shop has a public playlist
You know the kind of place that takes pride in its soundtrack? Maybe there are turntables in a corner, maybe there’s a record player. The people who work there look like they’re into music and go to cool concerts. That place probably has a Spotify playlist, and if they’re not trying too hard to prove themselves, it might be excellent.
Find dedicated curators of music
You get the point. Some examples:
The features I wish Spotify had
I would like Spotify to simply have a property in a user’s taste profile like ‘adventurousness’ or ‘noveltyseekiness’ which, when high, tells the algorithm to increase randomness in the recommendations, and bias more strongly to music the user has never heard before.
Another cool option would be to have algo generated playlists that correspond to various ‘anti-tastes’, different poles, adjacent taste profiles, ‘make it more obscure’ button, deep cuts, etc. In short, more ways to use the current algorithm1 in ways that better serve those with a low tolerance for repetition.
In the meantime, I guess I’ll just have to settle for the small satisfaction of getting Spotify to pay money to a bunch of artists every time I leave it running at 0 volume for a night of ‘algorithm realignment’.
Which, for the record I do want to say I find one of the best recommendation engines there is. If you compare it with Netflix or Youtube, Spotify is a shining beacon of recommendations done…better. But, not quite done right.
💯 you forgot the ‘Blend’ option, combining you and your friend’s taste. Not a mind blowing feature, still some useful new finds though.