I’m on Apple Music - but discovery comes mainly through the radio and Shazamming out in the real world - retail and hospitality mainly. I’m also working with a company that does curated music for brands so I find this subject endlessly fascinating 🤓
Another great newsletter! My main gripe is that I usually just shuffle my 1,321 liked songs (chaotic) but I don’t think it reallyyyy shuffles. I just hear the same songs over and over.
Hmm interesting takes. I think all of your suggestions are super valid and should be sent to Spotify R&D teams immediately - I hope they're already thinking of such things. That being said, I think the key to using their ai and algorithmic features (much like most other platforms) is that you have to put in a bit of your own work. Give a little, get a little. I have loved their algorithms ability to find new music and explore new genres, especially with the Discover Weekly playlist and the Radio feature. I'm continually impressed with the way it can create a radio based on one song and keep me entertained for hours.
THAT being said, I have a soft hypothesis: I think the algorithm/ai is confused with many popular songs. For example, if someone listens to Espresso by Sabrina Carpenter and adds it to a playlist with a bunch of obscure unknown artists, I think the algorithm's takeaway is still "oh this person is a pop normie". It seems to pull algorithmic learning from the song instead of the user. If that user went to the 'recommended songs to add to this playlist' section after adding Espresso, I bet it would recommend a bunch of pop girly songs bc of the algorithm's misdirection.
When your taste is a bit obscure to begin with, I think the algorithm has a bit more to chew on. Soft hunch, could be totally off.
right, it definitely could be dependent on genre and data for each genre. like people who like espresso --> like chappel roan --> whatever the algorithim knows mathmatically is the most logical artists based on sabrina carpenter. and for that, there are millions of data points and users who behave similarly. i do think it could be attributable to my music taste where my top artists are probably less "pop-forward" like peter cat recording co., olivia dean, jungle, etc. then it has less data to run off of, and i personally am usually hitting skip on a popular pop song. that being said, i should tell their team this, maybe i will shoot them an email lol. im interested on how they plan to present their future features and how they deploy ai for discovery.
I love Spotify, I really do. I love curating playlists, interacting with friends profiles and having songs I used to listen to back in 2013. However I do find that their suggestions and music discovery are awful... It is very annoying to have the same 10 songs suggested while creating any new playlist; I hardly ever found anything useful on my discover weekly; for as long as I remember it suggests music based on my location, except I don't listen to a single song from my country - the suggestions are never based on my taste. I'm glad I usually find music in more organic ways such as live sets, movies, friends, stores, etc. The features you mentioned are great and I also would love to see friend activity on mobile as I don't always listen to music on my laptop. Great post!!
same here! i feel like it is the same set of 20 songs maybe that is on repeat. i usually just go to explore other genres or try to look at what my friends are listening to when i am stuck without new music.
I think another reason people "don't care to beat the algorithm" is because the path of least resistance (aka the algo) is done for them. However, organic discovery is still available for those who want to put the work in, but Spotify presents the easy but arguably less effective way, which discourages the work. This reminds me of how putting a research paper into chat GPT is the easy way out, but the better paper will most likely result from actually doing the research yourself, if you equate the research paper to your Spotify playlist.
That said, I have some music discovery options on Spotify that I like and others I dislike. I agree that Daylist, Daily Mixes, AI DJ, and Discover Weekly regurgitate what you've already listened to. I do most of my music discovery through the Smart Shuffle Recommendations on my playlists, the Release Radar playlist, specific song radios they create, and also when my playlist runs out of songs and plays "songs recommended for you". With so many options for music discovery, maybe it's just about finding which ones work for you.
So happy I'm not the only one struggling with this one. I've seen some recent success in changing my algorithm by doing nothing other than listening to some truly random stuff. I've started new playlists that only feature relatively unknown songs from artists I haven't listened to in years. It's helped some... but I still get the same 15-20 song recommendations regardless of how hard I try not to listen to them.
It's so interesting because a lot of artists that I found on my own through actively searching on YouTube before the pandemic became really popular via social media like TikTok in more recent years - including Jungle, Steve Lacy, Kali Uchis, etc. I use Apple Music over Spotify and don't use their recommendations because they weren't the best in the past. But I do feel like my music taste has gotten so bland since Covid because I stopped searching for music on my own. Every song I download at this point is sadly just something catchy I heard on social media. I'm just busier now so I'd have to take out extra time in my day to actively search for unique artists, but I know it would be so worth it if I did.
I’m on Apple Music - but discovery comes mainly through the radio and Shazamming out in the real world - retail and hospitality mainly. I’m also working with a company that does curated music for brands so I find this subject endlessly fascinating 🤓
Another great newsletter! My main gripe is that I usually just shuffle my 1,321 liked songs (chaotic) but I don’t think it reallyyyy shuffles. I just hear the same songs over and over.
Agreed! Same thing happens to me too.
Hmm interesting takes. I think all of your suggestions are super valid and should be sent to Spotify R&D teams immediately - I hope they're already thinking of such things. That being said, I think the key to using their ai and algorithmic features (much like most other platforms) is that you have to put in a bit of your own work. Give a little, get a little. I have loved their algorithms ability to find new music and explore new genres, especially with the Discover Weekly playlist and the Radio feature. I'm continually impressed with the way it can create a radio based on one song and keep me entertained for hours.
THAT being said, I have a soft hypothesis: I think the algorithm/ai is confused with many popular songs. For example, if someone listens to Espresso by Sabrina Carpenter and adds it to a playlist with a bunch of obscure unknown artists, I think the algorithm's takeaway is still "oh this person is a pop normie". It seems to pull algorithmic learning from the song instead of the user. If that user went to the 'recommended songs to add to this playlist' section after adding Espresso, I bet it would recommend a bunch of pop girly songs bc of the algorithm's misdirection.
When your taste is a bit obscure to begin with, I think the algorithm has a bit more to chew on. Soft hunch, could be totally off.
right, it definitely could be dependent on genre and data for each genre. like people who like espresso --> like chappel roan --> whatever the algorithim knows mathmatically is the most logical artists based on sabrina carpenter. and for that, there are millions of data points and users who behave similarly. i do think it could be attributable to my music taste where my top artists are probably less "pop-forward" like peter cat recording co., olivia dean, jungle, etc. then it has less data to run off of, and i personally am usually hitting skip on a popular pop song. that being said, i should tell their team this, maybe i will shoot them an email lol. im interested on how they plan to present their future features and how they deploy ai for discovery.
I love Spotify, I really do. I love curating playlists, interacting with friends profiles and having songs I used to listen to back in 2013. However I do find that their suggestions and music discovery are awful... It is very annoying to have the same 10 songs suggested while creating any new playlist; I hardly ever found anything useful on my discover weekly; for as long as I remember it suggests music based on my location, except I don't listen to a single song from my country - the suggestions are never based on my taste. I'm glad I usually find music in more organic ways such as live sets, movies, friends, stores, etc. The features you mentioned are great and I also would love to see friend activity on mobile as I don't always listen to music on my laptop. Great post!!
same here! i feel like it is the same set of 20 songs maybe that is on repeat. i usually just go to explore other genres or try to look at what my friends are listening to when i am stuck without new music.
I think another reason people "don't care to beat the algorithm" is because the path of least resistance (aka the algo) is done for them. However, organic discovery is still available for those who want to put the work in, but Spotify presents the easy but arguably less effective way, which discourages the work. This reminds me of how putting a research paper into chat GPT is the easy way out, but the better paper will most likely result from actually doing the research yourself, if you equate the research paper to your Spotify playlist.
That said, I have some music discovery options on Spotify that I like and others I dislike. I agree that Daylist, Daily Mixes, AI DJ, and Discover Weekly regurgitate what you've already listened to. I do most of my music discovery through the Smart Shuffle Recommendations on my playlists, the Release Radar playlist, specific song radios they create, and also when my playlist runs out of songs and plays "songs recommended for you". With so many options for music discovery, maybe it's just about finding which ones work for you.
So happy I'm not the only one struggling with this one. I've seen some recent success in changing my algorithm by doing nothing other than listening to some truly random stuff. I've started new playlists that only feature relatively unknown songs from artists I haven't listened to in years. It's helped some... but I still get the same 15-20 song recommendations regardless of how hard I try not to listen to them.
It's so interesting because a lot of artists that I found on my own through actively searching on YouTube before the pandemic became really popular via social media like TikTok in more recent years - including Jungle, Steve Lacy, Kali Uchis, etc. I use Apple Music over Spotify and don't use their recommendations because they weren't the best in the past. But I do feel like my music taste has gotten so bland since Covid because I stopped searching for music on my own. Every song I download at this point is sadly just something catchy I heard on social media. I'm just busier now so I'd have to take out extra time in my day to actively search for unique artists, but I know it would be so worth it if I did.