Keeble

i work at a store that lets workers choose the music played and uses spotify premium to do so. historically this has meant a lot of spotify artist radio being chosen, which i don't care for for several reasons:

1. by spotify's own admission in their fine print, commercial considerations affect what spotify chooses to recommend to you or not. I'd rather not be marketed to by the major record labels thank you very much
2. they're usually only like three hours long before repeating. as repetitive as a typical store playlist

Personally, for playing music for a store as a worker i loathe playlists short enough where im hearing the same songs more than like twice a week. so my initial solution was a playlist called 626music (not a reference to stitch, as many people think, but im fine with that association) that i threw literally anything i thought should play in a restaurant, regardless of genre or mood or whatever. as a true lover of freeform radio, ive never cared for limiting oneself to genre all the time with mixes, it often ends up just being a marketing proxy and too often separates mixes into "all black" and "no black people" categories. eventually though, this got really big at a certain point:

this is big enough that we can use the smarter playlist utility developed by spotify engineer Paul Lamere that, essentially, allows you to do itunes smart playlist stuff with spotify playlists. here's an example of one of the playlists i generated:

lets make sense of this together, going in the order of the flowchart:

1. we start with the base "626music" playlist, in its 3000+song glory. that goes into
2. a de-duplicator, which first rids the playlist of any repeats if they happened to get in there
3. the tracks are all annotated with spotify data. see, spotify stores massive amounts of data about the songs within that it does not publicly displayed to the user, but via its api are indeed publicly attainable. some examples of the data spotify stores are: energy (how intense, noisy, loud, and/or fast a track feels), valence (A measure from 0.0 to 1.0 describing the musical "positiveness" or happiness conveyed by a track), along with basic data like artist and track popularity.
4. the tracks are shuffled, since the maximum output playlist size is exactly 1000 (though can also be whatever length you like shorter than this, too). this ensures that if the script is run multiple times, it will generate slightly different playlists each time
5. here's the important one. this takes those energy values that we found earlier and filters out any tracks with an energy level below 0.3/1 and above 0.7/1, leaving us only with tracks that spotify considers roughly "mid energy". this is how i chose to make this filter, but really you could filter in or out based on any of these values spotify stores.
6. the resulting 1000 songs are then sorted in order of reverse energy (i.e., the most intense songs first)
7. this is saved as a new playlist

what we're left with is a playlist that's completely free of industry interests that might impact how spotify feeds you its music, but still linked sonically. here's a roughly cd length segment of it when shuffled:

look at this selection, which defies any traditional genre logic of transitions. roy ayers, a funk/jazz/rnb musician from the 70s, goes straight into an electronic pop caroline polachek song. a prefab sprout song produced by thomas dolby (you know, the "she blinded me with science" guy) goes into a tito puente cover done on vibraphone. what links all of these is being, well, neither exactly chill nor inappropriately energetic for a crowd. i also used this site to schedule this playlist to update monthly.

and since there are all these playlists ive made, anything new i like i can just put in the main, gigantic playlist, and the child playlists will gradually update to feature the new stuff. and since there are 12 of them its easy to find one that matches a mood: one playlist of merely the 1000 most popular songs i chose for the big playlist. another with "chill" tracks. a third with "happy" stuff, a fourth for the 1000 oldest songs on the playlist. and so on and so forth.

ill end this by emphasizing that YOU CAN DO THIS TOO! if anyone would be interested, i can share some of the flows ive made on smarter playlists; all you would need to do is replace the source playlist with a different big playlist full of stuff you like (could be your saved tracks, could be a mass playlist you add any track you like, can be your most listened songs, etc). then, you can enjoy the joy of a well made smart playlist