100DB was a music discovery project I started. This is part of a series of posts I wrote to explore the thinking behind it.
Thanks to folks from the mailing list for letting me know I missed a whole bunch of platforms and functionality in my last sweep of music-recommending services. Hereās some follow-ups:
Spotify is the 800 pound gorilla here so itās not surprising that folks had some thoughts. Specifically, I mentioned in the previous Research post that I didnāt have any social accounts connected to Spotify, so I was lacking that signal. A couple people informed me that Spotify almost certainly doesnāt use any friend/social recos. Instead (and apologies for over-simplifying the technical explanation) theyāre basically running smart diffs on my plays and playlists, and community playlists that contain some of the same tracks. Theyāre ignoring fuzzy concepts like āgenreā and instead, essentially, filling in the blanks with stuff IĀ mightĀ have listened to, had I known about it. Neat!
Expanding on that: human curation plays a not-inconsiderable role for Spotify ā their recommendations feel human-like because they incorporate human intelligence in the chain. Thatās expanded on inĀ this 2015 article from The Verge, which is very much worth reading in its entirety, but this part in particularly grabbed me:
Itās still humans who are doing the song selection and arranging, but instead of outside experts, itās users like you and me. Generating a human-curated playlist for each of Spotifyās users would be a challenge of mammoth proportion. “We probably canāt hire enough editors to do that,” says Ogle. So Spotify uses each of its users as one cog in a company-wide curatorial machine. “The answer was staring us in the face: playlists, since the beginning, have been more or less the basic currency of Spotify. Users have made more than 2 billion of them.” In effect, Discover Weekly sidesteps the man versus machine debate and delivers the holy grail of music recommendation: human curation at scale.
In a sense, this is reassuring ā I donāt know exactly where Iām heading with this project but I do believe whole-heartedly in the human, rather than purely algorithmic, capacity to delight. Good to know that Spotify does, too.
Despite my previous claim that Soundcloud doesnāt have a recommendation feature, it does, apparently! But: it obviously doesnāt have enough info on my likes, dislikes, or listening patterns to make suggestions more depthy than āchartsā and āmoodsā. I donāt maintain a personal Soundcloud account, so itās unsurprising that they canāt piece together any kind of profile of what I might like. Iāll give them props for surfacing a big collection of exactly theĀ kindĀ of content that I basically visit Soundcloud for: DJ mixes.
If Iām being honest I donāt really know a lot about Mixcloud. On first blush, itās basically a Soundcloud-alike. I donāt have an account, so Iām not sure what kind of recommendation engine might be running under the hood. They do seem to be focused largely on DJ sets, and have aĀ deepĀ set of tags to drill down into specific genres and sub-genres. I found something I enjoyed listening to within three clicks, so the discoverability here is pretty dang good.
I also want to highlight Mixcloudās āplayerā functionality, which is doing three clever things I havenāt seen anywhere else: if Iām listening to a track, I can hover on the play button of another track to get a short preview. As soon as I lift off, Iām right back to where I was in the original track. The player itself keeps a history of everything Iāve been listening to ā a kind of built-in visual bookmarking (without logging in!). Finally, as soon as Iāve listened to enough music to createĀ someĀ kind of minimal profile, it pops up a suggested next track right in the player itself. Good product thinking, Mixcloud team!
If Spotify is the 800 pound gorilla, Apple Music is the 650 pound gorilla. The website for Apple Music shows me extremely nice screenshots of Apple Music being used on extremely nice Apple hardware, which is pretty telling. Itās also (still) running Beats 1, essentially a live radio station, which strikes me as distinctly un-Apple ā an appendix-like remnant of their acquisition of Beats. Appleās evidently the more popular service in the States (vs. Spotify), but I donāt know anyone whoād recommend it as qualitatively ābetterā, per se.
The thing I really want to dig into vis-Ć -vis Appleās music efforts is Ping, a service that operated for two years before being shut down. It seems like it was āmusic discovery via social networkingā . Nearly all of the platforms Iāve been digging into in these posts tend to eschew social networking in favour of straight-up delivery and various flavours of recommendations ā but thereās something really compelling (to me) about a person-focused feed.
Why does Google have Google Play MusicĀ andĀ Youtube Music? Re-read my lukewarm take on āYoutube Musicā, this is the same animal with different stripes.
I forgot to mention Turntable in my āplatforms Iāve never usedā section from the last post, and I bring it up now as itās worthĀ taking a peruse of the olā Wikipedia on this oneĀ ā specifically āOn November 22, 2013, it was announced that Turntable.fm would be shut down in December 2013 in order to focus on its Turntable Live service, which allows musicians to perform interactive online concerts.ā They shut down altogether a month and a half later.
I think Turntable is incredibly interesting for two reasons ā the first being the recreation of real-world behaviours in virtual space, e.g. āIf a user decided to click the “awesome” button, their character began to sway their head back and forth, simulating how a fan would react to a song they liked in a club.ā There are very few instances I can think of where a 1:1 mapping of real-world action to an app action makes sense. (Successful uses of this are widespread in musicĀ productionĀ rather than listening.) That said, digital-only proxies for real-world interactions are emerging, like the real-time emoji reactions on Facebook Live.
Secondly, ignore everything I just said, becauseĀ Marshmello performed a concert inside Fortnite last week. āBy one (unsubstantiated) estimate, 10 million concurrent users attended the show in the game’s Showtime mode.ā Turntable could clearly envision this weird future we live in, but couldn’t quite get there. What does any of this have to do with 100DB? I have no idea, except perhaps a reminder to think more ambitiously.