Spotify Vibes: Rethinking How We Enjoy Music
Since the proliferation of music streaming services about a decade ago, the fundamental user experience has remained largely unchanged. It’s 2018 and Spotify is mostly a collection of manually saved songs and albums, alongside a myriad of pre-made playlists. The golden promise of a personalized experience- “the right music at the right time” never fully materialized as the burden of the musical journey is largely shifted onto the user as choice paralysis detracts us from enjoying the music itself. Despite spending hundreds of hours with us every year, Spotify remains an amateur DJ who plays songs we like from time to time but is mostly tone-deaf to our mood and atmosphere.
The Missing Piece
“Those who were seen dancing were thought to be insane by those who could not hear the music.”
As of today, Spotify’s pinnacle of personalization is the Daily Mix section, consisting of several playlists sprinkled with new songs which you might also like:
As you can see, Spotify neatly groups songs into genre-specific playlists which is completely logical… the problem is that we don’t think logically when it comes to music. When I think of Bohemian Rhapsody I don’t think 70’s British Rock & Roll- I think of how it makes me feel, and those feelings and emotional states that influence our music selection are known as vibes.
While Spotify focuses on reverse-engineering song ingredients like beats-per-minute, sub-genres, melodic frequencies, etc… they are neglecting the most important aspect of the music experience- our feelings. No amount of data or algorithms could ever personalize the experience in the absence of emotional context. Sure, workout music could be framed in the context of a location: the gym, or activity: fitness, but perhaps it’s better described through an emotional lens: hype, motivation, grit. People don’t think to themselves “I really want to listen to a melodic, 2000’s pop song with a high BPM, and a female vocalist.” They think “I really want to listen to some feel-good, upbeat music as I go for a run.”
Capturing Vibes
The nature of vibes is extremely fickle and transient as they’re influenced by everything ranging from the time of day, location, friends, weather, life events, memories, and countless other factors. Spotify’s next inclination might be to determine our mood by shifting the algorithmic focus from the music to the user by running extensive analytics, tracking choice variation over time, correlating songs attributes to musical instruments, inferring feelings based on genre, looking for patterns between lyrical connotations…
…or they could just ask.
The Language of Vibes
Vibes are inherently abstract and contextual, which means they often evade the grasp of language. But what if users visualized their vibe using emojis? After all, they’re fun, brief, expressive and offer endless possibilities.
Creating a Vibe
Users can choose up to 3 emojis that express their current vibe, then continue to use Spotify just as they normally would. Meanwhile, Spotify is learning which songs users associate with the given vibe:
As the user skips, fast forwards, rewinds and replays songs, the algorithm automatically attributes those songs (or song segments) the user’s vibe and remembers it for next time that vibe is selected.
Imagine the next time you open Spotify, you no longer have to scour recently played songs, saved albums, or scroll through playlists to best suits your mood. Instead, you choose from a menu of your pre-defined vibes:
Assigning Vibes
If at any point a user is listening to a song and wants to“tag” it with a vibe:
Quick Curation
I redesigned the Vibe playlist with quick curation in mind by giving songs a card UI which allows users to quickly swipe and remove unwanted songs, which is important. The current workflow requires more steps and is not as intuitive.
Feeling the Blues
Each vibe is assigned a unique color gradient that emerges over time, which reflects the sentiment of the given music associated. Although this seems quite subjective, research indicates the synesthetic association between music and color is surprisingly universal and quantifiable.
Magic Behind the Curtains
The clean and simple UI is not worth much in the absence of great algorithms on the back end. If user actions such as tapping Next, rewinding a song, turning up the volume at a particular section, etc, are observed in the context of a vibe, interesting insights could emerge and improve the personalization experience.
Going Forward
In 2011, I wrote an article outlining how our smartphone sensors could pave a path towards customizing our music experience by correlating sensor data with our music selection in hopes of gaining insight. I still think the more our phones know about our context, the better DJ Spotify could become:
- Time- the music I want to listen to at 9 AM on a Monday varies drastically from 11 PM on a Saturday.
- Weather- gloomy skies and thunderstorm call for very different music than sunny days. (GPS)
- Location- at the office, I’m likely listening to music that helps me focus. At the gym, perhaps something to amp me up. (GPS, WiFi)
- Speed- walking, biking, or driving surely affect my music preference, (GPS, accelerometer, gyroscope)
- Friends- perhaps the biggest factor of them all is our present company since we’re likely to play music that suits everyone’s taste. Assuming my friends also use Spotify, even crazier correlations could be made. (Bluetooth / WiFi)
When combined, all those inputs truly are greater than the sum of the parts. Spotify could, in theory, know that I’m biking to work on a sunny morning with my co-worker Kris. This level of context would be a significant advancement in personalization, although privacy concerns would have to be addressed.
My Process
This concept came about from my personal frustration with choice paralysis everytime I open Spotify. I started inquiring how my friends and family used Spotify and found out nearly everybody developed some sort of a shortcut such as going directly to “Recently played” or going to their entire library and hitting shuffle. This extremely limited validation was enough motivation for me to explore this concept further. I started out by mapping out the iOS Spotify app in order to fully understand the navigation architecture:
Next, I mapped out Spotify’s various workflows and mental models in order to further my comprehension:
After gaining a better understanding of the Spotify’s architecture, came my favorite part- brainstorming experimental interfaces that could be used to capture a user’s vibe.
After exploring interactions such as roulette wheels, dials, drag-and-drop, tinder style matching, and even shaking the smartphone to extract a tempo, I settled on using color to represent vibes. I iterated several different approaches to a minimal UI color input mechanism:
After some user testing, it became obvious that people were struggling with assigning colors to vibes as they found it too abstract. Then the idea of emojis popped into my head, so I started exploring different presentations:
Ultimately I went with the dark background due to Spotify’s brand, and single rows for clarity, ending with the final product, which people found intuitive:
Although this concept is catered to Spotify due to my personal use, it’s worth noting that it could be applied to Apple Music, Tidal, Pandora, or other streaming services.