Designing A Mobile User Experience to Teach Spotify Algorithm Song Categorization

Music Streaming
MY ROLE
UX Strategist
User Researcher
UI Designer
UX Writer
PROJECT DURATION
4 weeks
PROJECT ADVISORS
Spotify Design Team
Overview
Spotify prides itself on giving its users the right music for every moment. So following their recent launch in Nigeria, Spotify wanted to understand the popular African and Nigerian music genres and sub-genres in Nigeria to make sure that their Nigerian users had the same awesome experience as their others.
Task
I was tasked with creating an in-app experience that allows users to teach the Spotify algorithm how to categorize songs across genres.
Solution
I designed a game that allows users to teach the Spotify algorithm how to categorize African and Nigerian songs according to their respective music genres.
Research Summary
  • Target Audience: Nigerians and people of other nationalities who actively listen to popular African/Nigerian music.
  • 10 user interviews
  • 12 usability tests
  • 3 critic sessions
  • 1 critic session with Spotify design team
Results
  • Increase in the amount of time spent on the app.
  • Increased discoverability of African/Nigerian music genres and artistes.
  • Increase in the number of African/Nigerian music listeners.
BREAKING DOWN THE TASK
This was my first time designing a user interface for artificial intelligence, so I started off by reading a few articles and research papers on AI-driven user interfaces. After my study, I had a clearer understanding of how to go about the task and sketched a few ideas that came to mind.
How does Spotify categorize African and Nigerian songs across these music genres?
Spotify audit
I carried out an audit on the categorization of African and Nigerian songs on Spotify which revealed some limitations.

This title is limiting and not encompassing. Not all African music genres are under Afro music.

Titles of playlists do not correspond to the existing music genres.

The “See more” button can easily be missed.

Spotify could not identify juju music as a Nigerian music genre.

Spotify identified Kwaito as a music genre but the genre has few playlists.

Spotify identified Fuji as a music genre but the genre has few playlists.

DEEPER INSIGHTS
Sometimes, these changes in genre while listening to music as were upsetting for users.

I carried out further synthesis on their behaviour and found out that they actually cared about genres but only when:
  • their playlist vibe gets disrupted.
  • they are looking for new songs.
  • they are creating a playlist.
  • they are not the only ones listening. (Eg. at hangouts, parties, etc)
How can I help users identify the genres of music they listen to so they are able to teach the algorithm?
Devising solutions
My solution was focused on helping users first identify the African and Nigerian genres they listen to and then providing them with a way to teach the algorithm how to categorize songs under those genres.

I explored two of the instances when users recognise a change in music genres for my solution;
  • when their playlist vibe gets disrupted
  • and when they are looking for new songs.
Exploring Users' playlist vibe DISRUPTION

Playlists based on only African genres users listen to are created by the algorithm.

The pop up message informs the users how to teach the algorithm song categorization.

When a song that doesn't belong in the genre comes up and disrupts users' vibe (listening experience)...

...they can swipe to remove the song from the playlist.

Exploring users' Search for new songs (Gamified)

“Play for free gift” attracts users and “African genres you like” let’s the user know what kinds of songs to expect

This provides more infor on the game. There is a looped transition between this card and the previous one.

The game comes in form of questions asking users to pick which genre each songs belong to.

Users wins hours of ads-free music at the end of the game and dismiss the card.

Usability Testing
I carried out 3 rounds of usability testings with a total of 12 participants.
Refining the designs based on testing insights
Playlist Vibe Disruption

Problem: Even though they found it an easier way to remove songs from a playlist, users aren’t used to the "swipe" action in music apps and kept trying other familiar methods. This made me realise that they didn’t read the pop up message informing them how to remove songs even though they paused to look at it.

Solution: I replaced the text in the pop up with an animation to capture users attention. The animation showed them how to remove songs from the playlist using the "swipe" action.
Problem: The "swipe" action was hard for users to learn as subsequent testings saw more users trying the options button or long press method.

Solution: I favoured the actions users were famaliar with for removing songs and utilised the opportunity as an entry point for the genre game (it performed better with users) but they only had one song to categorise, the removed song.

Search for New songs
‍
Problem: Even though they tapped all the right buttons, users didn’t understand what they were doing in genre game until they played it a second time.
They also felt like the game was a good opportunity for them to learn about music genres.

Solution: The copy at the start of the game wasn't clear on what the game was about. I changed it to reflect that and also make it clear that this was a teaching opportunity and not a learning one for them.

Problem: Users misunderstood the functions of the "Start listening" and "Listen later" buttons.

Solution: I changed the copy in the buttons to clearly communicate their function.

Problem: Users wanted easy access to all the songs from the game after playing. So I created an opportunity for them to find them in their notifications. They could also access their free gift via notifications when they opt to use it later.

The testings revealed that users rarely check their notifications.

Solution: I placed a playlist containing the songs on the homepage with an opportunity to continue the game at the end of the playlist, but without a free gift. 

I let the banner with the free gift remain on the homepage for easy access.


Final Design

Try out the Genre Game here!

Results
Increase in amount of time spent on the app.
‍
90% of users found the Genre Game fun and engaging. They also loved the free gifts for free version users.
Increased discoverability of African/Nigerian music genres and artistes.
‍
100% of users discovered African/Nigerian music genres they never new of and were delighted by this.
Increase in number of African/Nigerian music listeners.
‍
With increased discoverability, users are likely to find new African/Nigerian music they would love and share.
other notable insights
Reflections

Consumers of music on Spotify are probably not the best users to learn from.

Artistes are better equipped to provide information on their songs genres. They can be required to attach the music genres their songs belong to at the time of their release on Spotify.

Takeaways
  • “Think long term”

    The idea of a free gift as a reward for playing the genre game could be expensive for Spotify in the long run.
    ‍
  • “Recognise when to change direction with solutions”

    Especially when the ones at hand have expensive risks or too many bottlenecks, instead of trying hard to force a particular solution to work.

other Projects

Designing A Feature To Acquire And Retain Customers

Helping Users Onboard Faster And Easier

Engaging & Onboarding New And Potential Cus...

HomeResumeAbout

Let's Connect!

©2022, Ifunanya Ugochukwu