ROLE(S)
UX Research, Strategy, Interaction & Design, Copywriter
ASK
A solution that helps the Spotify algorithm accurately categorize African songs, genres and playlists.
SOLUTION
An in-app gamification solution that leverages on the knowledge of African users to teach the Spotify algorithm how to accurately categorize African songs and genres.
Explore how how I collaborated with my team and a Spotify design team in creating an in-app gamification experience that allows users to teach the Spotify algorithm how to categorize songs across African music genres.
I hope you stay to the end and enjoy the ride :)
DISCOVERY
The Problem
A quick search on Google tells us that there are at least 41 primary genres of music with hundreds and hundreds of sub-categories of music under these genres. The Guardian even takes it a notch further and tells us - "From charred death to deep filthstep, there are 1264 genres that make modern music".
This and a few other articles brings us to the conclusion that there isn't a definite number on the number of genres that exist. This also makes things a bit complicated for media & entertainment and audio streaming companies that provide music to users as a service. The simplest way out is to categorize most of these music under the most similar popular genre, but the con to this is that it messes up users experience while trying to find music.
In Africa, there are at the least - 10 primary genres of music, but majority of popular audio streaming companies are unable to identify these genres, thus making categorizations of these genres, generic. Users aren't satisfied with this, as it tends to mess up the quality of the music playlist they're listening to. For example, a user that wants to listen to songs in an Afrobeat playlist, could get recommended a song that belongs in an Amapiano playlist. How do we solve this?
The Project Brief.
Spotify, the company is the world’s largest music streaming service provider, with over 365 million monthly active users, including 165 million paying subscribers, as of June 2021.
Spotify provides access to over 70million songs, 2.2million podcasts and 4billion playlists.
Spotify wants to learn more about African genres, and apply this knowledge by accurately categorizing and recommending African genres and subgenres of music to its users.
Research.
1.
User Interviews
.png)
10 Users
2.
Usability Testing:
Low-Fidelity Wireframes
04 Users
Primary Research
- User Interviews
- Competitive Analysis
- Competitive Analysis
3.
A/B Testing:
High-Fidelity Prototype
04 Users
PRODUCT AUDIT
In order to fully understand user and business point of views, I decided to get started with secondary research by conducting an analysis to see how African songs and genres were categorized on the Spotify app. Here are a few screens and findings from the analysis.
From the audit, I was able to diagnose that -
Spotify's categorization of African music genres is almost inexistent, and songs are mostly categorized in playlists rather than genres.
Following our first secondary goal, my next step was to identify and define the most popular and most sought-after African genres.
Primary RESEARCH
I prepared a user recruitment plan/survey and interview questionnaire alongside four other colleagues. Within our recruitment plan, we outlined a script to be delivered during the user interviews. The script included an introduction, and screening questions that would help us in determining who fit into our ideal candidate profiles.
After we had recruited 10 users that fit best into our candidate profile, we proceeded to conducting the user interviews.
User Quote
I think Youtube music is doing a good job at classifying African/Nigerian music. I was able to find a playlist on Afrosoul & African pop but it can be better. I haven’t seen any on Amapiano or Kwaito.
- Joshua
Product Audit.
In order to fully understand user and business point of views, I decided to get started with secondary research by conducting an analysis to see how African songs and genres were categorized on the Spotify app. Here are a few screens and findings from the analysis.
From the audit, I was able to diagnose that -
Spotify's categorization of African music genres is almost inexistent, and songs are mostly categorized in playlists rather than genres.
Following our first secondary goal, my next step was to identify and define the most popular and most sought-after African genres.
Pattern MATCHING x KEY Insights
.png)
-
Users mostly listen to music according to their mood or activity.
-
Users don’t know much about genres.
-
Spotify isn’t easy to understand during first use.
-
Users know some genres but can’t accurately categorize songs under them.
-
Users believe Spotify has a wider range of music compared to most streaming platforms.
-
YouTube Music is the major competitor.
Competitive Analysis
I decided to look into YouTube Music to see what made some users speak so highly of the app's categorization and recommendations. Here's what I found.
Ideation.
PROBLEM STATEMENT
Our target audience (Nigerians) have some awareness on the most popular African genres, but are unable to accurately categorize most African songs into their appropriate genres.
how might we statement (hmw)
How might we help users learn how to differentiate between African music genres, so they might be able to accurately teach the algorithm how to categorize songs across each genre?
PRIMARY USER PERSONA
After coming up with a HMW and problem statement to help move the project forward, I went on to craft a user persona which would help in guiding design decisions.
.png)
Prototyping.
LOW-FI WIREFRAMES
USING In-App Tips
Following the first part of our problem statement - Helping users learn how to differentiate between African music genres, I created an approach to achieve this goal by teaching users about African music genres.
.png)
Usability Testing.
DECIDING ON THE FIDELITY FOR 1ST TESTING PHASE
I designed the wireframes to look like mid-fidelity wireframes, so users would be able to understand the context more during testing. Then I proceeded to prepare and get the prototype ready for testing with Spotify users on a moderated virtual session.
USER QUOTES
I'm ignoring the card because I feel it’d lead me to other questions, which is something I want to avoid.
- Nancy
I'm hoping that the AI affected will be just for my feed alone, incase I'm not right about the genre.
- Damaris
User Pattern x Research Insights
-
85% of people are mostly able to correctly categorize the songs they know and like.
-
Users think Spotify is just trying to test their knowledge with the questions
-
Most people scan, than read through tips.
-
The card isn’t noticeable and is hidden away.
-
Users want to receive before giving, which means offering them an incentive
BEHAVIOURAL OBSERVATION
Primary Research
- User Interviews
- Competitive Analysis
- Competitive Analysis
Iterations AFTER USABILITY TESTING
I followed user pain-points from the affinity mapping derived from the testing and iterated before converting the wireframes to high-fidelity mockups.
.png)
After I had converted the iterations to high-fidelity designs, I wasn't satisfied. Users were okay with the new solution but they weren't ecstatic, which was the behavioral attribute we were aiming for in order to achieve maximum success.
A part of the product strategy was that we provide a solution which users would look forward to, not something that would bore them and which they'd most likely skip, so it was time to think outside the box.
The Solution:
Spotify's Genre Categorization Game
BIRTHING THE IDEA
The idea for a game randomly came a few hours later when I decided to take a break from my PC and hangout with a few friends. You can imagine the elation that came with the idea, but there was one problem.
I didn't really know how I would put together a game that would be able to teach users about African music genres, while simultaneously allowing them teach the Spotify algorithm how to categorize songs across genres. So I decided to reach out to my mentor, Dara Oke (Lead Product Designer, Netflix).
We got on a call and brainstormed on designing for algorithms, but each idea we came up with had a limitation of its own, so we decided to go off the call so I could make more external research, which resulted in me creating a moodboard consisting of pieces of articles and videos that perfectly resonated and helped piece the final solution together. Yay!
GAME NOTIFICATION MODAL
When the user logs into their Spotify app, a modal comes up after a few seconds, catching the users attention with the interactions, and informing them about the game - "Play & Earn Points". It also houses texts and interactions that hints the user on what the game entails.
The CTA "LET'S GO!" is designed with the intention of being catchy and to evoke a feeling of excitement in the user.
HOW TO PLAY
This screen is transitioned in after the user clicks on the CTA on the modal. The screen contains information on how to play the game and an interactive design that shows that the user is to drag the music art and drop it on the genre it belongs to.
There is also a primary CTA to get started into the game.
The Genre Categorization Game
The user is taken into game mode after clicking the CTA on the How To Play screen. Here, the user drags and drops each music art on the African genre they think it belongs to. The number of genre they're asked to answer is shown in the slider, and there is an arrow button, which represents "skip" incase the user can't answer the current one and wants to move on to the next.
I used a fun and exciting sound as the background sound to make users feel more comfortable and also feel like they're playing a real game.
Game Score / How The Points Are Calculated
Users are assigned songs according to their playlists. They aren't asked songs they don't play often - remember, users are more likely to accurately categorize songs they like and listen to often. This way, songs in each users' categorization game can differ greatly.
The African genre for each song is finalized according to the most selected genre. For example, if 80 Nigerian Spotify users had the song "Essence by Tems and Wizkid" in their genre categorization game, and 56 of these users selected Afrobeat as the correct genre, while the remaining 24 user votes were distributed among the other three genres, Afrobeat would be the correct and final genre, and the users who selected Afrobeat would be assigned 10 points each.
Each user gets their points summed up at the end of the game, and they stand a chance to win more points in order to earn 12% discount on their next Spotify premium subscription.
.gif)
Final A/B Testing
I made the final iterations and proceeded to testing the two high-fidelity prototypes with new participants.
USER QUOTES
This game is really fun and insightful. At least now I can proudly say that King Sunny Ade is a Juju musician.
- Janet
If I see a game notification to play a game and earn 12% discount on my next subscription, I’m definitely playing that game.
- Abdul-Azeez
User Pattern x Research Insights
-
Majority of users listen to music from the background, while they keep busy with other things.
-
The cards are intrusive and distracting.
-
Fun facts can be easily missed.
-
Users still think Spotify is trying to test their knowledge.
-
75% of users are okay with Prototype A, provided it doesn’t happen very often.
-
Users were excited about Prototype B, and they were also able to seamlessly reach the goal without any problems.
Overall, about 80% of users leaned towards Prototype B - the genre categorization game.
ADDITIONAL/Final Iterations
.png)
background game notifs
Users get notified via their notification panel anytime the game comes on, which could be anytime between every 2-4 weeks (this way, business and user goals are in alignment and also achieved.).
With this method, we're also able to get to users who listen to music while doing other activities off or on their phone.
Fun Facts in Notification Panel
I also decided to include the fun facts as part of notifications that users get in their notification panel. With this feature, users are able to learn more about music, and this can in turn, contribute to the way they select genres in the categorization game.
Islamic Music Genres & Categorization
The image above shows a quote from another participant. It was pretty difficult to find Islamic/Arabic songs and playlists, and I realized it was an experience that had to be looked into and improved on Spotify moving forward. So I decided to design a few screens, just to show how they could be displayed and categorized.
Our Impact.
I gathered data alongside colleagues and compared it to our previous metrics. The results were as follows:
Playlist bounce rate reduced by:
42%
Increased user satisfaction by:
55%
Decreased song removal from recommendations by:
36%
Reflecting.
Working on this project, I learnt a lot about how really diversified users can be. While we are majorly focused on giving them the best experience, we have to be able to cater to their different needs while also making sure we're aligned with business goals.
Also, the game changer (pun intended) can be just a stroll or literally any relaxation activity away. I also saw the real value in team work. Most times, you really can't do it all by yourself. Getting opinions from different people and brainstorming with people your creative brain look up to is the ultimate game changer.
I do hope you enjoyed every part of this read as much as I did writing it :)
Interested in playing the genre categorization game prototype?
Click here
Go to top
.svg)
.png)
.jpg)
.jpg)
