Spotify wants African users to teach its alogorithm the different genres that are present and popular in Africa for a better experience.
.png)
The traditional music of Africa is rich and diverse, with different regions and nations of Africa having many distinct musical traditions. African musicians aim to express life from Afrobeats to Ndombolo which are played on the streets, in homes, clubs, party by tenagers and music lovers in different part of Africa. African music has evolved from record players to online streaming platforms. Most African either download the music to play on their phone or play via music apps
-min.png)
Spotify’s music genre classification extends majorly towards some countries and less for other like Africa. Though, popular music genre like Afrobeats and Alte are found easily on spotify but genre like Ndombolo and many others are either rare or not present at all.
As a Kimoyo Fellow, I was tasked to do an intensive research on African genre and subgenre of music and how African user can help teach the Spotify algorithm the genre of different African songs
-min.png)
To solve this, I had to come up with some research goals:
1. To discover and identify areas of success and challenges with genres from Spotify competitors
2. Identify the most popular African music genre and subgenre
3. Discover if users are having issues with the current in app experience
4. To identify solutions that could improve the product experience
Here are the methodologies I used to achieve this goals
.jpg)
I started off my research with installing Spotify and Youtube Music on my phone. This being my first time experencing what is like to stream music from this platforms, I wanted to put myself in the shoes of an existing user.
Then I Google search to know what African genres we have and what are the most popular to least. Then I compared the search result with everynoise. com to know the African genre Spotify already recoginses.
There are countries that their genres are well known by Spotify like South Africa, Ghana, Nigeria, Kenya, Zimbawe and others.
There some countries that have access to spotify but their genres are not known like Sechelles, Djibouti, Equatoria Guinea, Comoros, Chad. And not all the genre that are recognised on noise.com were available on spotify streamling app.
-min.png)
-min.png)
-min.png)
I also conduct competitive analysis with the platforms, I wanted to know how other Spotify competitors classify their music base on African genre.
After gathering information from my secondary research and getting myself knowledgeable about African genres and Spotify platform, I carried out 10 user interviews with my other fellows.
.png)
In other to get the best from all the insight, I converted all of my recorded user interviews and secondary research results into digital sticky notes using FigmaJam. After dropping all of my data onto sticky notes, I organized the information in search of repeated behaviours, pain points and similar trends of genres.
1. People find new music via recommendation and through searching for name of song or the artist.
2. People find a lot of new music on Youtube and prefers it because songs are more streamlined.
3. Playlists are defined by their “mood”, the word came up alot from different people
4. Recommend songs aren’t aligned with preference and mood.
5. Spotify African song mix are not well streamlined.
6. Similar trends like Afrobeats and African christain music were ranked as popular music that are mostly listened to
7. Playlist made for African music has too much genre variation.
8. Algorithm does not recommend African songs well.
.png)
From the trends, I curated a customer profile to better understand the customer I am creating value for based on customers’ jobs, pains and gains.
I synthesis the pain points into problem statement that could relate to finding a solution on Spotify algorithm and African genre.
.png)
Framing the problem statement into how might we, helped prepare my mind for ideation. I came up with several solutions, then I priotize my solutions using Impact-Effort scale.
.png)
During ideation, in order to empathize with the customer, I thought about how the customer would interact with the solutions, I started by sketching out the solutions on paper to get a better understand of how the solutions will play out.
Taking my major ideas to consideration, I created several low-fidelity prototype on paper, to sketch each flow of the ideas .
.png)
Then I converted the ideas from low-fidelity to mid fidelity so I can see how my ideas played out. I carried out usability testing with 12 participant for this solution. Each interview session with different iteration and 4 participant. I will be explaining my voted ideas with insight from the participant.
Solution1: Do a survey based on the genre users like
When a user select a the genres they want, it helps the algorithm to know the kind of genre the user likes and recommend songs based what was picked. It also makes the algorithm to know the popular genres in Africa. I carried out usability testing with 4 participant for this solution.
.png)
1. User wanted the choice to delete genre they don’t need.
2. User didn’t know the genre they picked are saved in the library they were confused at first
Solution2: Using voice input to ‘say the genre’
I wanted user to be able to say the genre of the song when asked what genre is this
1. User did not like the idea of Spotify recording their voice and there are too many vocals sound in Nigeria which the algorithm might get wrong.
Solution3: Adding gamification method with incentives
For this I came up with a different approach to add a ‘guess the genre’ game. When a user is listening to a particular song on Spotify a pop up with ‘guess the genre’ appears. The user can decide to pick a particular genre for the song or not. If the user select a genre, the user gets a point. When points accumulate over time the user can convert the point to money to pay for premium membership.
1. User wanted the “guess the genre” button to be easily seen.
2. User wanted to know from the beginning that there is incentive to guessing the genre.
3. Some user wanted to know if the genre they picked for a songs is right or wrong. They did not want the genre to just be added to the song.
Solution4: Playlist creation
I wanted user to state the genre of the playlist they want to create.
1. Users were not okay with this experience. Most of the participant say they are not sure of the kind of songs that would be added to a particular playlist.
2. Most often playlist consist of different genres.
Most participant preferred the gaming idea because of the points they would get and the choice of converting the point to money to pay for premium membership.
They also liked the fact that by selecting the genre of song, this would enable the algorithm recommend songs that are well streamlined.
After collating all paritcipant insight from the usability testing, I decided to go with the flow that solve the problem and the one that most user see has beneficial to them and Spotify which is:
1. Adding gamification method with incentives.
2. Carrying out a survey based on the genre users like.
"I think the point system is good, getting user to convert the point to money would get tem engaged".
Jeru Barnes
"With the guess the genre, lookin at the amount of people that will play this game, will have a biased set of genres. Think abou the choices and the way listeners will ener into the genres, how can that provide the best of genres from Africa".
Ashley Graham
Thinking of a way to come up with different solution on how to teach the algorithm different genres was a difficult one because even while you search Google you hardly get case studies on it. Based on the feedback from Ashley, will love to futher think of how best to solve this problem, so as to get better result.
Also adding a point system was kind of great for the users but I was having a problem with it affecting the business side of Spotify, I wanted to change it to something else like having the best user face and name on Spotify social platform or doing a leader board top user list. But I dug deeper into the point system we currently have in Nigeria (MTN services) after speaking to one of my pariticipant, I realise I could actually use the point system but make it one point per question until the user gets to 450 points which is the current Nigeria Spotify student plan.
Getting insights from participant that were not really familiar with spotify app and don’t stream music was difficult especially since Nigeria user have lots of pirated website to get music from. In other to get valuable insight I had to divert my questions from what I would ususlly ask a Spotify user, doing this gave me more ideas like stories from their childhood, how they came about the music they liked.
Related projects
