In my pilot study, I interviewed five Pakistanis about their general music consumption patterns and music discovery. It was done through convenience sampling, considering differences in ethnicity, language, age, literacy, tech-literacy, socio-economic status, and diasporic experiences. These interviews were done in Urdu and English. They included questions related to demographics to identify any underlying patterns. In addition to questions about the following factors:

• device usage
• mechanisms of music discovery
• mechanisms of sharing music
• average consumption time
• fluctuations in consumption
• streaming service usage and preferences
• changes in listening habits over time
• barriers in music discovery
• artists and genres they like/dislike
• possible guilty pleasures
• effects of concerts on discovery
• effects of religious, political and cultural factors
• understanding of music genres

After which I did a thematic analysis of these interviews to understand how Pakistanis consume music. Below I will summarize my major findings.

‍

Streaming services and social circles are major sources of music consumption

While the sources could be categorised into platform specific discoveries, we have social aspects in music consumption which can be seen by how friends, family and social media come into play in new song recommendations. Friends and family were separated as many of the recommendations came from in person interactions which were then translated to the digital realm. This includes sharing song links on Instagram or WhatsApp and creating playlists together on Spotify or YouTube.

Pakistani Dramas were separated from other multimedia because of their cultural impact particularly with two participants who were middle aged women. Additionally, Pakistan specific music programmes were mentioned by all participants, particularly Coke Studio, which they streamed on Spotify, YouTube or both. This was cited as their major source of finding newer artists, along with local artist collaborations.

‍

Frustrations with songs trending on social media

The infographic above also shows that the percentage of songs discovered through social media is generally small. Almost all participants mentioned the same song when they were asked what song they had heard recently despite belonging to different demographic groups and locations. They had discovered it through social media. However, participants went on to describe their dissatisfaction with songs discovered through social media. Younger interviewees mentioned that songs are being created to trend on Instagram or TikTok but they do not stick with them. One participant said:‍

Magar ye bhi hai na some gaanay they sound so good I read somewhere ke, it might be true, ke gaanay compatible banai jaatay hai to Instagram reels kyun ke quick content hai. And so they sound good, which attracts your attention to them. But the consistency is not there when you're creating songs again. Probably jiski wajah se I just hop artists and not loyally stick to one. Haan toh woh cheez thori si off lagti hai.

Translation: But I did read somewhere that some songs sound so good that they are made to be compatible with Instagram reels which is why they are quick content. They sound good which attracts your attention to them. But the consistency is not there when you're creating songs again. Probably why I just hop artists and not loyally stick to one. Yes, this is something I find slightly off-putting.

Another participant who does discover songs through social media trends such as GRWM (Get Ready With Me) videos, mentioned their exhaustion with songs trending on social media which makes them get sick of those artists and reduces their motivation to find new artists. They said:

And there's too many artists being overhyped nowadays. So, it's, it's really hard to find, like other artists other than ones that you already know quite well because everyone's talking with them. So, I think that's pretty exhausting to go through trying to find other artists.

Therefore, there is an oversaturation of certain artists on social media which also translates to streaming services which acts as a barrier to music discovery.

‍

Lack of meaningful music discovery

Every participant was asked about the amount of songs they listen to in a day and month. After which they were asked how many new songs they find in a month. This question was asked to separate unique instances of music discovery in context of their general listening habits.

In general, they discovered 2 songs per month. This could indicate the lack of stickiness of the new songs they hear on social media. Two of the younger participants who were mainly Spotify users mentioned using Smart Shuffle to find new songs. Smart Shuffle is an AI feature that adds songs to a playlist based on similarity of existing songs. But despite that their song discovery was small. Only two participants who were very active in their exploring songs quoted 14 and 200 songs per month. Therefore these findings demonstrate a need for meaningful music discovery.

‍

Artist collaborations and concerts were important the in discovery of local artists

Some participants mentioned going to a handful of local concerts in the last three years where they discovered other artists. Some of them were memorable enough for them to explore post-concert on social media and streaming services. They are loyal followers of those artists now. Regardless they acknowledged that they gained exposure to the local music scene, particularly the different types of genres, and it made them interested in finding out more.

Coke Studio season 14 was mentioned by one participant who was very resistant to listening to rap songs because rap "did not match my vibe at all". It clashed with his idea of the purity of classical songs and literature, despite some of the rappers using famous Urdu poetry. However, one of his favorite band vocalists collaborated with a rapper duo which made him gain some respect for them. In general, that season was a major push in making rap mainstream in Pakistan through collaborations with pop artists. Therefore, providing evidence of the power of collaborations in the local music scene.

The next step was collecting music data to explore artist collaborations and defining genres in the local context. Hamnawa is the only known organisation which has data about Pakistani music industry which made it invaluable to this project. Their database was used to tag artists with their genres. Spotify was used to collect data about tracks of individual musicians in order to account for all artist collaborations and understand their interconnectivity.

Spotify API was run on a Django server to collect the following information about each artist:

• Album name
• Track name
• URI
• Featured Artists

‍

Gephi, a graph and network data visualisation software, was used to understand the combined dataset. A brief exploration of 7 artists quickly showed the potential complexity of a larger network graph. Especially given that the Hamnawa dataset consists of 400+ artists. In order to reduce project scope, a suitable context was required.

For this purpose I chose Coke Studio Season 14 (2022) due to the popularity of that season. Coke Studio Pakistan is a TV program which features live studio recorded music performances by various artists. Anyone with an interest in Pakistani music would know of Coke Studio which established sufficient familiarity and interest. Moreover, new artists are often discovered through artists one is familiar with. Similar to the way smaller artists are discovered when they open for a more popular artist at a concert. Hence, all 27 artists featured in that season were selected.

Javascript libraries such as D3, Cola, and Cytoscape were used to find the optimal visualization. Attempts were made to combine libraries to borrow features from each library. However, that often led to unintended consequences and D3 generally had the best execution. Unlike the other libraries, D3 had a greater degree of interactability.

Usability testing was conducted with 20 users in total. Below are the interface iterations and results of testing.

Iteration #1 - Interactive Graph

Summary

1. D3 library was implemented to create an interactive graph
2. CSS styling was adjusted to increase readability
3. Different interactive states were accounted for such as when a node is dragged or unselected

Limitations

1. Displaying all of the artist names creates clutter
2. More importantly, the graph had to be supplemented with more data for it to be usable. Unless the user is a music enthusiast, they are unlikely to even remember the names of artists without a visual reference

‍

Iteration #2 - Embedded Tracks

Summary

1. Spotify Tracks were embedded such that whenever a node is selected the songs of that artist and the featured artists are shown
2. Only one song is shown per connection/link to avoid overwhelming the users
3. Graph initial state was added to establish context and reduce cognitive load. Only the names and tracks of Coke Studio Season 14 artists are shown when the graph is first displayed.
4. Context was established using heading, subheading and color key text

Limitations

1. Testing with 7 users revealed that people become fixated on the graph and its interactivity and do not explore songs
2. Project context has to be explained in some cases because people do not seem to be reading the content

‍

Iteration #3 - Improved Hierarchy & Copywriting

Summary

1. Rearranged the interface components such that users now interact with the entire interface
2. The copy of Spotify tracks was changed such that it now reads ‘Arists that {selected artist} has collaborated with’ instead of the prior ‘{selected artist}’s songs’

Limitations

1. A localized zoom feature might be needed because users seem to want to zoom in to the network graph

‍

Future Works

• Users mentioned different use cases for the network graph such as seeing the evolution of Pakistani music over time, adding their favorite artists to the graph or picking specific contexts such as Karachi rap. Highlighting the need for more customizability which can be catered to now that the proof of concept is complete.
• The genre key was believed to be an interactive feature by all participants. They believed they would be able to pick certain genres and dig deeper into them. Some of them do not know how to classify music so it could also teach them that. This finding leads me to believe that users will be more active in their music discovery if they are given the opportunity.
• Most participants started remembering other local artists upon seeing the graph. However they could not identify others as the artist image or music video was not present. Therefore, indicating that displaying track covers may not be enough to encourage recognition.
• Some of the participants mentioned culture specific music categories such as qawwalis and ghazals being missing. While these were categorized under the Traditional Desi genre for the sake of simplicity. This project needs a reclassification of genres and subgenres to fit the local context as too much context was being lost after abstraction.
• There were issues with artist/node labels including overlapping text which would require more spacing between the nodes. Some elderly participants had trouble clicking on nodes because of their size too. Therefore, pointing towards the ability to zoom into the graph and providing relevant affordances.

This project started in my Data Visualizaton (AD-609) course. However, I engaged with the owner of the Hamnawa dataset, Zeerak Ahmed, and kept him involved throughout the process. The overall feedback was positive and showed that this project achieved its goal not only in terms of usability, but also enjoyability. Below are some of the remarks I received during testing:

• “I can play with this all my life and not get bored”
• "I can imagine seeing this in an art museum or gallery"
• "You should send this to Spotify..."

The project was only deployed on a local server as it is under an NDA. In future, I hope to increase the scale of this project and make it open to the public. Going beyond the lab setting to gain a holistic understanding of the music experience is my next goal.

‍

‍