Introduction

Recently I came across an HN post and I was amazed to know that Spofity does provide our streaming data. It got me all worked up and I also wanted to see what my listening pattern looked like. What did I listen to the most? So without further wasting a millisecond I logged into my Spotify account and requested for it. You can request yours by heading over to : spotify's privacy page.
Note: To get lifetime of your data select the checkbox for Extended streaming History
Two days ago, I received this mail from spotify, that my requested data has been prepared and is ready to download.

The Mail I received two days ago

Data Overview

The Zipped file that I received contained following files, this is my streaming data of last year, the extended version of the data is still being prepared and will arrive soon. 

   -rw-r--r-- 1 dell 197609      81 Feb 16 14:38  Follow.json
   -rw-r--r-- 1 dell 197609      78 Feb 16 14:38  Identifiers.json
   -rw-r--r-- 1 dell 197609     148 Feb 16 14:38  Identity.json
   -rw-r--r-- 1 dell 197609   16094 Feb 16 14:38  Inferences.json
   -rw-r--r-- 1 dell 197609   82759 Feb 16 14:38  Marquee.json
   -rw-r--r-- 1 dell 197609       2 Feb 16 14:38  Payments.json
   -rw-r--r-- 1 dell 197609  200468 Feb 16 14:38  Playlist1.json
   -rw-r--r-- 1 dell 197609  672409 Feb 16 14:38  Read_Me_First.pdf
   -rw-r--r-- 1 dell 197609   51572 Feb 16 14:38  SearchQueries.json
   -rw-r--r-- 1 dell 197609 1420974 Feb 16 14:38  StreamingHistory_music_0.json
   -rw-r--r-- 1 dell 197609 1405227 Feb 16 14:38  StreamingHistory_music_1.json
   -rw-r--r-- 1 dell 197609  446850 Feb 16 14:38  StreamingHistory_music_2.json
   -rw-r--r-- 1 dell 197609   38384 Feb 16 14:38  StreamingHistory_podcast_0.json
   -rw-r--r-- 1 dell 197609     332 Feb 16 14:38  Userdata.json
   -rw-r--r-- 1 dell 197609   38854 Feb 16 14:38  YourLibrary.json

The files we'll be focusing are StreamingHistory_music to analyzing what songs and StreamingHistory_podcast for podcasts that I played and listened to.

  • In last one year, I precisely streamed 23063 songs.
  • Total time listening to the songs accumulated to total of 867.57 hours
  • I listened to 206 podcasts.
  • Total time listening to podcasts was about 72 hours.

    • Insights

    Songs

    After performing the anlysis on the data available, I found out the following things:

  • Daniel Pemberton became my most streamed artist after being played for 495 times, summing up to 1232 minutes.
  • Kedarnath by Pandvaas was my most played song.
  • Shaayad by Taba Chake was my most listened song.
  • Between 23:00 - 24:00 hours is the peak time during which I've spent the most time listening to songs
  • Highest listening time occurred during the 30th week when divided into 52 weeks
  • It was revealed that the highest number of listening hours occurred during the month of July
  • Total listening time maxed at 11th of the month.

    • Podcasts

  • Joe Rogan's podcast was my most listened podcast.
  • I tend to listen to the podcasts between 16:00 - 17:00 hours

    • Analysis

    Let's see some code and graphs to understand how I made these conclusions above, shall we? For the sake of the this analysis, I've used python for visualization on jupyter notebook. You can find the code in this tutorial in this repo

    I wanted to make sure about the time zone at which this data was recorded, so I contacted the spotify support. The team replied, the data is recorded in UTC timezone. To convert it into my local timezone i.e. IST, I had to add 5.5 hours to it.

    df['endTime'] = pd.to_datetime(df['endTime']) + pd.Timedelta(hours=5.5)

    Now we are in IST. Feels pretty good.

    Listening Time Distribution over time

                
    # Extract date from 'endTime' column
    df['date'] = df['endTime'].dt.date
    
    # Group the data by date and calculate total listening time for each date
    
    daily_listen_time = df.groupby('date')['msPlayed'].sum() / (1000 * 60 * 60)  
    # 1000 * 60 * 60 to convert ms to hours

    # Plot the time series
    plt.figure(figsize=(12, 6))
    daily_listen_time.plot(color='green')
    plt.xlabel('Date')
    plt.ylabel('Listening Time (Hours)')
    plt.title('Total Listening Time Over Time')
    plt.grid(True)
    plt.tight_layout()
    plt.show()

    My Top 10 Most Played Artists of All Time

            
artist_data = df.groupby('artistName').agg({'artistName': 'size', 'msPlayed': 'sum'})

artist_data.columns = ['play_count', 'total_listen_time_ms']

artist_data['total_mintutes_listened'] = artist_data['total_listen_time_ms']/(1000*60)

Most_played_artists = pd.DataFrame(artist_data.sort_values(by = 'play_count',ascending = False)['play_count'])

plt.figure(figsize = (15,5))
plt.barh(Most_played_artists.head(10).index,Most_played_artists.head(10).play_count,color = 'green')

plt.xlabel('Number of times played')
plt.ylabel('Artist Name')
plt.title('Top 10 Most Played Artists of all time')

for index, value in enumerate(Most_played_artists.head(10)['play_count']):
    plt.text(value, index, str(value))

plt.gca().invert_yaxis()  # Invert y-axis to display the artist with the highest play count at the top
plt.show()

    My Top 10 Most Listened Artists of All Time

            
artist_data = df.groupby('artistName').agg({'artistName': 'size', 'msPlayed': 'sum'})
artist_data.columns = ['play_count', 'total_listen_time_ms']
artist_data['total_mintutes_listened'] = round(artist_data['total_listen_time_ms']/(1000*60))
Most_Listened_artists = pd.DataFrame(artist_data.sort_values(by = 'total_mintutes_listened',ascending = False)['total_mintutes_listened'])

plt.figure(figsize = (15,5))
plt.barh(Most_Listened_artists.head(10).index,Most_Listened_artists.head(10).total_mintutes_listened,color = 'green')
plt.xlabel('Minutes Listened')
plt.ylabel('Artist Name')
plt.title('Top 10 Most Listened Artists of all time')

for index, value in enumerate(Most_Listened_artists.head(10)['total_mintutes_listened']):
    plt.text(value, index, str(value))

plt.gca().invert_yaxis()  # Invert y-axis to display the artist with the highest play count at the top
plt.show()

    My Top 10 Most Played Tracks of All Time

            
track_data = df.groupby('trackName').agg({'trackName': 'size', 'msPlayed': 'sum'})
track_data.columns = ['play_count', 'total_listen_time_ms']
track_data['total_mintutes_listened'] = artist_data['total_listen_time_ms']/(1000*60)
Most_played_tracks = pd.DataFrame(track_data.sort_values(by = 'play_count',ascending = False)['play_count'])
Most_played_tracks.index = map(lambda x: x.split('(')[0] , Most_played_tracks.index)

plt.figure(figsize = (15,5))
plt.barh(Most_played_tracks.head(10).index,Most_played_tracks.head(10).play_count,color = 'green')
plt.xlabel('Number of times played')
plt.ylabel('Track Name')
plt.title('Top 10 Most Played Tracks of all time')
for index, value in enumerate(Most_played_tracks.head(10)['play_count']):
    plt.text(value, index, str(value))

plt.gca().invert_yaxis()  # Invert y-axis to display the artist with the highest play count at the top
plt.show()

    My Top 10 Most Listened Tracks of All Time

            
track_data = df.groupby('trackName').agg({'trackName': 'size', 'msPlayed': 'sum'})
track_data.columns = ['play_count', 'total_listen_time_ms']
track_data['total_mintutes_listened'] = round(track_data['total_listen_time_ms']/(1000*60))
Most_played_tracks = pd.DataFrame(track_data.sort_values(by = 'total_mintutes_listened',ascending = False)['total_mintutes_listened'])
Most_played_tracks.index = map(lambda x: x.split('(')[0] , Most_played_tracks.index)

plt.figure(figsize = (15,5))
plt.barh(Most_played_tracks.head(10).index,Most_played_tracks.head(10).total_mintutes_listened,color = 'green')
plt.xlabel('Minutes Listened')
plt.ylabel('Track Name')
plt.title('Top 10 Most Listened Tracks of all time')
for index, value in enumerate(Most_played_tracks.head(10)['total_mintutes_listened']):
    plt.text(value, index, str(value))

plt.gca().invert_yaxis()  # Invert y-axis to display the artist with the highest play count at the top
plt.show()

    Hourly Listening Time

            
df['hour'] = df['endTime'].dt.hour
# Group the data by hour and calculate total listening time for each hour
hourly_listen_time = df.groupby('hour')['msPlayed'].sum() / (1000 * 60 * 60)  # Convert ms to hours

# Plot the results with adjusted x-axis ticks
plt.figure(figsize=(15, 6))
bars = plt.bar(hourly_listen_time.index, hourly_listen_time, width=0.8, color='green')

# Adjust x-axis ticks to be in between the numbers
plt.xticks(np.arange(25) - 0.5, range(25))

plt.xlabel('Hour of the Day')
plt.ylabel('Listening Time (Hours)')
plt.title('Listening Time Distribution by Hour of the Day')
plt.grid(False)
plt.show()

    Weekly Listening Pattern

            
def categorize_time(hour):
    if hour < 12:
        return 'Morning'
    else:
        return 'Night'

# Extract the week number from the 'endTime' column
df['week_number'] = df['endTime'].dt.isocalendar().week

# Extract the hour component from the 'endTime' column
df['hour'] = df['endTime'].dt.hour

# Categorize each listening session as morning or night
df['time_of_day'] = df['hour'].apply(categorize_time)

# Group the data by week number and time of day, and calculate total listening time for each category
weekly_listen_time_split = df.groupby(['week_number', 'time_of_day'])['msPlayed'].sum() / (1000 * 60 * 60)  # Convert ms to hours

# Plot the results
plt.figure(figsize=(15, 6))

# Plot morning listening time
plt.bar(weekly_listen_time_split.unstack().index, weekly_listen_time_split.unstack()['Morning'], color='orange', label='Morning')

# Plot night listening time on top of morning listening time
plt.bar(weekly_listen_time_split.unstack().index, weekly_listen_time_split.unstack()['Night'], bottom=weekly_listen_time_split.unstack()['Morning'], color='black', label='Night')

plt.xlabel('Week Number')
plt.ylabel('Listening Time (Hours)')
plt.title('Weekly Listening Time Distribution by Time of Day')
plt.grid(False)
plt.xticks(range(1, 53))
plt.legend()
plt.tight_layout()
plt.show()

    Monthly Listening Pattern

            
df['month'] = df['endTime'].dt.mon
# Group the data by month and time of day, and calculate total listenitime for each category
monthly_listen_time_split = df.groupby(['month', 'time_of_day['msPlayed'].sum() / (1000 * 60 * 60)  # Convert ms to hours

# Plot the results
plt.figure(figsize=(15, 6))

# Plot morning listening time
plt.bar(monthly_listen_time_split.unstack().indemonthly_listen_time_split.unstack()['Morning'], color='orangelabel='Morning')

# Plot night listening time on top of morning listening time
plt.bar(monthly_listen_time_split.unstack().indemonthly_listen_time_split.unstack()['Night'bottom=monthly_listen_time_split.unstack()['Morning'], color='blacklabel='Night')

plt.xlabel('Month')
plt.ylabel('Listening Time (Hours)')
plt.title('Monthly Listening Time Distribution by Time of Day')
plt.grid(False)
plt.xticks(range(1, 13))
plt.legend()
plt.tight_layout()
plt.show()

    Listening Pattern as per Date

            
# Extract the day of the month from the 'endTime' column
df['day_of_month'] = df['endTime'].dt.day

# Group the data by day of the month and time of day, and calculate total listening time for each category
daily_listen_time_split = df.groupby(['day_of_month', 'time_of_day'])['msPlayed'].sum() / (1000 * 60 * 60)  # Convert ms to hours

# Plot the results
plt.figure(figsize=(20, 8))

# Plot morning listening time
plt.bar(daily_listen_time_split.unstack().index, daily_listen_time_split.unstack()['Morning'], color='orange', label='Morning')

# Plot night listening time on top of morning listening time
plt.bar(daily_listen_time_split.unstack().index, daily_listen_time_split.unstack()['Night'], bottom=daily_listen_time_split.unstack()['Morning'], color='black', label='night')

plt.xlabel('Day of the Month')
plt.ylabel('Listening Time (Hours)')
plt.title('Daily Listening Time Distribution by Time of Day')
plt.grid(False)
plt.xticks(range(1, max(df['day_of_month']) + 1))
plt.legend()
plt.tight_layout()
plt.show()

    Most Listened Podcast

            
top_podcasts.columns = [ 'total_listen_time_ms']
top_podcasts['total_mintutes_listened'] = round(top_podcasts['total_listen_time_ms']/(1000*60))
plt.figure(figsize = (15,5))
plt.barh(top_podcasts.head(10).index,top_podcasts.head(10).total_mintutes_listened,color = 'green')
plt.xlabel('Listening time(in minutes)')
plt.ylabel('Podcast Name')
plt.title('Top 10 Most Listened Podcasts')
for index, value in enumerate(top_podcasts.head(10)['total_mintutes_listened']):
    plt.text(value, index, str(value))

plt.gca().invert_yaxis()  # Invert y-axis to display the artist with the highest play count at the top
plt.show()

    Peak Podcast Listening Hours

            
podcasts['hour'] = podcasts['endTime'].dt.hour
# Group the data by hour and calculate total listening time for each hour
hourly_listen_time = podcasts.groupby('hour')['msPlayed'].sum() / (1000 * 60 )  # Convert ms to minutes

# Plot the results with adjusted x-axis ticks
plt.figure(figsize=(15, 6))
bars = plt.bar(hourly_listen_time.index, hourly_listen_time, width=0.8, color='green')

# Adjust x-axis ticks to be in between the numbers
plt.xticks(np.arange(25) - 0.5, range(25))

plt.xlabel('Hour of the Day')
plt.ylabel('Listening Time (minutes)')
plt.title('Podcasts - Listening Time Distribution by Hour of the Day')
plt.grid(axis='y', linestyle='--', alpha=0.7)
plt.show()

    I know it is small and easy analysis but what it offeres is personalization as the data represented me, so I could see and verify how accurately the numbers can represent us. I recalled all the times I listened to those songs and podcasts. Although while working I tend to use youtube over Spotify. IMO everyone should try doing their own analysis and see through themselves.

    This is it for this article, see you next time.