My SiriusXM tracker for ‘The Heat’ radio station has now been running since the summer of 2016 and has collected a good amount of data. It seemed like a great opportunity to make use of this data for getting some insight into what songs/artists/albums have been popular throughout the months, and also get some Spark practice under my belt.

For more information regarding the tracker you can see my previous blog post or view the live web client here. There is also a clone tracking The Highway station. The source code for following post is available at the end.

Data Quality

Before a song is stored in the database it is ran through some filters to ensure that only real songs are kept, and things like promos, different channel advertisements, DJ Mixes, etc. are discarded. A song that passes the filters then is checked against the Spotify API, and if it exists there the Spotify metadata is recorded as well. We can say that the stored songs, which also have Spotify data can be regarded as good. Nothing is perfect of course, and sometimes songs are not found on Spotify because of different spellings between SiriusXM and Spotify or some bad data has slipped through the filters.

From all the songs tracked, 97.27% have Spotify metadata. From the remaining 2.73% that do not, the top 3 played songs among them are by Beyoncé. Since Beyoncé’s music is not on Spotify, we can safely assume that those 3 songs can also be counted as good. Bringing the amount of good data to 99.12%. Looking very good!

I remember there was a brief period of time in which the tracker had gone down due to a change in the Spotify API policy, now requiring access tokens. Looking at the tracked songs per day we easily can see the gap, between May 29, 2017 and June 11 2017. Other than, that it looks like smooth sailing and there is consistent data throughout the recorded timeframe.

Overview

• First tracked song: Nothing Is Promised by Mike WiLL Made-It, Rihanna on June 20th, 2016.

• Last tracked song: rockstar by Post Malone, 21 Savage on January 8th, 2017.

• Total songs: 200,871

• Total different songs played: 1,508

• Total artists played: 1,041

Top Songs

Most Played:

Most Played per Month:

Most Played Song in a Single Day:

Top Albums

#1 4501 times.

#2 3745 times.

#3 3717 times.

#4 3647 times.

#5 3576 times.

Source Code

The source code and Spark project for the analysis above can be found on my GitHub.

Spark on!