A song that performs well on this measure is exceeding what we'd expect for the song given how much it's playlist. This tells me how many streams a song did compared to what I would've expected it to do if I just knew it's playlist reach. This gave me a measure I'm calling 'Streams Above Expected'. I then calculated how many streams the song actually got compared to this value. From this I calculated an expected number of streams for each song - simply 6% of the playlist reach for each song. I used the median of this rate for the top 20 songs as a typical conversion rate of playlisting into stream counts. I then calculated the rate of streams from playlist reach - weekly streams divided by average playlist reach. I averaged playlist reach for the Friday-Thursday period for each song. Just some quick notes so people can replicate this. I adjusted the weekly streams of each song based on this playlist reach to create a new version of the Top 20 chart that removes the bias of playlisting and just shows true popularity of songs. Playlist reach is the number of users being reached by playlists that the song is featured on (i.e 100m means 100m users follow playlists where the song can be found). I got the weekly chart of the most streamed songs in the US and using SongStats I calculated the average playlist reach for the week for each song. I had a little time this week and decided to fix Spotify's charts by adjusting for this playlisting. This isn't really a problem on Apple - Apple has some playlists of course, but usage of these playlists is far lower than on Spotify where millions of people follow playlists like Rap Caviar. This means 30% of Spotify's streams are decided by people other than the user - aka Labels. Somewhere around 30% of Spotify streams come from playlists rather than user selection.
0 kommentar(er)
