Music royalties are generated in a many ways and from a growing number of platforms and can therefore be an attractive income stream for investors. For example, performance royalties can be generated when a song is played on:
●Elevator music services
●Internet radio/non-interactive streaming (e.g., Pandora)
●Online interactive streaming services (e.g., Apple Music or Spotify)
●Terrestrial radio (your favorite FM station)
●Television (the actual placement of a songs in TV, film, or commercials)
Additionally, visual media with which music is used can be synchronized and also generate royalties but these terms are usually negotiated. Examples of synchronization royalties are: ●TV shows
●Advertisements (web and TV)
●Films and trailers
●DVDs and Blu-Rays
Another revenue source for songs comes from mechanical royalties generated from the sale of:
●Record sales (vinyl, CDs)
●Online interactive streams (e.g., Spotify)
●Recorded cover songs
The administration and collection of royalties is a collaborative global effort with over 150 collection societies world wide. Each of these collection societies collects royalties from the above mentioned sources. As you can imagine, collecting the royalties is complicated but trying to analyze and forecast the value of each revenue stream can be even more complex. This is why Crescendo Royalty incorporates Machine Learning algorithms to evaluate songs.
As each revenue stream will decline or grow at a different rate it becomes important to determine what decline/growth rate each revenue stream has. Forecasting each revenue stream can be done by evaluating how the historical revenue stream has been trending but this is usually insufficient as the historical data for an individual song could be to short to accurately forecast the future revenue. However, when a individual song is compared to other songs with similar characteristics then a forecast becomes much more reliable.
Imagine trying to predict the future revenue of Drakes 2018 hit song "Gods plan". If you only have the historical data from that song it would be almost impossible to have a reasonably accurate forecast. Now imagine if you just took two other songs into consideration like the data from his 2012 hit song "The Motto" and his 2010 hit song "Whats My Name". All of a sudden it starts to become more possible to try and predict future revenue streams when you take into consideration 10 years of data from similar hit songs. A Machine learning algorithm can analyze similar songs along with many variables from multiple data sets.
In order for machine learning algorithms to be built, data needs to be accessible which is why Crescendo pulls data from over 7 different sources. The Machine learning algorithm then analyses these millions of data points. The variables that Crescendo Royalty analyzes when evaluating songs for an acquisition include:
1) Historical revenue for each source that generates revenue.
2) Number of platforms the song is played on (Youtube, Spotify, Pandora, etc).
3) Total number of streams since release
4) Age of song.
5) Number of albums released by artist to date.
6) Number of songs in a given album?
7) Number of featured artists in the album.
8) Number of songs released by the artist.
9) Album/song release cycle (how often do they release new stuff?)
10) Artist career length
11) Genre / musical characteristics
12) Does the artist have any tours scheduled?
13) Does the featured artist(s) have any tours scheduled?
14) Social media trends for artists.
15) Social media trends for song.
16) Historical stream counts since May 2019 for an artist, on Spotify.
17) Historical stream counts for the past 12 months for an artist, on Pandora Radio.
18) And more to try and forecast future royalty revenue with a reasonable degree of accuracy