In this composer's talk, we are going to examine the Spotify Popularity Index. It's very important because in order to get on Discover Weekly, the main algorithmic recommendation tool of Spotify, one has to have a high enough Popularity Index, as pointed out by Andrew Southworth. Nobody so far discovered the whole truth about it. But with the tools of mathematics and logic, I reverse-engineered it, figuring out the mathematical details of the algorithm that chooses the tracks to be recommended. I will present the formula that Spotify uses to compute the Popularity Index, so I will dismiss all the legends about it and confirm the few truths that were correct, and add crucial new insights. All my claims can be fact-checked, as all scientific theories.
UPDATE: My software for computing the Popularity Index is no more publicly available. Spotify every now and then modifies the Popularity Index, and I don't have the time to update the software. Fortunately, Spotify doesn't seem to change the algorithm, but only tweak its parameters, so everything I said, conceptually, remains valid. If you are nevertheless interested in the software, contact me.
