Abstract
Digital delivery of songs has radically changed the way people can enjoy music, the sort of music available for listening, and the manner by which rights holders are compensated for their contributions to songs. Subscribers can enjoy an unlimited potpourri of songs and sounds, uniquely free of incremental acquisition or switching costs. This shift reveals listening patterns governed by affinity, boredom, attention budgets, etc. Listening patterns can be driven instantaneously, dynamically, organically or programmatically (playlists, for example). Listening demand is in a new paradigm, with a commensurate change in revenue implications. These new listening phenomena deprecate past orthodoxy around content curation in which a listener made a single purchase of a song. This point-of-sale model is now insufficient: demand revenue is proportional to song affinity—e.g., by how often a song is listened to within a time interval—and can be modeled as a time dependent process. We explore modeling digital on-demand demand and employ a fully Bayesian probabilistic model that: (1) yields estimators for multi-level effects on song demand and (2) naturally joins with multi-stage Linear Optimization scheme to optimize the same.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call