ETV Trinity ’07 Entrepreneur Interview :
Tim Westergren, Pandora.com
I hereby certify that the following piece of work complies with the Own Work
Declaration form already issued and with the University’s Rules and Regulations
relating to plagiarism and collusion, as listed in the Essential Information for
Students and the MBA Course Handbook.
Student ID: 8268
Tim Westergren is founder and CEO of Pandora.com, an internet based music 1recommendation and radio service. I will conduct a R.A.C.E. framework analysis of
Pandora chronologically, as R.A.C.E elements map well to clearly definable periods
of Pandora‟s history.
Pandora Timeline 1988 - Present
I have divided my report into 4 phases: (1) A Basis For Recognition, (2) Recognition,
Appropriation and Control, (3) B2B Flop and Exploitation, and (4) B2C Success and
Exploitation. Regardless of recent media attention, I treat Pandora‟s recent regulatory conflict as a sub component of it‟s B2C phase.
Despite travelling extensively for business and in support of a grass roots campaign to
reverse regulatory rulings by the US Copyright Royalty Board, Tim was kind enough
to take an hour to speak with me from his London hotel room. I wish to express my
gratitude to him, and to Michele Husak, Director of Communications at Pandora, for
their support in facilitating this interview. All quotations are taken from my phone
interview with Tim Westergren, 19.05.07, unless otherwise noted.
Technical and Business Model
How Pandora Works
Pandora is a website that allows users to create and listen to multiple personalized
internet radio stations. Users create stations by searching for songs or artists they like
in the Pandora database. If found, Pandora uses the Music Genome Project (hereafter
MGP) to find similar songs and artists, which are then played over the personalized
1 Recognize, Appropriate, Control, Exploit
station. As a user listens to a station, they are able to rate each song positively by
clicking a “thumbs up” icon, or negatively by clicking a “thumbs down” icon. This
data is then used to refine the list of songs from which the MGP chooses subsequent
songs for the user. In this manner, a user is able to find new music similar to their
existing tastes, gaining exposure to unknown artists and major recording stars
equally… an impossibility in traditional radio.
How MGP Works
The MGP relies on a genre specific “genome” of several hundred song characteristics
(genes) to classify a given song. Examples of genes would include Blues Roots,
Gravely Female Vocalist, Grunge Recording Qualities or Flat Out Funky Grooves.
Musical experts employed at the MGP listen to each song for 20 to 30 minutes to
determine each gene of the song on a scale of 1 to 5. The „genotypes‟ of songs a user
rates favourably are used to calculate a user preferences vector. This vector is used to
query the MGP for a ranked list of song recommendations.
How Pandora Makes Money
Pandora has two user based revenue streams: ad supported, and ad-free subscriptions
Phase 1: A Basis For Recognition (1988 – 2000)
The Pandora and MGP concepts ultimately arose out of Westergren‟s passion for
music and musicians, and his experience using a technical tool of the film composer‟s
Westergren‟s background in the field extends at the very least to 1988, when he
received his BA in Computer Acoustics and Recording Technology from Stanford
University. Following this, his career included stints as a musician, recording artist,
producer, and studio owner, providing him with experience in the field and a deep
respect for musicians. Later in this phase, he worked as a film composer. The
technical process by which film composers identify suitable music for a score
provided the basis for what would later become the MGP: iteratively swapping songs
(mixed tapes, CDs, online, etc.), getting feedback, picking new songs based on
relatedness, and swapping again. After several rounds of this, a composer is able
narrow down and “glean the tastes” of a director. The experiences of this phase gave him the background necessary to recognize the Pandora opportunity.
Phase 2: Recognition, Appropriation, and Control (2000)