COLD-START PODCAST SUPPORT

OPPORTUNITY

Pandora was developing a new podcast streaming service with recommendations to customers, similar to how they present music. As the ability to predict consumers’ tastes is at the heart of Pandora’s brand, Pandora needed to launch the initiative with an initial predictive algorithm, even though usage data did not yet exist.   

METHODOLOGY

• Online survey among n>4,000 active Pandora customers determined preferences across podcast categories, styles and attributes. Then, known customer demographics and music preference data was appended to the podcast survey’s dataset.

• A predictive model was developed, using advanced analytics (i.e., canonical correlations), to determine podcast preferences from known customer demographic and music preferences.

IMPACT

DMG delivered a predictive algorithm for podcast preference that Pandora used to built a “genome” for all 4MM+ Pandora customers, driving suggestions of podcasts for the launch of their new streaming service.