Relevance and personalization are key to reaching audiences. Brands often use content targeting as a tactic to align with an audience. Brand presence in a space that is native to the desired consumer is the ultimate relevance play.
We're excited to introduce Pandora’s Content Recommender. The Content Recommender lives within our proprietary audience insights tool, Audience Explorer, and utilizes a sophisticated data science algorithm to align brands with mixtapes that we know'll resonate with their target audience.
Pandora’s Music Genome Project personalizes every aspect of a user’s listening experience so no two artist stations are the same.
One listener’s “Beyonce Radio” will be worlds different than another listener’s. In knowing how different and personal music tastes are, how does a brand navigate through Pandora’s list of mixtapes to find sponsorship experiences that will resonate best with their customer?
Pandora data scientist, Monica Bhole, hand-crafter of the content recommender, sat with us for an interview to share her perspective:
With Pandora collecting around a billion data points every day, how did you decide where to begin?
Monica Bhole: Data scientists at Pandora have been working with these data points for quite some time now. We get listening signals from listeners and have a pretty good understanding of how we can use these signals to provide a great, personalized listening experience.
How does the content recommender work?
MB: We leverage machine learning models to recommend mixtapes at a listener level and then aggregate them to the segment level. This ensures that the stations we recommend are a great fit for specific listeners in each audience, instead of picking something that is simply an “average” for everyone.
Listeners music habits can change and are generally different from one another. How do you account for this?
MB: We do this a number of ways. The first step is that we make an effort to rank all of our mixtapes at the listener level. While listeners may not own these particular stations, we can get a good sense of whether they would like it based on the stations they do own. We try to focus on the stronger listener-station relationships, to ensure the listening experience is a good one. From here we aggregate these rankings to a segment level. By working in this way, advertisers are able to sponsor content for any combination of target audiences. Our recommendations can help provide good listening experience for many people in the target audience despite different user tastes.
By incorporating recent listening we can account for changes in listening patterns and preferences. For example, when listeners start to move toward holiday music in the winter we pick up on this and recommend holiday-oriented stations.
How do you validate the accuracy of recommendations?
MB: There are a couple ways to validate the accuracy. First, the algorithm for these recommendations is similar to what we do when recommending featured playlists to our premium listeners. Through testing, we have seen that those recommendations perform very well.
The second thing that we did was to compare our results to previous mixtape sponsorships. We compared listeners’ familiarity with the songs on the stations we recommend to their familiarity with the songs on the mixtapes that were actually sponsored for them. We found that overall listeners were much more familiar with the songs on the stations we recommend, indicating that these recommendations will provide an even better listening experience.
Why is this content recommender so revolutionary?
MB: First, the ad-segment recommendations stem from a personalized recommendation at the listener level. As a result, we can be more confident that listeners in the segment would like the station if they were to listen to it. This is just so much more exciting than picking a station that is simply “average” for everyone.
The second is that many of these stations are actually “fraternal twins” to featured playlists that we provide and recommend (in a similar fashion) to premium listeners. These stations are created by our amazing curators who are truly music experts and by sponsoring them, advertisers are bringing a unique, curated, experience to listeners that can only be found on Pandora.
And guess what? It works.
Our analysis shows listeners spent twice the time listening to songs on the mixtapes that were recommended for them versus a random mixtape, which can translate to twice the time spent with your brand.
Pandora uses data science to position brands for success. Take the guesswork out of content sponsorships and enjoy data-backed recommendations today. Get in touch!