In a new interview with the Wall Street Journal, Pandora Chief Scientist/VP of Playlists Eric Bieschke says that the popular digital music service has refined its algorithms enough over the past decade to see a notable effect on listening hours. Due to user interaction, such as ratings songs with a thumbs up or thumbs down, and requests to skip songs, Pandora has increasingly been able to attract listening to tune in longer.
Bieschke also noted that the entire process is not automated, as every song in its library is coded by its employees. He told the WSJ that "humans understand music far better than machines do. That could change some day, but today, if you want to know if a song makes you laugh or dance or cry, you need to ask a human."
Pandora has made "measurable gains in the amount people listen based on improvements in algorithms… Percentage point increases year after year after year," Bieschke told the paper, adding that the company is still working on improving the relevance of its ads to users. "We’re just getting to the ad personalization experience."
Currently, the average Pandora user listens to 19 hours of music per month, and Bieschke revealed the company is also trying to determine when people are listening to the service in groups, such as in the car or at parties. This information would also be used to better target ads.