December 3, 2018
With over half-a-million podcasts today, discoverability is often difficult for consumers. To make the process easier, Pandora has come up with its algorithm-based Podcast Genome Project, first announced a year ago and just released in beta. Pandora rose to success with its initial Music Genome Project, and the Podcast version is similar, in that it uses more than 1,500 tags to create a recommendation engine. Also similar to the Music Genome Project, the Podcast Genome Project relies on humans as an adjunct to the algorithms.
According to Wired, Pandora chief executive Roger Lynch said the Podcast Project relies on machines more heavily than its music counterpart. “Machines can actually determine content, determine intent, there’s many more things they can determine about a podcast than they can about a song,” he explained. More specifically, Pandora “leans on natural language processing to parse the content of a given podcast episode, assigning descriptors related to content type, production style, the host profile, and lots more.”
If the user gives a thumbs-up on an episode, said chief product officer Chris Phillips, Pandora will present another, similar one. Currently, the Podcast Genome Project has “nearly four times as many descriptors” as the music version.
Humans play a role as “guardrails,” given that algorithms may not be able to discern satire or fact from fiction. But the “granularity” of the Podcast Genome also raises issues. A Questlove Supreme podcast dealt with Atlanta politics at the 50-minute mark, meaning a listener interested in that topic would have to “sit through nearly an hour of unrelated conversation, or scrub until they found the relevant discussion.”
Creating a recommendation engine for long-form content also presents challenges, since “unlike songs, many podcasts are a substantial time commitment,” meaning “you can only listen to so many in a day.”
“With a music recommendation, if you get one song wrong or one song is off slightly, it’s OK,” said podcast advertising network Midroll Media chief executive Erik Diehn. “Each song is only two or three minutes. You can give it a thumbs-up or thumbs-down, and fine-tune along the way. But when you’re making a choice about a podcast, the stakes are a little higher. You have a smaller set of samples to choose from as you try to figure out what somebody likes or doesn’t like.”
Addressing that issue, Lynch said, “the company is encouraging podcast producers to focus on shorter formats, bite-size morsels that represent less of a commitment, and clear the way for less intrusive advertising.” Pandora believes there is “even more potential in targeted advertising for podcasts,” despite inherent challenges.
“All the ads that we deliver on Pandora are targeted,” said Lynch. “If you compare that to podcasts, most podcasts are downloaded. There’s a measurement issue. Do you know if the podcast was even listened to? And how long was the podcast listened to? The answer to all of this is, you don’t know.”
In an earlier post, ETCentric described Pandora and Spotify’s entry into the streaming podcast market.