Stitcher Aims To Build Talk Radio’s First Search Engine

With help from new transcription technologies, Stitcher is unlocking discovery features for audio content.

Stitcher Aims To Build Talk Radio’s First Search Engine

At one point, “binders of women” and “Mitt Romney’s tax return” were among the most discussed election-related topics on talk radio. With election day less than a week away, focus has switched to the U.S. national debt, the Benghazi attack, and the final presidential debate.


Those three topics now lead Stitcher‘s “trending” list, one component of a special election center feature that Stitcher launched in June. The list also shows which episodes of Stitcher’s 10,000 podcasts mention the trending topics and, if you wish, starts playing those episodes at the minute the relevant keywords are mentioned.

For now, trending topics are a feature buried within a temporary feature at the corner of the Stitcher app. But the technology behind them reveals the potential for discovery to impact talk radio the way it has music, video, and written news.

The key is real-time translation technology sold as a service, which allows Stitcher’s algorithms to work with audio clips as though they are text. It unlocks for talk radio some of the recommendation strategies used in other content categories.

“Only now, and never before, do we understand what’s being talked about in the content,” Stitcher Director of Product Colin Billings tells Fast Company.

Combine that understanding with years of data about user listening behavior and real-time trends from the web, and you have the ingredients for a recommendation engine–something that is, Billings argues, even more important in talk than it is for music.

“If we can say to a user, we know that you love Marc Maron, and we know he’s on another podcast, and he’s talking at minute 24, that shortcuts all of the problems in discovering that content,” he says. “Somebody doesn’t have to be watching for it, someone doesn’t have to read that episode description to determine that Mark Maron is in the podcast, someone doesn’t have to start listening and wait for him to come on, because we can tell them, hey, he starts at minute 24.”


Stitcher has already built recommendation features such as a “smart station,” which serves up content based on past listening behavior, and is making new additions with every release.

But it is also eyeing the search aspect of discovery. Just as Stitcher’s trending topic feature for the election allows users to compare several podcasts’ takes on a key issue and to pinpoint the place in each podcast where the discussion of that issue begins, the startup could provide specific audio results for any topic.

Rather than browsing podcasts, you’d be able to search their contents as easily as you search articles on Google News.

Instead of a passive listener, you’d become an active consumer.

“Instead of what was 10 years ago, which was an experience where people pressed one button on their dashboard and listened to whatever came out of the speaker, they now have on-demand,” Billings says. “Where we want to go in the future is that…you can actually find things that are specifically about what you’re interested. If you want to hear about what is going on with the hurricane, people can [eventually] think about searching for that and finding content that is immediately available–which is not something that many people did or even thought of earlier before these technologies existed.”

[Image: Flickr user Ben]

About the author

Sarah Kessler is a senior writer at Fast Company, where she writes about the on-demand/gig/sharing "economies" and the future of work.