Nieman Foundation at Harvard
What’s in a successful succession? Nonprofit news leaders on handing the reins to the next guard
ABOUT                    SUBSCRIBE
Nov. 13, 2018, 11:01 a.m.

Pandora wants to map the “podcast genome” so it can recommend your next favorite show

Plus: SNL pokes fun, Conan O’Brien tackles a new medium, and why we need more podcast transcripts.

Welcome to Hot Pod, a newsletter about podcasts. This is issue 185, published November 13, 2018.

Pandora’s Podcast Genome Project enters the wild. Sydney Pollack had a great line in Michael Clayton where he wags his finger at George Clooney’s down-in-the-dumps fixer protagonist saying: “Fer chrissakes, Michael, you’ve got something everybody wants! You have a niche!”

That line popped into my head when I first heard that Pandora was planning to graft its famed Music Genome Project onto the podcast universe. I mean…it makes sense. If the company was going to start properly distributing podcasts, this would be the way in. It’s great to have a niche, a thing only you have in the world. If you were born with a hammer for an arm, why wouldn’t you smash everything?

This morning, Pandora’s podcast offering, powered by the “Podcast Genome Project,” begins rolling out beta access to select listeners on mobile devices. Chances are you won’t see it: The feature will first appear to about 1 percent of users before expanding out over time. But it’s coming, and you can find the landing page here.

The beta rollout comes shortly after Pandora hired its first podcast chief, the lawyer Lindsay Bowen, formerly of Cowan, DeBaets, Abrahams & Sheppard LLP, and about two months after SiriusXM announced that it was going acquire the company. It also comes almost a full calendar year after Roger Lynch, who became Pandora’s CEO in September 2017, first signaled his intent for the streaming music platform to get seriously involved in the podcast ecosystem in an interview with Variety.

That intent doesn’t come out of the blue, of course. As some avid readers might remember, Pandora had deployed two significant experiments with spoken audio in the past: the first being a streaming partnership with Serial and This American Life and the second being an original production, the weekly podcast Questlove Supreme.

“The goal is to do something similar to what we’ve done for music over the years,” Chris Phillips, the company’s chief product officer, told me when we spoke last week. “To provide effortless discovery that’s also personalized.” Something similar, but not the same. Phillips tells me this isn’t a situation where existing technology is simply refashioned to fit a new context: “More like same concept, different tools.”

The Music Genome Project is a fairly well-documented Internet artifact, but to briefly explain the thing: It’s a technologically-facilitated effort to break songs down into their component elements and to then group them to create meaningful clusters of contiguous listening experiences — which allow for the discovery of similar new-to-you songs. Some of these elements are technical (tempo, musical families, etc.); others are more conceptual (genre, “female singer-songwriters,” etc.). Creating those clusters is a mostly automated process, but Pandora also houses a team of musicologists tasked with helping to form better epistemological rubrics. The team is “small but mighty” (or at least that’s how Phillips describes them) and the arrangement illustrates the marriage of technology and human expertise in pursuit of an edge in user experience. Very utopian, but also, concordant with a suspiciously familiar philosophy underpinning the enterprise: the belief that you can quantify and codify something previously considered subjective to unlock some higher level of achievement. (Alternate hed: Moneyball, but for music. Musicball?)

On the surface, the Podcast Genome Project exhibits much the same skeleton. Podcast episodes will be ingested, transcribed, and analyzed for taxonomical elements that are then grouped into discovery clusters. In our conversation, Phillips talked about things like content topics, themes, energy level. (Two examples given: “true crime” and “animated conversations about cooking.”) But the podcast skeleton won’t feature the same heart: There’s no spoken audio equivalent of a small but mighty musicology team pumping blood and wisdom through the whole system. Instead, there will be a more conventional group of human beings tasked with providing curatorial guidance and quality assurance. (So, antibodies.) That absence of an expert “podcastology” team — oh god I’m so sorry — is noteworthy, and I think probably crucial: It may well be the determining factor in whether Pandora’s podcast discovery technology will be revolutionary or merely additive.

Nonetheless, I fancy the notion. I remain unconvinced that podcasting has an existential discovery problem (once again, I’m more of the unpopular opinion that it has an existential marketing problem), but having another vibrant space to learn about new podcasts is undeniably a good thing for both publishers and listeners. Plus, for a podcast publisher, accessing potential new audiences on Pandora would theoretically require little more than plug-and-play. Which is, you know, pretty dope.

Of course, the True Promise is for the listener, and all the unique consumer adventures that an effective Podcast Genome Project can potentially create. The platonic ideal of the Pandora experience is something that sits firmly between the paralyzing freedoms of on-demand and the punishing captivity of linear radio. Take music, for example — something I enjoy tremendously but simply don’t have much bandwidth to do my own research. My life on Spotify is perhaps best described as a complete failure of imagination: the same playlists, the same albums, the Top 40 charts, over and over again. On the other hand, commercial radio is a straight-up hellscape: more ads than music, and when you do get to the tunes, it’s the same ten songs ramming your eardrums, because that’s how you make stars, right? The promise with Pandora is essentially better radio, featuring a scalable human-machine cyborg curator instead of the more specific hit-or-miss taste of a mortal DJ. Or worse, the capitalist imperatives of the music industrial complex.

Forgive the fan fiction, but: In theory, with Pandora, you’d have a situation where a listening session of, say, The Woj Pod leads me to efficiently surface and sample other (weirder) hoops podcasts like Horse and Buckets — ones that I’d otherwise have to plumb the murky depths of Google search results to learn about. (As an aside, one potential metric for discovery gambits like this would be its ability to elevate “weird podcasting.” Note to self: Revisit this idea later.) Upon discovery, I’d be in a position to engage in two follow-ups: First, I can add them to my “collections” — which is Pandora’s way of functioning like a straightforward podcast app — and second, I can give it a little thumbs up to help Pandora learn more about the stuff I like. Table-stakes stuff, really, especially in the age of tech companies knowing more about me than my mother does. But given that podcasting is still an Internet mosquito preserved in amber in so many ways, it’ll be cool to see those standard tools applied to podcasts at scale.

All right, publishers, let’s talk brass tacks. There are a few important things to note.

First of all, there is some trickiness around what will be included in Pandora’s podcast offering, at least for now. The product enters public beta with a series of launch partners: APM, Gimlet, HeadGum, Maximum Fun, NPR, Parcast, PRX+PRI, reVolver, Slate, The New York Times, The Ramsey Network, The Ringer, WNYC Studios, Wondery, and Libsyn, plus This American Life and Serial. Which is to say, Pandora isn’t beginning with an open platform, and inclusion depends on a series of discussions and negotiations. For now, to be distributed through the platform, you need to either be part of the aforementioned list of publishers or be hosted on Libsyn. A spokesperson told me that not all content partners available on Pandora are listed in the press release, and that it’s still a dynamic process at the beginning of the beta launch. Definitely expect more inclusions over time, but for now, I’d check with my hosting provider to see what’s up if I were you.

Another front to watch: monetization. Pandora’s podcast product enters public beta without any advertising tools, which are still being developed with the intent of rolling out sometime next year. How would podcast advertising on Pandora work? The details are still being worked out, I’m told, but Phillips discussed a potential scenario where it comes down to whether a publisher has an advertising deal with the company. In this hypothetical future, if a publisher does have a deal with Pandora, then the platform will strip the midroll ads baked into their episodes and swap it out with whatever podcast advertising experience the company comes up with. (The company is currently building the necessary tools to allow for those strip-and-swaps.) If not, those midrolls will be preserved. Again, this is how advertising might work, and Phillips notes that the company is in close contact with various podcast companies to figure out the best way to execute these relationships.

Publishers will also get enhanced analytics of whatever listens happen on Pandora, including episode completions, audience demographics, and so on. Again, table-stakes stuff, mosquito in amber, etc. etc.

For good measure, I asked about how the platform will handle content policing. (I was thinking, specifically, of The Alex Jones Problem that popped up over the summer.) Phillips acknowledged that it’s a tricky issue. “We try to be thoughtful and balanced,” he said, when it comes to the broader issues of censorship and policies. They do, however, have strong policies around hate speech.

Finally: original content. I’m told that there are no immediate plans for more original Pandora podcasts beyond Questlove Supreme…for now. “Watch this space,” Phillips said when I raised the question. Which, you know, sounds like they’re definitely going to do more stuff at some point in the future. I mean, come on: Spotify’s doing it, Google did it at one point, iHeartMedia literally bought a whole podcast company to keep doing it.

So will Pandora’s Podcast Genome Project end up being a significant boost for the podcast industry? Or will be slow on the take, as Spotify was? Obviously, I have no idea. More broadly, it’s been a long time since I bought stock in any flashy narratives about new “inflection points” for podcasting. Every time I see something that makes a little voice in my head go “this could be big,” I try to take the voice out back and bury it beneath the shed. (It’s just good practice.)

Still, there’s something about Pandora’s Podcast Genome Project that strikes me as particularly interesting, if only because its approach seems genuinely untested within the context of podcasting. As such, it wouldn’t surprise me if this was the thing that could well bring podcasting to a new place, even if I won’t buy into the possibility right this second. But even if it doesn’t, I won’t be blaming Pandora for under-cooking the pursuit. They have a niche, a place in the world. And they’re leaning into it.


  • Conan O’Brien’s podcast, Conan O’Brien Needs a Friend, launches next Monday, and the production has circulated a star-studded guest list preview. Stitcher (née Midroll Media) is handling ad sales, and there’s an interesting connection there: Adam Sachs, the former CEO of Midroll, is now the COO of Team Coco, O’Brien’s production company.
  • ICYMI: “Some podcasters are reporting that they are seeing a steep decline in listenership, specifically when using Apple’s Podcast Analytics tool,” per 9to5Mac. Apple is reportedly on it.
  • The New Yorker published a pretty chunky piece on podcasts yesterday (titled “How Podcasts Became a Seductive — and Sometimes Slippery — Mode of Storytelling,” or “Binge Listen” in print). It’s a strange one, in that it could be read as a limited analysis of the kinds of shows New Yorker staffers typically listen to. But it also feels emotionally true, especially in its findings of entrepreneurial power (public radio “vow of poverty” be damned) and potential narrative pitfalls. The ending is superb.
  • I don’t think I’ve mentioned the actual date yet, but: Post Reports, The Washington Post’s daily news podcast, officially drops December 3. Also, as host Martine Powers noted on Twitter, the team will be “managed, hosted, and majority-produced by women of color.”
  • Found this Vulture interview with Homecoming creators Eli Horowitz and Micah Bloomberg, on how the podcast was adapted to television, pretty interesting. Of note: The second season and the TV adaptation were written at the same time, and there will be no third season for the podcast. One swapped out for the other, it seems.
  • Surviving Y2K, Dan Taberski’s followup to Missing Richard Simmons, dropped today. It’s really good.
  • WNYC Studios tells me the new Jason Reitman film, The Front Runner, about Gary Hart’s 1987 campaign for Democratic presidential nomination that was taken down by a scandal, was apparently directly inspired by a Radiolab episode, “I Don’t Have To Answer That.” Wild.
  • Who Weekly’s Lindsay Weber and Bobby Finger with the goods: “The Safe Space of the Celebrity-on-Celebrity Podcast.” Also, whoever gave Weber and Finger a column at Slate should be given a raise.
  • You probably saw this, but in case you didn’t, here you go. Think it could’ve been a hundred times better? It’s SNL.

The case for transcription [by Caroline Crampton]. One thing about producing audio for the Internet that a lot of people I talk to really like is the relative tranquility of the whole experience, especially when contrasted with all other forms of online publishing. There’s no comments section, and audiences generally have to explicitly listen to a whole show to hear what’s actually said, rather than being able to Ctrl+F their way to outrage. As one female political podcast host told me, the comparative isolation really “cuts down on the number of drive-by haters.” (The flipside of this isolation, one might say, is podcasting’s infamous problems of shareability and discoverability. You can’t have everything, I suppose.)

But there are plenty of good reasons for expanding the publication of podcast transcripts. Chief among them is accessibility: having text alongside audio means that people who are hard of hearing or with auditory processing issues can still enjoy the show, as can those who aren’t fluent in the original broadcast language. (The regular English translations are how I enjoy the Spanish-language NPR podcast Radio Ambulante, for instance.) Increasing the accessibility of podcasting to these groups is innately the right to do; it’s a moral need as much as it is an audience growth tenet. And then there are the secondary considerations: notably, SEO and archiving. Plus it’s just good practice to preserve your stuff in several forms, as I’m sure the good people at the Preserve This Podcast project would tell you.

For some, transcriptions are baked into the DNA of the podcast’s goal. Amanda McLoughlin and Eric Silver, co-hosts of the Join the Party podcast and founding members of the podcast collective Multitude, told me that transcribing is foundational to their show. “In pre-production for Join the Party, we were designing the show to be more accessible than any other Dungeons & Dragons podcast out there — or any podcast in general,” they said. “We knew we had to have transcripts from Episode 1 to make our episodes accessible to people of all abilities, processing styles, and language backgrounds.” The transcripts were part of a broader effort to open up their subject (in this case, the playing of Dungeons & Dragons) to more people, no matter their background or subject knowledge. To this end, in addition to the transcriptions, they also purposefully introduced LGBTQ+ characters early on, cut down on the pop culture in-jokes, and released versions of their episodes that explained how to play.

However, producing accompanying transcripts isn’t yet a widespread practice in the podcast industry. Bigger publishers mostly don’t provide, or provide them sparingly. Sometimes, if you search long enough, it’s possible to find a listener who has done it for themselves (e.g. these transcripts for Dirty John) — but the majority of podcasts are simply left untranscribed. There’s an obvious reason: Producing transcriptions can be expensive and time consuming. I’ve seen rates of around £1.10 per minute ($1.50) for human transcription services; there are also various paid-for automated options that cost less, although they can be far less reliable. Even formatting and posting the final transcript can be onerous for already overworked producers.

The amount of work involved in producing useful transcripts also depends on the type of show. For narrative formats, it’s theoretically fairly easy to assemble a transcript that effectively serves as proper representation of the final episode, whereas for conversational podcasts with several speakers — in which the enjoyment occasionally comes from its chaotic cross-talk — it can take a lot longer. I’ve found these huge disparities in the show I’ve worked on. For Shedunnit, where the episodes are scripted and sub-20 minutes and the interviews are already transcribed for editing purposes, it’s a matter of a couple of hours — whereas for some of the roundtable shows I produce, creating a complete transcript can take all day, or even longer if I’m adding footnotes and links. When I’m in the middle of trying to untangle all the overtalking and track down precise sources, I can see why there are those who would rather not bother.

McLoughlin and Silver argue that just time and labor resource intensiveness is no reason to skimp on transcribing podcasts. “Accessibility is a right, not a privilege. Making great transcripts involves extra work for podcasters, but doing it demonstrates that you truly care about all of your listeners,” they said. They also pointed me towards this guide they published with the Bello Collective, which details some cheap and even free ways of doing it. (Private YouTube videos are great, the automatic caption service does most of the work for you!)

Counterintuitively, they posited, it’s actually the relative ease of taking a podcast from idea to final episode that has stopped transcripts going mainstream. “We all love that podcasting has such a low bar of entry for creators…Transcripts and other features that can improve accessibility, discovery, and potential audience growth are left by the wayside because they’re not required by your podcast host to launch a show.” They also feel strongly that transcripts are not to be released as paid-for extras to Patreon supporters or similar, writing in the aforementioned guide: “Accessibility should never depend on a listener’s income.”

Personally, I find podcast transcripts a great aid to memory — I frequently use them to track down something I half-remember from an episode before I pass it on to a friend. I think McLoughlin and Silver are right about the accessibility argument, too. Just as it is standard for TV to offer subtitles and cinemas to have signed screenings, maybe it’s about time the podcast industry took responsibility too.

Image by Hyper Glu used under a Creative Commons license.

POSTED     Nov. 13, 2018, 11:01 a.m.
Join the 60,000 who get the freshest future-of-journalism news in our daily email.
What’s in a successful succession? Nonprofit news leaders on handing the reins to the next guard
“Any organization that is dependent on having a founder around is inherently unsustainable.”
Worldwide, news publishers face a “platform reset”
Some findings from RISJ’s 2024 Digital News Report.
The strange history of white journalists trying to “become” Black
“To believe that the richness of Black identity can be understood through a temporary costume trivializes the lifelong trauma of racism. It turns the complexity of Black life into a stunt.”