Podcasting 2020 – The Rise of the Machines
December 5, 2019 · By Wil Harris · 6.4 minutes to read
For all its merits as a rapidly evolving digital medium, it’s possible to argue that podcasting has been an almost contrarily analogue format to date. As Publishers, lack of detailed engagement metrics mean we measure our audiences’ listening behaviour with surveys and studies, like print or television, rather than with clicks and impressions, like more mainstream digital media. As Consumers, new content recommendations are almost entirely driven by peer-to-peer word of mouth or chart activity, like the radio days of old. And as Advertisers, while Google and Facebook have driven innovation across video and social with ever more sophisticated ways of buying audiences with innovative new advertising creatives, the host-read-straight-into-the-show podcast format has not only persisted but, arguably, thrived in its isolation over the past year.
But if 2019 was the year podcasting finally went mainstream (question for the reader to ponder - was it?) then 2020 looks like it might be the year it finally gets smart. Like, robot smart.
If, As Marc Andreesen famously posited, “Software is going to eat the world”, then a combination of algorithms, data, AI and the applied innovation of these technologies is going to start to nibble away at all elements of the podcasting experience, and everyone sat around the dinner table is going to have to start eating or be eaten.
The Consumer Experience
For so long dominated by the Apple Podcasts app - which, while sporting a facelift, hasn’t seen any significant innovation since its inception over a decade ago - the consumer experience of podcasting is changing and becoming far more akin to other forms of online audio than it has been previously. For this we have to thank the rapid expansion of 4G infrastructure, compounded by the music industry’s wholesale move into subscription models, for training users to stream their choice of audio, rather than download it.
Downloaded audio, the legacy podcast paradigm, has traditionally been a black box of consumption behaviour, with occasional insights delivered by whatever titbits Apple deemed to give. But with more and more podcasts being consumed via on-demand stream, rather than ahead-of-time download, the opportunities to collect and utilise data to deliver a better experience start to open up.
Take Spotify’s latest experiment: Daily Playlists for podcasts. Data shows that contextual playlists for music on Spotify - ie those driven by deep usage data - make up over a third of all listens . The ability to generate contextual playlists for podcasts on the fly has the possibility to transform how we discover new shows to listen to, and move away from the more traditional curation models (that’s certainly what Spotify seems to be hoping).
Google has already started rolling out in-podcast search, and aims to bring this to a wider set of shows in 2020. With Google’s in-house voice-to-text automation making hosts’ voices as searchable as the traditional written blog, finding a key moment and being able to jump straight to it could open up a wave of casual drive-by listening that has so far eluded podcasting. ‘Casual listening’ and ‘podcasting’ have not historically been comfy bedfellows, but this kind of ‘light touch’ consumption but has been crucial to the survival of other online media entities – even as its popularity has disrupted traditional publishing models.
Even Apple – which for so long has sat on a treasure trove of data and done nothing with it - is starting to get in the game. Its recent low-key release of host tracking within its Podcasts app shows that the manipulation of metadata present the opportunity to follow people and topics, not just shows and feeds - something that the Twitter hashtag has crucially provided for years.
I would be remiss if I didn’t mention Entale’s own efforts in bringing data and AI to the podcast experience. Afte launching our platform last year which allowed podcasters to increase engagement with their content by manually adding supplementary information to their audio, be that images, maps or articles ,we are now highly focused on taking those 'curated’ experiences and replicating them in a machine learning environment. By doing so, we can surface and display relevant, entertaining or insightful extra content to any podcast available, in the open ecosystem which brings with it greater opportunities for storytelling, discovery & monetisation.
Of course, more data and more automation means more ads. Much has been made of the fact the podcast advertising market is growing rapidly, but even if it hits the $1bn in revenue predicted for next year, it will arguably still be under-monetised compared to other formats, with the IAB estimating that while podcasting accounts for 40% of the listening for online audio overall, it brings in less than 15% of the ad revenue.
Innovation in advertising aims to close this gap, and is predicated on having better, and more, automated and data driven systems to increase effectiveness. Acast has been particularly forward-looking in this space - not only was it one of the first providers to let advertisers buy audio programmatically, it has also experimented with personalised audio within that format, through its partnership with A Million Ads, which released its first podcast product earlier this year. DAX, the digital audio arm of UK commercial radio giant Global Radio, has data partnerships with a clutch of providers in order to target users with its ListenerID technology. With data and information at the heart of their strategy, it’s no wonder that both these companies are growing their podcasting revenues at an outsized rate within the market.
Even the host-read ad isn’t immune from machine-driven innovation, as Nick Quah wrote about recently. Attempts to scale the host-read ad rely on marketplace automation, and the programmatic insertion and rotation of host-read ads through 'traditional’ ad-servers - in effect making the host-read just another creative spot to be targeted and rotated based on listener data – is rapidly becoming popular with podcast hosts like ART19.
Of course, more advertising means more opportunities for ad-blocking. We’ve already seen the first tech that claims to be able to detect and skip ads, using Shazam-esque analysis of audio data. As more and more audio moves to streaming, how long before we start to see an arms race between unskippable ads (a la YouTube) and technology built into apps or browsers to combat it?
Even the decidedly analogue process of podcast creation will start to feel the touch of technology next year. Already we are seeing innovation in this area - Descript will use AI to allow you to re-edit and even overdub a piece you’ve already recorded using some stunning AI to synthesise and blend your voice. And that’s without even getting into text-to-speech synthesis - companies such as iSpeech may only be in the early stages of delivering a programme that is listenable, but the rise of short form and atomised content (created for the stackable morning briefings that both Google and Amazon are implementing for their smart speakers mean that a large chunk of purely informational consumption could be created purely by AI.
As we fret on a global scale about the role that microtargeting, user data and automated content creation is having in our society at large, it seems odd to contemplate that a format so far mostly devoid of it needs more. But it’s clear from just about every survey of consumer trends that podcasts are only increasing in popularity and are sure to continue to do so for the foreseeable future, and supporting that requires not just experiences that can scale beyond the hand-curated, but data and monetisation that can support that growth.
I think it’s increasingly important that we realise, as an industry, that the old adage of “what got you here won’t get you there” is going to apply with increasing urgency. As an industry, it’s surely crucial that we embrace and adopt the changes coming to the world of podcasting as a key way to get better experiences into the hands of more people. I, for one, welcome our new robot overlords.
—Wil Harris is the CEO of podcasting platform Entale Media, founded the YouTube multi-channel network Channel Flip, and has also worked for Condé Nast. He has appeared on many podcasts including This Week in Tech. He's based in London, UK.