When it comes to streaming media, we’ve entered the era of the personalized playlist.
When streaming services first appeared, they typically fell into one of two broad categories. There were on-demand services, like Netflix, Spotify, and YouTube, in which the user chose what content to play next (with perhaps some guidance from a recommendation algorithm); and there were lean-backs services like Pandora and Rdio, which generated algorithmic playlists based on limited inputs from users. But streaming services today increasingly are looking to personalized playlists to differentiate themselves and maintain user engagement.
This week saw the launch of NFL Now, a new cross-platform streaming service that will deliver an always-on, personalized stream of news, highlights and live content to users based on the teams, games or even players they’re interested in (see here for a hands-on demonstration).
Although the service will be available on mobile platforms, NFL Now will emphasize long-form content, such as documentaries, live press conferences and locker room shows. Unlike on-demand game highlights, much of the content on NFL Now will be created specifically for the service and mixed into individual users’ feeds based on their profiles, making the experience more like a personalized TV channel than an on-demand video service.
“We will have as much live content as we can,” NFL Now GM Cory Mummery told Gigaom.
While the NFL is creating its own content, startup Pluto.tv, which launched in April, is creating playlists of content scraped from third-party platforms like YouTube, Vimeo and Daily Motion, using a combination of algorithms and human curation.
“We built our own meta player that can pull in other players so that the experience of watching content from different sources appears seamless to the user,” Pluto co-founder and CEO Tom Ryan told me.
Pluto currently offers over 100 curated channels organized by topic, such as Cats 24/7, Guns & Explosions, Women’s Fashion, Stand-Up Comedy, and various genres of music video. The lineup also includes some branded channels, such as a Funny or Die channel, QVC and GoPro.
The melding of machine learning and human curation to create personalized playlists is also at the core of Beats Music’s strategy, and was a major factor behind Apple’s $3 billion acquisition of the company.
“We think its the first subscription service that really got it right,” Apple CEO Tim Cook said at the time of the acquisition.
Since then, Google has acquired music “concierge service” Songza, which creates playlists based on a user’s mood, setting, activity, time of day and other contextual factors, and Amazon launched the heavily playlisted Prime Music service.
Microsoft, meanwhile, whose Xbox Music service has been something of an also-ran, recently opened up third-party APIs for the service to developers and launched an affiliate program. The program and APIs will allow developers to draw on Xbox Music’s 38 million song library and metadata to create new music apps and services, including, presumably, new approaches to curation.
The service providers all naturally hope that introducing a human element to the process of content selection will deepen users’ engagement with the service and lead to more time spent on the sites or in the apps. And at some point, they hope that deeper engagement will translate into easier monetization.
But human curation is also a labor intensive, and therefore expensive process. Ultimately, machines running algorithms are cheaper than humans exercising taste. The winner will be whomever figures out how to use machine learning to make human curation scalable enough so that user engagement grows faster than costs.
“With enough data machines can do a pretty good job at recommendations for some types of media, such as text,” Pluto’s Ryan said. “Music is sort of in between because music naturally has more of a human connection.” As for video, “we not sure what the right mix is yet.”