Google’s acquisition of Songza is widely regarded as a buy-in of Songza’s unique “Concierge” delivery of personalized music, based on each user’s activity and mood combined with other data signals. The backbone of Songza programming is assembled by a team of 60 music specialists, and their work is supplemented by a Big Data-fed algorithm which refines the output based on the listener’s history of song votes and song skips.
In Songza, two people might select the Mainstream Indie station at the same time, as a soundtrack for their desk work on a Wednesday morning, and receive two different listening experiences. Each stream would be essentially informed by the human specialists (“curators”) who assign songs to that station, and refined by each user’s listening history — plus other differing signals like the local weather. (Songza, uniquely among music services, has an arrangements with The Weather Channel to weave that information signal into the song-picking algorithm.)
Songza’s formula, and those of other music services, which determine what a listener hears, all fall under the broad term “curation.” Unlike a museum curation, which presents one experience for all visitors, music-service curation is mostly a technology achievement, and depends on user behavior to deliver user-specific results.
Curation strategies
All music services that provide some degree of interactivity deploy a curation strategy. The underpinning of that strategy is subject to trend and vogue. The early trend of data-driven music algorithms remains an essential component of customized listening, but has lost some marketing luster. In 2014, the industry meme is focused on bringing human taste and intelligence into the formula.
- Beats Music, more than any other service, has fashioned its marketing on the appeal of celebrity human curators programming the playlists. Browsing the service, you encounter listening experiences that are more like branded programs than playlists. Beats launched in January, has about 250,000 subscribers, and was snapped up by Apple in May.
- 8tracks, a profitable listening service with over 6.5-million active listeners, which was reportedly a subject of Google’s acquisition interest, is all-human in its programming. The 8tracks mixtape model encourages users to upload music and assemble playlists, tag them for sharing, and build personal audiences. (See RAIN’s interview with founder David Porter on music curation.)
- Slacker Radio devotes resources to create human-assembled playlists, some of them hosted with pre-recorded voice breaks between songs, DJ-style. Jack Isquith, SVP and head of content programming, told us that those streams out-perform the more impersonal playlists: “We’re making a big bet that you really need hosts and the human voice, the sense of humanity actually in the listening product. We have a new country station. Adding a host who is heard every couple of songs with just a bit of context raised listening by 20 percent.”
- Pandora is committed to a unique approach anchored by its Music Genome Project, in which teams of people analyze musical attributes of songs to create an unusually granular catalog system. When combined with user behavior (song voting and song skipping), the platform seeks to provide a steady stream of perfectly personalized music. Despite the intense reliance on human effort, Pandora is essentially a technology company. It leverages human analysis, not human taste-making.
The celestial jukebox
The arguable downside to the Pandora’s Genome is the slow build: Pandora’s music collection is less than two-million tracks after years of development. Spotify, Rdio, and other “celestial jukeboxes” embody more than 10 times that number. From Pandora’s perspective, it is trading music quantity for curation quality.
It is that “celestial jukebox” model, intended to eventually provide a comprehensive catalog of recorded music, which has given rise to algorithms as guiding forces in music programming. Spotify, Rdio, and Rhapsody are good examples. All three built their curation models on technology provided by The Echo Nest, the market-dominating provider of music intelligence to online services. The Echo Nest reverses Songza’s model of human-first, technology-second curation. Essentially a data-crunching company whose work is supplemented by music experts, The Echo Nest was acquired by Spotify in March. (Rdio and Rhapsody bailed out of their respective partnerships.)
The beating heart
Interestingly, Google did not (reportedly) open merger talks with Pandora or Spotify, the two leading music services in the U.S. (iHeartRadio, which enjoys high brand recognition second to Pandora, is the public-facing brand of Clear Channel, and presumably off the table.) Google can afford anything, and already operates a music store (Google Play Music), a celestial jukebox service (All Access), and a gigantic crowdsourced media platform (YouTube). Google is preparing an all-music version of YouTube for launch this summer.
With all these guns locked and loaded, Google focused its shopping trip on relatively small Songza and 8tracks, two of the most interesting music-curation plays in the industry. Google’s All Access platform is Google-y through and through, algorithmic and impersonal. Songza’s emphasis on context around music choice — understanding what the listener is doing and feeling — could be a warm beating heart in the steely Google infrastructure.
Tuning the formula
Google is not the only outlet trying to get a piece of the music-context action, Concierge-style. Rhapsody, iHeartMusic, and Beats Music have all borrowed the concept, and offer playlists based on user-confessed activity and mood. Each of those three companies has found an engaging way to do it. But none of them has the multi-year history of building data around understanding user context.
Tuning that combination of technology and humanity is the ongoing challenge of effective music curation. These days, there is a reaction against technology layers, partly because they result in a perception of sameness — Spotify, Rdio, and Rhapsody are indistinguishable to many people. So: Enter the human touch, the marketing of human expertise, and acquisitions of small companies that seem to be getting it right. But like a coin with two sides, digital music curation always needs the both — human brain and machine brawn.