Apple Music Launch: Too Bad Steve Jobs Is Not Around

Today Apple launched its Apple Music app to much industry buzz about how the company might shake up the current streaming music market. But it may have wished Steve Jobs was still around to nail the user experience and the timing of the launch. Apple Music combines four core services:

• Connect: Social media site for musicians and fans to interact.
• Radio: Beats 1 live stations, genre-specific stations, and user-created stations based on a seed-song.
• New: Curated playlists of new albums or artists (1989 from Taylor Swift showed up at the top. They are really giving in to her!)
• For You: Recommends playlists based on your genre and artist preferences, and your iTunes library. The playlists are curated by music experts.

Artist following via Connect and Apple Music’s radio stations (with limited skipping) are part of the free plan, so enjoy. Apple bets that it can convince you to pay $9.99/month for its streaming services, although for now there is an introductory three-month free-trial.

Apple enters the paid music streaming arena with a leg up: almost one billion iTunes users and the possibility to mine their purchase and consumption behavior. If you are an iTunes user, Apple can easily narrow down your music taste to a couple genres and a few hundred songs, and even track which ones you have listened to more frequently. Early reviewers of Apple Music already noticed how the For You curated playlists nailed their music taste. And with the ability for iPhone users to request songs via Siri, just voice your request for narrow searches of its library of more than 30 million songs. With a potential edge in personalization of its streaming service, will Apple win the battle for paid streaming over Spotify and Pandora ?

The simple answer is not yet, for two reasons: The cluttered interface and the disintegrated recommendation engines. Many early reviews are already noting that, even though the design looks cool, the screens are cluttered and confusing. James Dilellio, a faculty colleague at Pepperdine University, Apple lover and heavy Pandora user, stated: “It takes me too much to get what I want. I like the simplicity of Pandora’s interface to create stations based on a seed-artist and how Pandora then picks the songs.”

If the interface makes getting streaming recommendations cumbersome, the recommendation engines themselves are good but not ground-breaking. To explain, here’s a quick primer on recommendation engines. There are two types based on the algorithm used: consumption-based and content-based. Spotify’s recommendation engine uses a consumption-based technique called collaborative filtering, where two users that listen to the same or similar music will be considered “neighbors” in music taste and offered the same or similar recommendations, including songs one has listened to that the other one hasn’t.

In contrast, Pandora uses a content-based algorithm that makes recommendations based on hundreds of features analyzed for each song, such as genre, rhythm, and beats per minute (see Pandora’s Music Genome Project). It takes significant time and investment for Pandora’s music experts to classify songs along hundreds of features, while the algorithm helps identify the features that matter most for each user to make song recommendations.

Recommender system experts suggest a hybrid of consumption-based and content-based logic can be most effective, because it combines the advantages of both techniques. Consumption-based algorithms recognize the social element of music and capture the wisdom of the crowds, while content-based algorithms measure artistry and acoustics. Socially-driven recommendations gravitate towards popular songs for the mainstream listener, while content-based logic can lead to novel song recommendations from new, niche, or undiscovered artists.

Apple can leverage iTunes to enter with a ground-breaking hybrid recommendation engine. For consumption-based logic, iTunes offers valuable data to mine consumption patterns for micro-segmentation: billions of user-created playlists, consumption behavior, purchase behavior, and even demographics. For content-based logic, iTunes has more than 30 million songs across billions of user-created playlists to determine what song features matter to individual listeners. To top it off, adding human-curated playlists can address the imperfections of the engine’s “cold start” recommendations for non-iTunes users, while the machine learns about their taste.

Does Apple Music’s song recommendations live up to this potential? The simple answer is no. I tested Apples Music’s recommendation engines with Jesse Bockstedt, a faculty member at University of Arizona who is a music aficionado and has done some very interesting research on recommendation engines. We both downloaded Apple Music and formed a preliminary impression on the quality of its recommendations.

To activate For You’s streaming, Bockstedt first selected genre and artist preferences in a slick bubble interface. He reflected on For You's first recommendations: "It was either songs and albums in my iTunes library or curated playlists that were just based on the genres I selected in the start-up screens. I was expecting more here, especially since they have access to my entire music library, playlists and playing history. I was expecting more interesting recommendations." While some may say For You nailed the recommendations, Bockstedt was not excited because he was expecting new, novel songs in a steady stream.

He was also expecting song recommendations a la Pandora or Spotify, and that was nowhere to be found… until after searching around he found that, if you click on the three dots to the right of the song that’s playing, you can “create a station” with the song as the seed. Then, you have to jump to Radio and you will see the new station. This user-created station functionality is a stripped version of the existing iTunes Radio streaming service, with the difference that there are no ads (that will likely change after the free trial period). By the way, I checked with DiLellio to see if he had found the seed-song creation option and he had not, hence his frustration as a heavy Pandora user, used to the simpler interface.

It appears that Apple has simply incorporated the former Beats Music functionality with curated playlists in For You, and the iTunes Radio recommendation engine in Radio’s user-created stations. Still, let’s assume that the For You streaming marginally outperforms the Spotify and Pandora algorithms, based on the addition of curation from experts and some leveraging of iTunes playlists. Bockstedt performed some experiments on music recommender systems that can shed a light, with his co-authors Shawn Curley, Gedas Adomavicius, and Jingjing Zhang: “What we found is that online recommendations from a trusted source bias a consumer’s preferences and willingness to pay in the direction of the recommendation, even if the recommendation itself is not 100% accurate.” So any marginal improvement in a recommendation is likely to go unnoticed, as consumers adjust their taste to the recommendation engine. This does not sound good for Apple, because it has missed the chance to launch a killer engine that integrates consumption-based, content-based, and curated recommendations to really sway consumers away from competitors.

If Steve Jobs would have been around, patching multiple music services would have been unacceptable, forcing recommender engine integration to make the interface look slick and simple, yet powerful. Nevertheless, Apple Music is here to stay, and it will be a force to reckon with as it leverages its massive iTunes user base and multi-media platform. The challenge for Spotify, Pandora, and other competitors will be to compete with Apple Music for new streaming customers, especially those who already have an iTunes membership. But Apple Music’s launch could have been a killer app to steal market share, and instead, it enters the battle with a bunch of services patched together. For Apple Music to be a hit a la Steve Jobs, it needs much improvement.

Do you like Apple Music's song recommendations? Please comment.