This article is part of CMO.com’s September series on the state of media and entertainment. Click here for more.
Where do you start when you want to kick back, relax, and be entertained? How about watching one of the thousands of movies or TV shows now available on services such as Netflix, HBO Now, Showtime, BBC’s iPlayer, or Amazon Prime. Or maybe you’d prefer listening to some of the 30 million tracks on Spotify, Apple Music, Tidal, Google Play, Pandora, or Amazon Music Unlimited. If there aren’t enough human attention minutes to warrant all the premium content being produced by the media and entertainment industry—and let’s not forget all the content freely available on social platforms—where does the entertainment industry go next to ensure its new content gets watched or heard?
The obvious place to start is discovery or recommendation services. Amazon has been recommending books, music, and movies based on our browsing and purchase history, as well as ratings and reviews, for close on two decades, and it works, at least from a sales point of view—35% of Amazon.com’s revenue is being generated by its recommendation engine. But how often have you bought something based on these types of recommendations and wondered why it doesn’t always deliver?
The prediction algorithms used to deliver recommendations are not only of huge importance to online retailers but also to entertainment companies—the more accurate they are in delivering what consumers want, the more they buy.
In The Ocean Of Tunes
In 2005, Last.fm became one of the first services to successfully launch a music discovery and recommendation service. It uses collaborative filtering to recommend you music based on your own listening profile, then compared it to the listening habits of others with similar tastes and interests to provide recommendations. Although the site is still with us today, the recommendations it provides can often seem a little too obvious. Would Bob Dylan fans be pleasantly surprised to know they might also want to check out Neil Young or Van Morrison, when they were looking for something a bit more tangential?
The emergence of music streaming platforms such as Spotify, Apple Music, and the Jay-Z-owned Tidal has made discovering music an even bigger challenge for consumers, with tens of millions of tracks to choose from. So it’s no surprise that these sites have invested in music recommendation and discovery services.
The man responsible for developing the code that powers Spotify’s discover service, Edward Newett, explained to Wire Magazine how it delivers recommendations: “There are two parts to how the algorithm works: on the one side, every week we’re modelling the relationship of everything we know about Spotify through our users’ playlist data. On the other, we’re trying to model the behaviour of every single user on Spotify—their tastes, based primarily on their listening habits, what features they use on Spotify, and also what artists they follow.” The service has been hugely popular. In May 2016, Spotify claimed that more than 40 million people had used its Discover Weekly service, streaming just under 5 billion tracks in less than 12 months.
AI To The Rescue
The emergence of AI offers those in the recommendations space new ways of delivering even better recommendations.
One of the companies at the forefront of these new discovery tools is Belgium–based Musimap. Its technology has helped the company decode music’s DNA and then understand listeners’ contextual needs to help solve the complex problem of music discovery.
Chief executive Vincent Favrat believes the company can provide better recommendations than the services that have preceded it because it not only brings together both human intelligence (HI) and artificial intelligence (AI), but it also takes into account listener’s context and emotions to deliver recommendations.
“We’ve leveraged the collective knowledge of music experts to identify the most important artists and the most important tracks in the global library of music,” Favrat said. “The library consists of 1.6 million tracks from 70,000 artists, which provide the backbone of the recommendations. The AI is then used to optimise the recommendations, to provide another layer on top of the human input.”
As Favrat explained, a key part of the optimisation is to understand the listeners emotions: “Ninety percent of our choices are not led by our rational brain, but they are led by our mood and feelings. Music is a very important medium in helping us understand our emotions.”
Capture The Essence
Favrat believes that understanding consumers’ emotions in relation to music can have a broader application than just music recommendations. He believes that marketers can use the emotional intelligence gathered from understanding our feelings in relation to music to then capture the essence of the consumer, enabling brands to deliver relevance by delivering programmatic ads with a contextually relevant soundtrack that reflects the end-consumer’s tastes, interests, and emotions.
Musimap also takes a psychological approach to its music recommendations. “That means that, for example, the energy level of a song is an important element that reflects how your body moves while listening to music,” Favrat explained. “It’s quite obvious that you don’t listen to classical music the same way you would listen to heavy metal or pop music.”
The first service Musimap has released is aimed at music supervisors who are responsible for licensing music for TV, film, video games, and adverts. “What we do is we use music as a door to understand users and their emotions, and we can tag the emotional profile of a potential user. That means we can also tag the emotional profile of a song, of an album, of an artist, or a brand. The brand can be mapped and matched to the right music, while an advertisement, for example, can also be tailor-made for a specific demographic, which is very powerful.”
Favrat believes that “when you deliver perfect recommendations, it’s the exact opposite of spam. And what we want to achieve is to help brands to stop spamming people and making contextually relevant recommendations.”