Should music services recommend new music ?

Recently in Forbes, George Howard 1 published a long article explaining why, in his opinion, music services (understand Spotify, Pandora, Beats and others) are wasting time recommending new music. His point of view is simple :

No one wants to “discover” new music. We may say we want to discover new music (…) but the reality is we’re just not predisposed to do so. (…) This shouldn’t be surprising; we’re not really predisposed to want anything new.

Howard even suggests to put his reader to the test :

Think about your favorite band/song. Chances are it’s an artist/song that you formed your affinity for during your teenage years.

Since 2011, services such as Youtube or Spotify have made a great deal about helping their users discover new music. Why ? Because pure-music-streamers know for a fact that after you searched for all your favorite artists and songs by typing their names in the searchbox… well… you’re kind of done! No-one can type the name of a song or an artist they never heard about!

Hence the focus on music recommendation. 2

Music discovery has always been a challenge. In this digital world, there are basically two ways to provide an efficient music discovery service.

1. Algorithms

Since the first days of LastFM gathering data about the music we play on our desktop, organizations like The Music Genome (behind Pandora) know that “if you like this, you will like that”. With the amazing power available through the APIs 3 provided by most music services, music – including music we never heard of – can reach new listeners by hundreds of ways.

Please don't make me discover the 'New King of Pop'!

Please don’t make me discover “The New King of Pop”!

But algorithms have their own limits. Because I love the Thriller album, does it mean that I want to listen to this artist that the press calls “The new Michael Jackson“? Probably not. Give me the one and only real King of Pop!

And if an algorithm carefully analyzes the sound of “You are not alone” and plays me any crappy R&B sound-alike, I’ll jump on the skip button!

You see, as a matter of fact, I do like three songs from Beyoncé (1,2,3), but that doesn’t change the fact that I hate 99% of R&B productions from the last 10 years… How can an algorithm serve me well if I start by choosing “If I were a boy”?

The new music I may like may be somewhat connected to the music I already love, but maybe not in a way that any algorithm can understand. If songs or artists suggestions from music services are only based on what I already like/love, then there’s no way this will be a real service : I’ll always like the original more than the “new thing”.

George Howard explains :

If you’re constantly forced to contextualize new music via its relationship to music you are already familiar with, this new music is almost axiomatically going to fall short. Why bother with new music that approximates something you already like when you can just stick with something you know you like, and thus reduce any risk of being disappointed.

2. Recommendation / curation

I think we have invented 5% of what could be invented to drive active discovery.
Axel Dauchez – Deezer, CEO

The only other way to discover new music is to let people tell you about it. Intermediaries. That was your music shop dealer who knew all about your tastes. That was that hot chick in high school who was making [what-was-not-yet-called] mixtapes and duplicating them for a selected number of disciples. And that is now every possible source you decide to trust. Because… this is the keyword : who you trust to give you reliable advice on good music to listen to.

While this whole discussion is considering one type of music (commercial), strangely enough, in some other music worlds, aficionados are on a constant quest for new music! Think about classical music lovers, for example. I’m not saying they are searching for new composers (duh!) but they are some of the biggest buyers of new recordings, because each interpretation differs from another, and they can own ten versions of the same opus in their music library. Pop music fans, on the other hand, cherish mostly two versions of a song : the radio edit and the album cut.

The biggest enemy of everything-on-demand hides into your brain : laziness.

Most of the times, when we do discover new songs or new artists, we do it… passively! And what is a better medium for passive listening than our good old radio? It’s no surprise then to discover that again in 2014, US listeners discover new music mostly on AM/FM broadcast radio! Look at this chart :


I understand that Spotify really wants to offer an amazing service to… record companies. So I understand this focus on new music discovery. I understand also that from a geek point of view, everything is in the data.

Digital music services try to reinvent the wheel. They think that the Graal of successful music recommendation is hidden into the Big Data they collect from us. I think they are wrong. I think that what’s hidden in these Big Data is the secret of music consumption. And that’s already something amazing to work on.

Radio – in its different avatars – is irreplaceable for music discovery, because of two aspects :

1. It satisfies listeners laziness.

When radio carefully chooses a new song, there’s a human being behind that choice. And as we all know, these are the only 3 minutes when the station really takes a risk : a new song is an untested song. By providing these 3 minutes of discovery, radio stations do the job for you!

2. It embeds the power of its brand into the recommendation.

On Spotify or Youtube, you can decide to accept David Guetta or Bruno Mars as influencers. That’s cool. But when your favorite radio chooses a song and tells you it’s hot, it comes from a brand that “always existed” and that is deeply identified as the expert is a particular music genre. A brand will always be stronger than a single person.

Digital Music Services try to solve a problem that we radio operators solved half a century ago.


  1. Associate Professor of Music Business/Management at Berklee College of Music, and the COO of Norton, LLC, the parent company of Concert Vault, Daytrotter, and Paste Magazine, former President of Rykodisc, and one of the original founders of TuneCore.
  2. This 2011 video from Youtube demonstrates the focus on music discovery, by all possible means, including curation/recommendations from famous artists.
  3. In computer programming, an application programming interface (API) specifies how some software components should interact with each other. It’s the way for apps to exchange informations between each others.
Share this Story

Leave a Reply

Your email address will not be published. Required fields are marked *

Join me on Facebook !

" Denis develops a global vision where radio fully integrates digital opportunities."
Thierry Chamoux, EMG

" Denis is a respected expert in the Radio industry. His strategic vision of the different areas of the business is highly appreciated."
Julien Pauchet, Radio France

" If you are looking for someone to deliver sharp insights into the evolving audio landscape, I would certainly recommend that you get in touch with Denis."
Matthew Carver, EGTA

" Denis’ presentation was crisp, tangible, and very entertaining."
Martin Baumgartner, Swiss Radioworld