30 days of ideas – 25: Next-gen personalised music radio


Sadly, this is just a workaround until someone invents the real thing

I teach music programming for radio. And while it’s true that music programming on broadcast radio is typically safe and boring, there are some very sensible and useful practices that understand how people make meaning from music, and how music fits into people’s lives that I think online services could really learn from.

And while not everyone wants to go through their music collection each night, finding the right songs to listen to the next day, and putting them in order – I think there’s something in between that and simply hitting the ‘shuffle’ button that could be useful.

Real personalised music radio
Existing ‘personalised music radio’ services don’t understand (or don’t factor in) the things that make radio programming an actually useful and worthwhile endeavour.

It’s not about just playing songs that ‘sound like’ other songs you like (Pandora) and nor is it about playing songs that you’ve played in the past, interspersed with songs other people with similar tastes have listened to – but in an entirely random order (Last.fm).

It’s about choosing the right music and then putting it in an order that makes sense.

‘Music makes meaning’ is my music programming mantra. And meaning doesn’t only happen on a song by song basis. You can create meaning over time with sequences of songs. It’s a construction of narrative.

This is about the difference between “a good next song – and “the right next song”. And there’s no reason we can’t have that.

Actual music programming – not just random assortments
I’m not going to go into massive detail here about programming theory and the psychology of playlists, but I’d like to suggest that anyone who wants to do anything serious with the next generation of music consumption at least has a good think about it.

I’m talking about automated music programming. iTunes as a real radio station – not just a jukebox. A sort of simplified and guided RCS Selector for home consumers.

The idea is to use radio programming techniques and practices, as well as existing software technologies to make a playlist you can leave on all day, be sure of hearing your favourite songs, and never get bored or turned off by the selection.

Here are a few concepts that would be helpful to integrate:

A-rotates: You probably have between 6 and 12 ‘favourite songs’ at the moment. Over the course of a day, you might wish to hear these ones more than once. They’re your superhits.

Gold: These are your all-time favourite songs from way back. Oldies but goodies. Used sparingly, you get the ‘Oh my god – I haven’t heard this in AGES!’ effect.

Rules: In order to make a playlist work, you need to set some rules. Don’t play me two songs by the same artist within an hour. Don’t play me the same song at the same time each day. Make sure that 30% of the music you play me is New Zealand music.

Dayparts: There’s a difference between music you want to listen to during the day, and music you want to listen to at night. You can even break it down further into morning and afternoon – or even breakfast-only ‘get up and going’ songs.

Mood: You want a variety of moods, but you don’t want to go from a really introspective and melancholy song straight into ‘Walking on Sunshine’ by Katrina and the Waves.

There’s a superb iPhone app called MoodAgent which does a great job of identify the ‘feel’ of a song and plays you a selection of songs of that same mood. My ideal system would let you transition moods over time, but without jarring jumps and sudden energy drops.

Perhaps an API or even a collaboration with the MoodAgent people would address this issue. They’re certainly partway there – but again, it’s about sameness, rather than meaningful variety.

As easy it can be, and as complex as it needs to be
While there’d be some key settings that could be defaults, it’d be good for ‘power users’ to be able to get in and mess with them. But the key to this would be to build an initial playlist based on your favourite music.

Choose 10 tracks.

That’s your A-list. From that, the software can work out a bunch of other useful stuff.

For a start, using the iTunes Genius playlist builder (or some similar algorithm), it could easily populate a B-list of other stuff you’re likely to enjoy based on a Venn diagram intersection between further songs that are suggested by that A-list, and what is currently in your music collection.

You could, of course, then go through that lot and chuck out anything you don’t want in there – and maybe add some other stuff too. It’s good to be flexible.

Then there’s the Gold stuff.

It could go through and grab old, famous music from your collection – and you could add and subtract as necessary.

The trick here is to have quite a lot of Gold, but only use it sparingly. Draw on about twenty tracks that get included in the rotates for a week or so – then rest those while a different twenty get spun into the mix.

Now add a Flavour category.

Pick a bunch of songs that you’d like to have pop up every now and then, even though they don’t fit the regimented and automated structure of your playlist.

Love Joni Mitchell, Prince, Black Sabbath and Kylie Minogue – but your main playlist is mostly indie rock? There’s really no way your algorithm is going to guess that. So throw them into your Flavour playlist, and they’ll get included.

Added cleverness
You could do smart things with new music – anything you’re still getting used to. Tracks that have been added to your collection recently could be automatically sprinkled in the mix – as could, with a bit of cleverness, recommended tracks streaming from an online service like Last.fm.

We’ve mentioned dayparts – only playing certain songs at certain times of the day. The playlists could also be weather aware.

Sunny outside? Why not automatically feed in data from online weather services, and ratchet the overall mood score up a bit to reflect the brighter day. Or only ever play ‘Friday I’m in Love’, ‘Manic Monday’ or ‘Sunday Bloody Sunday’ on the right days of the week.

Change the station
Of course, your radio station is essentially based on a selection of your 10 favourite songs. But you could just as easily have a different station built out of a different 10 songs, and you could mess with that to your heart’s content.

Once it’s all set up though, you can change the playlist from day to day – or just choose another 10 A-rotate songs on a Sunday night so your week of listening’s all sorted and ready to go.

Finally, I’d suggest an ‘Over It’ button – which is not so much a ‘permanently delete it from my playlist’ option – but just an instruction to give that song a rest for a week or so.

There are, of course, lots of further levels of detail that we could go into, but the point of this is to make it far more programmable than existing services and software – without making it difficult or daunting.

And you’d want this to easily transfer over (or be replicated on) your iPod as well.

The idea is to have your own personalised radio station. Not the poor approximation to that we have at the moment, but rather, as if you’d hired a radio music programmer, and asked them to make a radio station just for you, using your own music collection as a starting point.

Free to a good home
If you have the programming skills to make this work, and you’re interested – then please go ahead and make this. The idea’s up for grabs.

And while, to the untrained eye, this all might seem like an unnecessary complication to what is already a pretty well-covered selection of tools and services… trust me: done right, this will make your iPod psychic.

Table of contents for 30 Days of Ideas

  1. The other way of following first
  2. Now we’re up and dancing
  3. 30 days of ideas – 01: Keymash
  4. 30 days of ideas – 02: Radio Alerts
  5. 30 days of ideas – 03: Only Famous (a romantic comedy)
  6. 30 days of ideas – 04: Modcasts
  7. 30 days of ideas – 05: Numberless Calendar
  8. 30 days of ideas – 06: SpringCleanr
  9. 30 days of ideas – 07: Street Gallery
  10. 30 days of ideas – 08: Smart Business Cards
  11. 30 days of ideas – 09: Recordings in Concert
  12. 30 days of ideas – 10: Vinyl scanner
  13. 30 days of ideas – 11: Photo Stack-and-Scan
  14. 30 days of ideas – 12: A Box of Cool
  15. 30 days of ideas – 13: Karaoke-Tube Celebstar Idol
  16. 30 days of ideas – 14: I Made You A Tape
  17. 30 days of ideas – 15: Newspaper download codes
  18. 30 days of ideas – 16: Pebble Splash
  19. 30 days of ideas – 17: Digital radio, somewhere useful
  20. 30 days of ideas – 18: Public domain music collection
  21. 30 days of ideas – 19: Blog cast-list automator
  22. 30 days of ideas – 20: The Retirement Pile
  23. 30 days of ideas – 21: Nationalise EMI
  24. 30 days of ideas – 22: The Stainless Steel Rat (the movie)
  25. 30 days of ideas – 23: WordPress Bandcampify template
  26. 30 days of ideas – 24: Rollercoasters as public transport
  27. 30 days of ideas – 25: Next-gen personalised music radio
  28. 30 days of ideas – 26: New Music Trust
  29. 30 days of ideas – 27: Tamagotchi Gardening
  30. 30 days of ideas – 28: Charity shop clothing subscription
  31. 30 days of ideas – 29: ‘Now Playing’ social music app
  32. 30 days of ideas – 30: House of Spare Ideas
  33. Mixtape for You by Ray Kuyvenhoven
  34. What can you do in 30 days?

2 Responses to “30 days of ideas – 25: Next-gen personalised music radio”

  1. Lee Jarvis says:

    In a funny kind of way, I feel that this “mood-interpreting, music-knowledge-junkie, perfect-next-record-selector” is the sign of a good DJ. That can mean a dance club DJ, but also a great radio jock. When people say to me, “ah I’d love to DJ but don’t think I can” I always say that the technical aspects are easy to learn – if you can count to four you can learn to beat-match. However, there is this huge leap from someone able to play one record after the other, and a great DJ who has this connection with and understanding of their listeners. A good one is an educator and an entertainer simultaneously.

    Bringing this back to a technological interpretation, I do listen to and analyse to Pandora / Last FM and enjoy testing their recommendation systems, but similarly, they have always been that same leap away from the intelligent and near-psychic process which you are describing. You are right that if this could be developed it would be the perfect system.

    I guess I am torn between wanting such a system and being willing to pay well for it, and wanting it to be an unachievable goal because I believe in and desire the HUMAN touch and experience and skill of said humans ;) Either way I hope to pay close attention to the developments.

    Best, Lee.

  2. Jim Kerr says:

    Love the idea. The implementation is the hard part. Not technology-wise, but user-experience wise. If there is one thing that we are consistently seeing in the music space is that music consumption is a passive experience (often called a “lean back” experience). Anything that requires proactive work by the user to “get it right” is annoying. There are plenty of people that find the simple process of clicking on “skip this song” buttons on Pandora to be annoying.

    As Lee outlines above, this process has historically been handled by radio, and–despite the negative press–radio has done an increasingly good job of it over time. Still, it is not PERSONAL radio, and that’s the subject of your post. I’m guessing there is a technological solution, but it would require some in-depth behavioral data and analysis, more than perhaps even Pandora has on file.

Leave a Reply