Smarter Playlists: automate your music discovery, playlist strategy, and library organisation

Smarter Playlists is still the best way to ‘automagically’ create and update playlists on Spotify. The tool, made by Paul Lamere of music data firm The Echo Nest (now Spotify), provides you with countless ways to source music, combine it, filter it, sort it and turn it into playlists.

I hinted at the value of Smarter Playlists / Playlist Machinery when I wrote about playlist strategy in a previous post titled If you want to start a music brand, don’t wait until the pandemic is over. Here’s how to use it.

Music discovery

Not everyone needs a playlist strategy, but everyone reading this is crazy about music and always curious to explore more. Here are some examples of recipes that surface gems.

New Music Friday… but high-energy from around the world

Fridays are when new music is released and Spotify helps surface that new music in numerous ways. It has its the algorithmic Release Radar which lets you listen to tracks from artists you personally follow. It also has New Music Friday playlists for specific territories that are editorial and mostly pop-focused.

I love seeing how trends emerge and are adopted around the world and have a soft spot for high-energy music, so I created a weekly tool to scout new tunes.

A lot happening in this screenshot, so let’s break it down by steps.

Firstly, all of the data streams in from the left and streams out (to a Spotify playlist) on the right. In between, there are various steps which either combine data (e.g. tracks from different playlists), filter, or sort it.

  1. First I added a number of Sources. The Sources are Spotify’s New Music Friday (NMF) playlists from various regions. You copy the playlist URI and add it to the box. I’ve changed the box names to the region it’s sourced from.
  2. Since the international NMFs also tend to feature the world’s biggest pop stars, who I’m already familiar with, I took the global New Music Friday playlist (which has over 3M followers) and connected it to the mixer with a red line. This ‘bans’ all the tracks on the global NMF playlist and essentially filters out the global hits from progressing down the workflow, in case they’re present on any regional playlist.
  3. Since I’m working with 7 input sources, I set the mixer’s max tracks to a few thousand. Otherwise it clips to a low number by default.
  4. It’s Friday – I want energy (tbh, I always want energy). So I took the energy filter and set it to ‘most energy‘. This filters out all tracks that are not energetic.
  5. Next, I’ve sorted the stream by artist popularity and picked ‘reverse’, so that the most popular artist shows up on top of the list. This is counter-intuitive, but it makes sense if you dive into how they rank artist popularity numerically. I do this, because if people visit the playlist and play track 1, it makes it more likely it fits current trends and expectations and people are less likely to move on to another playlist.
  6. But life shouldn’t be too predictable. So I’ve used ‘weighted shuffle‘, which lets you set the degree to which you want the list to be randomized. If you want things to be roughly in order of popularity, you set it to 0.1.
  7. In the above recipe or formula, I save the output to an existing Spotify playlist in my collection. I’ve chosen to overwrite, but you can also select to append. Additionally, you can choose to create a new playlist altogether.
  8. Hit the play button to run your workflow, check if the output makes sense in the Tracks tab and also check your Spotify library for the playlist.

👉 Playlist | Program

I’ve used the scheduler to update it weekly, because I was happy with the result and I imagine I can build a following with the playlist. You can find the scheduler by going to the Program section after saving your playlist recipe.

Scout labels’ playlists for unknown talent

Labels usually have regularly updated playlists which showcase their new releases. If you’re curious about musicians that are less well-known, you can set a filter that removes all tracks by artists that are too ‘popular’ (according to Spotify) for your taste.

The above example features 3 prominent drum & bass labels and is set to append less well-known artists’ tracks to a playlist on a weekly basis. (for the connoisseurs: some of the artists in the playlist are indeed quite legendary, but somehow don’t index high on Spotify’s popularity scoring)

👉 Playlist | Program

Playlist strategy

This toolset is also excellent for simplifying the work that goes into maintaining playlists one might use to build their following. Here are two examples.

Sourcing scene playlists for fresh music

Let’s say you read my recent post and are now building a new music brand. You already have a feeling of what it should sound like and are familiar with popular & less popular playlists in your scene. Your flow might look something like this:

I’ve added red dots to the playlist boxes to make it clearer which is which. In the big group, I have 8 different playlists (Wixapolo, Hardtekk, Lobsta B, Clubland, Pumping, Makina, Hard Dance Interpretations and an old playlist I no longer update) that get randomized and duplicate tracks removed before the mixer picks 50 tracks from them.

I’ve split 3 playlists from that path. For Lento Violento, I want to limit the amount of tracks that may show up, so the mixer on the left is set to a very low number, so only a couple of tracks enter the pool. For the Hyperpop playlist, I only care about the high BPM tracks that may be in there. Lastly, there’s trash rave, which is a big pool of music I add music to regularly. I want this playlist to dominate the flavour of the final playlist, so I’ve seperated it, so I can make sure the mix from the 10 playlists on the left have about a 50/50 ratio to the trash rave playlist.

Artist separation makes sure the same artist doesn’t appear multiple times in a row.

Enjoy some of the goofier bpms of dance music.

👉 Playlist | Program

Turn one big playlist into daily instalments

Let’s say you’ve been collecting loads of music into one big playlist, but you want to turn that into a highly engaging format that people come back to daily. This one is really simple.

For years, I’ve been compiling various types of Club Music into one big playlist – from Jersey Club to Juke to Bmore, you name it. Let’s turn it into a brief playlist that people can come back to daily.

Shuffle the input, so you don’t end up with only the top tracks (update: in this case, the ‘sample’ selector does the same as ‘shuffle → mixer’). Remove duplicate artists, since it updates daily, so keep it varied. That’s it. Don’t forget to set it to update every day via the clock icon in the Programs tab.

If you’re ready to move, give the resulting playlist a listen.

👉 Playlist | Program

Library organisation

Not everyone’s building playlist brands, but you may have a library that could use some organisation.

‘Focus music’ playlist based on what you know

This one was shared by Antoine Marguerie, a designer at Base Secrete.

I’ve rebuilt it and it takes familiar music (less distracting), filters the stream to only include low BPM tracks, removes some duplication, removes any lyric-heavy tracks, and takes a 100 tracks to add to the focused work playlist. A good way to reconnect with music you’ve already discovered.

For me, the result still requires some fine-tuning, because sometimes Spotify gets the BPM wrong and thinks a 160 bpm track is 80 bpm. This may not be an issue for most people, but my music taste makes those false positives quite likely to appear in my library. You could address that with energy and danceability filters.

👉 Program

Cleaning up a playlist with lots of albums

One of Spotify’s strengths is the convenience with which you can build playlists. Just drag and drop albums into a playlist and you’re done. The result is a playlist with albums all grouped together. In case you don’t want that, here’s something you can do.

This takes an unorganized source playlist, puts the most popular tracks towards the top and then shuffles things around with ‘weight’ (meaning you can set how random you want things — less random preserves the rough order of the list). In this formula I sent it to a new playlist, because I wanted to hold on to the source playlist.

👉 Playlist | Program

Your turn

The Smarter Playlists has FAQs and many additional examples. Start playing around and think of how you may put this to use. By automating, you’re programming, since this tool is a lot like a visual programming language. You can drop your programs in the comment section below, or drop them in this Twitter thread. Don’t forget to make backups in case you’re overwriting playlists.

If you thought this post was valuable, please consider joining the 2x-week MUSIC x newsletter and / or joining the MUSIC x Patreon.

Mood augmentation and non static music

Why the next big innovation in music will change music itself — and how our moods are in the driver’s seat for that development.

Over the last half year, I’ve had the pleasure to publish two guest contributions in MUSIC x TECH x FUTURE about our changing relationship with music.

The first had Thiago R. Pinto pointing out how we’re now using music to augment our experiences and that we have developed a utilitarian relation with regards to music.

Then last week, James Lynden shared his research into how Spotify affects mood and found out that people are mood-aware when they make choices on the service (emphasis mine):

Overall, mood is a vital aspect of participants’ behaviour on Spotify, and it seems that participants listen to music through the platform to manage or at least react to their moods. Yet the role of mood is normally implicit and unconscious in the participants’ listening.

Having developed music streaming products myself, like Fonoteka, when I was at Zvooq, I’m obviously very interested in this topic and what it means for the way we structure music experiences.

Another topic I love to think about is artificial intelligence, generative music, as well as adaptive and interactive music experiences. Particularly, I’m interested at how non-static music experiences can be brought to a mass market. So when I saw the following finding (emphasis mine), things instantly clicked:

In the same way as we outsource some of our cognitive load to the computer (e.g. notes and reminders, calculators etc.) perhaps some of our emotional state could also be seen as being outsourced to the machine.

For the music industry, I think explicitly mood-based listening is an interesting, emerging consumption dynamic.

Mood augmentation is the best way for non-static music to reach a mass market

James is spot-on when he says mood-based listening is an emerging consumption dynamic. Taking a wider view: the way services construct music experiences also changes the way music is made.

The playlist economy is leading to longer albums, but also optimization of tracks to have lower skip rates in the first 30 seconds. This is nothing compared to the change music went through in the 20th century:

The proliferation of the record as the default way to listen to music meant that music became a consumer product. Something you could collect, like comic books, and something that could be manufactured at a steady flow. This reality gave music new characteristics:

  • Music became static by default: a song sounding exactly the same as all the times you’ve heard it before is a relatively new quality.
  • Music became a receiving experience: music lost its default participative quality. If you wanted to hear your favourite song, you better be able to play it, or a friend or family member better have a nice voice.
  • Music became increasingly individual: while communal experiences, like concerts, raves and festivals flourished, music also went through individualization. People listen to music from their own devices, often through their headphones.

Personalized music is the next step

I like my favourite artist for different reasons than my friend does. I connect to it differently. I listen to it at different moments. Our experience is already different, so why should the music not be more personalized?

I’ve argued before that features are more interesting to monetize than pure access to content. $10 per month for all the music in the world: and then?

The gaming industry has figured out a different model: give people experience to the base game for free, and then charge them to unlock certain features. Examples of music apps that do this are Bjork’s Biophilia as well as mixing app Pacemaker.

In the streaming landscape, TIDAL has recently given users a way to change the length and tempo of tracks. I’m surprised that it wasn’t Spotify, since they have The Echo Nest team aboard, including Paul Lamere who built who built the Infinite Jukebox (among many other great music hacks).

But it’s early days. And the real challenge in creating these experiences is that listeners don’t know they’re interested in them. As quoted earlier from James Lynden:

The role of mood is normally implicit and unconscious in the participants’ listening.

The most successful apps for generative music and soundscapes so far, have been apps that generate sound to help you meditate or focus.

But as we seek to augment our human experience through nootropics and the implementation of technology to improve our senses, it’s clear that music as a static format no longer has to be default.

Further reading: Moving Beyond the Static Music Experience.

Spotify’s strategy to become a habit-forming product

Last Friday, Spotify unveiled its newest feature: Release Radar – a personalized playlist of newly released music, updated every Friday. It’s reminiscent of Discover Weekly, but Release Radar’s recommendations are always newer tracks. My first impression is that it’s much more likely to recommend music from artists you’re already familiar with.

As Spotify keeps rolling out features like this, and competitors no doubt follow suit, the implications for the music business will be significant. Matt Ogle, who’s behind both of these playlists, revealed last March:

There are 2,000 artists for whom Discover Weekly is currently 80% of their streams, and something like five or six thousand for whom Discover Weekly is half of their streams.

But I’d like to zero in on Spotify’s product strategy and why features like Discover Weekly and Release Radar are so important for the service. It has everything to do with the power of habit.

Habit Loop Spotify

Discover Weekly creates a perfect habit loop. The routine is listening to your refreshed playlist. The reward is the release of good hormones due to interesting new finds, and perhaps the social currency of sharing. The cue, or trigger, is simply the fact that it’s Monday and the start of a new week.

On Sunday, another habit loop is triggered. To prevent losing newly discovered gems, users log on to save tracks from Discover Weekly to their playlists. Loss prevention is one of the strongest motivators.

Spotify Discover Weekly Habit chart

Spotify’s bet is that they can create another habit, focused on different days of the week, by releasing a new feature in the style of Discover Weekly. Being able to consistently drive traffic back to your product is great if you’re ad-supported, might help to convince free users to upgrade to premium, and helps premium users justify the recurring cost of their subscription.

Now, instead of 2, there will be 4 cues.

Spotify Discover Weekly + Release Radar chart

Friday is a great day for Release Radar for two reasons:

  1. Easy to remember: it’s the last day of the week and people have the weekend on their minds.
  2. Since last year, Friday is the global release day for new music.

Here are two hacks I made that bring some cool additional automation to the new Release Radar playlist:

Curious how Release Radar works? The wonderful folks at Hydric Media, who are behind the hit music app Wonder, created a free tool called Playground, which opens up all the different parameters of Spotify’s Echo Nest API powering the Discover Weekly and Release Radar playlists.

How has your Release Radar experience been? I’ll show you mine, if you show me yours. Send me a tweet: @basgras.