Libraries in The Stream

It is funny that now Spotify “allows” us to like more than 10,000 songs, but offers no practical way to browse or manage a collection of songs this large.

The liked songs library is presented as a single list, which you can choose how to sort and there is a filter search. Other than that, it show some 20 songs per screen (I counted 19 on my monitor and ‘actual size’ view). To scroll 10,000 songs will take more than 500 screens. 

The other main library sections, Artists and Albums, are less essential, at least in my opinion. Following artists should give you updates when they release music, and listening to albums is much less prominent than listening to songs. And you can always like every song in an album, so they appear in your main Liked songs library…

Liked songs is very useful: it is the most practical way to keep track of all/most things from your listening history that you actually enjoyed. Other methods to access your listening history in Spotify are controlled by Spotify and their implementation changes all the time. 

The liked songs serves not only as a history of the listening but also as a history of the listener. As time goes by and the liked songs are accumulated, there is some personal memories being stored there as well, an audio scrapbook maybe.

The liked song library also has a symbolic value: it is a very small pond of personal, familiar songs within a gigantic, ever-expanding ocean of unfathomable content (more than 60 million songs and more than 40 thousand songs added daily).

Of course, a big priority of Spotify is to create ever improving recommendation tools, advanced enough that they may “know” in advance the songs, artists, genres, playlists the user will like. Maybe if you are trying to build the “perfect” recommendation system, the archive of liked songs is just a distraction. Time spent listening to liked songs is time not spent listening recommended playlists or other innovative promotions. And already-liked songs probably also have little algorithmic value…

But that gives another valuable distinction to our pond metaphor: it serves not only to protect us from the vastness of the unknown; it is a safe harbour from the impersonality of the deep data/machine learning…

In the past few weeks I have been experimenting to see which mainstream streaming service offers the best combination of functionality and sound quality (Spotify, Apple Music, Tidal, Qobuz and Deezer each have strengths and weakness) and as I contemplate switching to one or another as the main service, the only data that feels relevant to migrate from one to another is the collection of liked songs. (Soundiiz and SongShift work both fine).

The Best Albums of 2016, Now and Then

A fun exercise. The 100 best albums from 2016 according to But two different snapshots: the ranking as it was at the very end of that year, versus the ranking of these albums as of July 30, 2020.

Radiohead’s “A Moon Shaped Pool” remains number one, but David Bowie’s “(Blackstar)” went from number two to number three. That’s a -1  position change and a -50% ranking change.

The highest absolute jump was “Telephone” – Nomame with +67 positions. The highest relative jump was “Nonagon Infinity” by King Gizzard & The Lizard Wizard: 80%.

Highest absolute fall: “Cautionary Tale” from Dylan Leblanc: -346 positions. Highest relative fall: “Suicide Songs” from Money: -449%.

One caveat: these ranking combine “recognised” charts (traditional lists from mainstream publications and websites) and charts from members of the site’s community.

Spotify 2Q 2020

A quick summary for easy reference:

  • 138 mm premium subscribers
  • 170 mm ad supported MAUs
  • Total 299 MAUs
  • Total growth:  5% quarter over quarter; 29% year over year.
  • Within Premium, average revenue per user (“ARPU”) of €4.41 in Q2 was down 9% Y/Y (down 7% excluding the impact from FX rates).
  • At the end of Q2, our workforce consisted of 6,049 FTEs (full-time employees) globally.
  • Q3  2020 Guidance:
    • Total MAUs: 312-317 million
    • Total Premium Subscribers: 140-144 million
    • Total Revenue: €1.85-€2.05 billion
      • Assumes approximately 260 bps headwind to growth Y/Y due to movements in foreign exchange rates
    • Gross Margin: 23.1-25.1%
    • Operating Profit/Loss: €(70)-€(150) million
  • Q4  2020 Guidance:
    • Total MAUs: 328-348 million
    • Total Premium Subscribers: 146-153 million
    • Total Revenue: €2.05-€2.25 billion
      • Assumes approximately 360 bps headwind to growth Y/Y due to movements in foreign exchange rates
    • Gross Margin: 23.7-25.7%
    • Operating Profit/Loss: €(45)-€(145) million

Consumption trends by platform are beginning to normalize as well; in-car listening at the end of the quarter was less than 10% below pre-COVID levels having recovered from a 50% decline at the trough in April. 

…expansion of our Microsoft Xbox Game Pass relationship

Separately, the introduction of our TV app on Comcast’s proprietary X1 set-top box marked the first integration with a pay-TV provider in the US, demonstrating continued progress against our ubiquity strategy. 

Product and Platform 

We continue to accelerate product innovation in order to enrich the consumer experience and serve the right content from the more than 60 million unique tracks and more than 1.5 million podcasts on our platform. Through experimentation, we aim to scale features that will lead to improved intake, retention, conversion, and thus higher LTV. For example, we rolled out lyrics this quarter in 26 countries across Southeast Asia, Latin America, and India covering more than 100 million of our users. 

We continue to make improvements to our join flow and onboarding experience and have seen a demonstrable improvement to both short term and long term retention as a result. Once on the platform, we constantly add or improve features such as our recent announcement that we have removed the cap on the number of tracks a user can download, now offering the ability to save an unlimited number of songs, albums and podcasts to their collection of favorites (up from a max of 10,000 tracks previously). 

With COVID-19 protective measures still in place, we launched a Listening Together microsite that visualizes when two Spotify listeners start to play the same song at the exact same time – which happens on average 30,000 times every second on Spotify. A further enhancement to the user experience, Spotify launched a Group Session feature that allows up to five Premium users to share control over the music being played. Group Session participants can control what’s playing in real-time as well as contribute to a collaborative playlist for the group. Another product feature we’re experimenting with is Canvas, which turns formerly static song pages into engaging video-art showcases with 8 second visual loops. Spotify users and artists can now share Canvas artwork to Instagram stories, a feature that is unique to the Spotify platform and sharing experience. 

Ubiquity remains a core strategy, and we continue to find ways for consumers to seamlessly connect with our platform. This quarter we expanded our Spotify Free offering through Amazon Alexa devices beyond the US, Australia and New Zealand. Amazon Alexa devices will now support Spotify Free in Austria, Brazil, Canada, France, Germany, Ireland, Italy, Japan, Mexico, Spain and the UK. Additionally, we are also excited to be launching Spotify (Free and Premium) on Alexa in India. 


Today, 21% of our Total MAUs engage with podcast content, up from 19% of MAUs in Q1 2020, and consumption continues to grow at triple digit rates Y/Y. We see strong MAU growth in podcast content across all regions for Spotify. Overall supply of new podcast content recovered in Q2 after a slight impact related to COVID-19 in the previous quarter. There have been a healthy number of releases for Catalog, as well as Spotify Originals within the quarter. We launched 110 podcast playlists across 6 markets (including 49 new O&E podcasts outside the US) on a variety of themes and topics to continue to drive podcast discovery for users. Currently, Spotify’s podcast catalog has over 1.5 million shows, 50% of which launched in 2020. Our acquisition of Anchor last year has helped accelerate content growth on the platform with approximately three quarters of new podcast releases being powered by Anchor. Spotify announced an additional 9 exclusives like Do You See What I See and Rapot to our Creator Accelerator Program in Indonesia, launched our first original podcast, Search Engine Sex, in Australia, and released our first exclusive, XRey, in Spain. 

As COVID-19 has impacted the creator community and the way in which users interact with Spotify, we created user experiences that allow listeners to adapt to the current environment. 

Spotify launched an “At Home Music & Podcast Entertainment” hub, which has had over 20M unique visitors. Additionally, we created a “COVID-19 Guide” podcast hub to serve as a resource to those wanting to engage and understand more about COVID-19 related information and topics. As we mentioned last quarter, we launched the Spotify COVID-19 Music Relief project, through which we have partnered with organizations that offer financial relief to those in the music and creator community around the world. We pledged to match dollar-for-dollar public donations made to these organizations, up to a total Spotify contribution of $10 million. We also launched Artist Fundraising Pick (a feature that enables artists to raise money to support themselves, their bands, their crews and charitable organizations) in April, and to date have seen more than 91,000 artists take advantage. 

Two sided market place

The number of artists and their teams utilizing our Spotify for Artists tools on a monthly basis has grown to more than 690,000, growth of 68% Y/Y, and these creators are finding ways to unlock new channels for discovery. Growth in the number of artists making up our top tier (those accounting for the top 10% of streams) is accelerating; that cohort now stands at over 43,000 artists, up 43% from 30,000 one year ago. Our product and platform are driving discovery, diversifying taste, and helping up-and-coming artists reach new audiences. Gone are the days of Top 40, it’s now the Top 43,000. We continue to add new features to our suite of Marketplace products to better serve the needs of these creators and their teams. We also expanded our Sponsored Recommendation functionality to Canada this quarter, and look forward to launching in further markets soon. We also began piloting new targeting features in the Sponsored Recommendation product that allow for more flexibility to suit the unique needs of marketers. Conversion remains high, and the number of artists taking advantage of these features grew more than 37% Q/Q. 

Ranking Music Streaming Subreddit Communities by Members

As of July 28, 2020:

July 29 update for Video Reddit (just a few):

Late Era / Welcome to Chicago

Intriguing idea for a podcast:

Andy Cush: Like Winston said, the impulse to do Late Era definitely grew out of Welcome to Chicago: to give some sort of consideration—whether it’s relatively serious thought or just cracking dumb jokes—to music that might as well not exist to 99% of the sane listening population. To me, on some level, both podcasts are oblique reactions to the perverse abundance that streaming services offer us as listeners, even if we’re not talking about streaming directly most of the time on the shows. In music, as in just about every other area of our lives, nearly all of recorded history has become incredibly easy to access—oppressively so, sometimes—with a void of context that can make trivial things seem important and important things seem trivial. By diving into these bodies of work that are more or less forgotten in the cultural memory, but still just as readily available for listening as the Abbey Roads of the world, we’re willingly exposing ourselves to some of the weirdest and most disorienting contradictions of our new reality as music listeners, and trying to make sense of them.

Don’t sleep on Chicago 19:

Winston Cook-Wilson: Since November 2018, the three of us have been co-helming the controversial music and humor podcast Welcome to Chicago, which is an examination of the 36-and-counting-album discography of the band Chicago. This was inspired by a joke on our text thread after seeing a music writer post a picture of the 1989 album Chicago 19 with the caption “Don’t sleep on Chicago 19,” as if it was some deep Detroit rap mixtape people needed to check out. The project has become increasingly high-concept, featuring tons of guests, invented side characters, scripted true-crime episodes and about an album’s worth of original music. We’re still doing that and may be doing it until we are old men.

Good point:

Sam Sodomsky: As a music fan, I have always had a collector’s mindset. If I like an artist, I want to know everything they have ever done. I have probably listened to, say, Neil Young’s Fork in the Road more than most band’s best albums. When I start reading a rock memoir, I always skip ahead to see how much they get into their later albums, because usually those are the most mysterious to me. Why did they choose that cover art? Why is there an entire song about grocery shopping? Most times they don’t get into it. I also find that eventually everything comes around. The same way that I can hear Van Morrison’s ’80s albums with fresh ears, maybe someone decades from now will hear that Who album from last year and feel totally mystified by it. Maybe not. Either way, it’s good to get outside the canon mentality, especially when you’re dealing with these legacy artists.


What’s one tip that you’d give a music podcaster starting out right now?

Winston Cook-Wilson: Don’t do a music podcast about the band Chicago.

Music vs Video Streaming

It might be fun to make a brief comparisons between music and video streaming services.

If only for the reason that, since each video service has an unique content library, its users are much more used to use (!) several different services almost at same time. So, between Netflix, Amazon Prime Video, Apple TV+ and, in the U.S., Disney +, Hulu, HBO Max and Peacock) there is a lot of room to explore different user interfaces, features and even design philosophies.

In the music area, on the other hand, all services have basically the same libraries, basic functionalities and pricing plans, so there is much less incentive to subscribe to more than one service. And user libraries and listening histories grow over time, serving as a barrier to discourage people from switching.

A very recent article by Alan Sepinwall articulates frustrations with video streaming as of July 2020:

“The part of the streaming shell game that I’ve never been able to fully understand — and that has somehow gotten worse with each passing year and each new service debut — is just how bad the user experience is on all of them. It’s been 13 years since Netflix began offering streaming content, with Hulu and others soon to follow, yet the user interfaces consistently seem designed to make finding what you want to see — whether continuing a binge or discovering something new — a Herculean effort.”

“Simply put, there are a lot of basic practices that all streamers should be following, and that most of them don’t seem to understand in the slightest. Here’s our four-point plan to optimize user experience.”

Here are his four points:

1. Always make it easy to resume a binge.
2. But don’t make it too hard to go back and re-watch when necessary.
3. Give the customer more control of the browsing process, and a layout that makes sense.
4. Make searches easy.

For our purposes, item 3 is particularly interesting:

Each (dis)organization strategy is serving a different master, and that causes different problems. Netflix has sworn undying allegiance to its algorithm, believing the computer knows what you want to see better than you do. You can’t customize which genres or subgenres (“Witty Workplace Comedies,” et. al.) are displayed on the home screen, because the algorithm is doing the work for you. But the algorithm easily gets confused (my “Because you watched Star Trek” bar is full of foreign-language drama series, few of them with even a hint of science fiction), or takes the wrong lessons from what you’ve already watched — as in the case of a friend who can’t stop getting true crime recommendations because he enjoyed American Vandal, which existed to make fun of those kinds of shows. Netflix, like so many of the others, goes out of its way to make it impossible to find new movies or shows, since “new” is defined as “newly-added to this service.” So, actual fresh releases like Old Guard and EuroVision will be listed right alongside the 2010 box-office flop How Do You Know. Trying to skim by genre is useless, both because the algorithm picks what you see first, and because the same content keeps popping up from genre to genre. (As I write this, the original Alyssa Milano version of Charmed is listed as a TV Comedy, among other things.)

Netflix is an interesting case because this is a company, unlike all its competitors, integrally committed to a single service and its related user experience. They sweat over every detail, and deemphasising category browsing is a calculated decision. I wonder how much Spotify thinks similarly about balancing recommendation and browsing efforts…

Then again, there are differences. A video streaming library may be comprised of some 50 thousand titles, while music libraries have more than 50 millions songs. Similarly, video watchers need one or two good recommendations per day/session and music listeners can easily need dozens of good songs for the same period.

Another point is that there are two companies that are major players in both markets: Apple and Amazon. So it is interesting to observe how their video services are evolving. About Amazon Prime Video, Sepinwall writes:

“Why in the world does Prime Video still separate shows by season, so that you have to add them one at a time to your Watchlist?”

“Amazon has both the best bonus feature of any streamer — X-Ray, where pausing a scene gives you the name of every actor and character in that scene, the name of any songs playing, and other details — and one of the hardest interfaces to navigate, befitting a service that’s an afterthought in a much larger business empire.”

For Apple TV+, I quote Lucas Shaw writing for Bloomberg:

Apple TV+
What’s gone right: A buzzy show out of the gate in “The Morning Show,” endless cash reserves and a willingness to buy movies that studios can’t release in theaters.
What’s gone wrong: The service hasn’t been able to sustain its early momentum, and is still looking for an actual programming strategy or target audience.
Verdict: When’s the last time someone asked you about a show on Apple TV+? The service is something of an afterthought — not as big a misfire as the Arcade gaming service or Apple News +, but not a clear hit like Apple Music. In the end, that may not matter all that much. Apple Inc.’s push into services looks good in an earnings release, where the company has reported booming revenue and proven to Wall Street that its growth isn’t done.

(By the way, this description of Apple TV by Dustin Curtis is just brilliant.)

Besides Apple and Amazon, there is Google, which finds itself in an odd position. Youtube is a leader in music and video, but not when subscriptions are involved. As Wikipedia summarises, their subscription strategy has been convoluted so far:

YouTube Premium (formerly YouTube Red) is a subscription service offered by the video platform YouTube. The service provides ad-free access to content across the service, as well as access to premium YouTube Originals programming produced in collaboration with the site’s creators, downloading videos and background playback of videos on mobile devices, and access to the YouTube Music music streaming service.[5]

The service was originally launched in November 2014 as Music Key, offering only advertisement-free streaming of music videos from participating labels on YouTube and Google Play Music.[6][7][8] The service was then revised and relaunched as YouTube Red on October 31, 2015, expanding its scope to offer advertisement-free access to all YouTube videos, as opposed to just music.[9]

YouTube announced the rebranding of the service as YouTube Premium on May 17, 2018, alongside the return of a separate, YouTube Music subscription service.[10][11] Later in the year, it was reported that YouTube was planning to make some of the original content associated with the service available on an ad-supported basis.

Three Thousand Albums of The Year (2013 to 2018)

Let’s have some fun with data!

The “Album of the Year List” Project by Rob Mitchum is a great data set in convenient spreadsheet format. From its GitHub page:

What is this project?

My yearly effort to collate the album of the year lists from music publications into a single spreadsheet. It started as a good way to catch albums I missed over the past year, and to have all the lists in one place for easy browsing. One could also use it to make some arguments semi-supported by numbers about music criticism and trends and so forth, if one wished.

The methodology:

It’s mostly data entry. When a publication posts a list, I put the top 50 in the spreadsheet. I only choose lists that meet the following criteria:
– A list voted on by a staff, not a single person.
– A list that covers multiple genres, not just metal, rap, etc.
– At least 50 albums are listed (though I only take the top 50).
– It’s a website or magazine that I’ve heard of (or it looks legitimate).
I just list albums 1-50 according to their rank on the list, no 50 points for 1st, 49 for 2nd funny business. So low numbers are good, like golf. If a list is cowardly and unranked (looking at you, NPR), I just give every album listed 25.

There are lists for the years between 2013 and 2018. I did a bit of number crunching (actually just number counting) and found that the six annual lists of top 50 albums lists add up to to a total of 174 lists where 3,160 albums are mentioned.

  • 1,902 albums are mentioned in a single list, 60% of the total.
  • 466 album appear in 2 lists, 14,7% of the total.

That’s almost 800 albums mentioned in at least three top 50 year-end lists!

There are five albums that appeared in 30 lists or more, all of them in 2016:

  • David Bowie – Blackstar
  • Frank Ocean – Blonde
  • Solange – A Seat at the Table
  • Radiohead – A Moon Shaped Pool
  • Angel Olsen – My Woman

A summary by year:

Top 50 # Lists # Albums
2013 24 441
2014 35 588
2015 20 420
2016 36 602
2017 27 549
2018 32 560
Total 174 3160

Mitchum also lists albums mentioned in top 10 lists. There are 170 top lists collected from this period and 724 albums are mentioned. That’s nearly a quarter of the top 50 albums.

  • 462 albums appear in only one top ten list, 64% of them.
  • 106 albums appear in two lists, 14% of them.

That’s 156 albums mentioned in at least three top 10 year-end lists, an average of 26 albums per year.

There are five albums that appeared in more than 17 lists:

  • War on Drugs – Lost in the Dream (2014)
  • David Bowie – Blackstar (2016)
  • Frank Ocean – Blonde (2016)
  • Beyonce – Lemonade (2016)
  • Kendrick Lamar – DAMN. (2017)

A summary by year:

Top 10 # Lists # Albums
2013 23 96
2014 33 147
2015 19 92
2016 36 117
2017 27 131
2018 32 141
Total 170 724


A graph of total albums by number of lists they appear:


I am quite sure there must be a lot of good to great music in those, 150, 700 or 3,000 albums. Imagine how cool would it be to go through this universe of music without having to search each album one by one and then add them to one (or a hundred) arbitrary playlists first…

My Dream Streaming Service

Everybody dreams about the perfect streaming app right? Right? Or is it just me? Anyway, here is a list of my favorite features from the current apps I am familiar with:

  1. The amount of information from Spotify (number of plays for every song, playlist; number of followers for every artist
  2. The like/heart/add button is always visible (Deezer)
  3. Easy way to share/search user playlists (Spotify)
  4. Music videos (Tidal)
  5. Curated playlists (Apple)
  6. The name of the editor of curated playlists (Deezer)
  7. Archive of previous versions of a playlist (None)
  8. Deep integration with (Spotify)
  9. 5-star ratings besides hearts/likes (Apple)
  10. Streamed songs are automatically saved to a cache and available in offline mode (Qobuz)
  11. Offline mode easy to navigate and visualize (Apple)
  12. For any song playing, easy to see a list of related recommended songs (Tidal used to have, now hidden in mobile app)
  13. Spotify Connect (Spotify)
  14. Autoplay similar songs (Spotify, Apple in iOS 14)
  15. Real music recommendations by particular people with a particular taste and not particularly trying to “maximise engagement” (Qobuz)
  16. Synchronised queue across devices (Deezer)
  17. Booklets (Idagio, Qobuz)
  18. Track/album credits (Tidal)
  19. Daily mixes (Spotify, Tidal)
  20. Good organisation of classical music (Idagio)
  21. “Kids” mode, “party” mode where the music played does not “pollute” the recommendations (none)
  22. Easy way to see, select, change streaming quality (Tidal, Deezer)
  23. Friend activity (Spotify Desktop)
  24. “Flow” (Deezer)
  25. Support for anthologies (none)
  26. Discover Weekly (Spotify)
  27. Lossless audio (Tidal, Qobuz, Deezer)
  28. HD audio/MQA (Tidal, Qobuz)

07/29/20 Update: a Rquiem for Rdio

Highlights from Spotify Investor Day (March 2018)

Spotify Investor Day, held in March 2018, right before Spotify’s stock Direct Listing.

Here is the full presentation, a 2 hour video.

Here is the full slideshow, 152 slides. (By the way, the site used to host the files, works really well.)

Some highlights from the slides:

  • 3,500 employees, 40% of them R&D
  • 35 MM tracks
  • >200 petabytes of data
  • >250 partners
  • 30% U.S. music listens happens in the car
  • >50 million use Spotify in the car
  • 3 billion music related events collected every day
  • 49 minutes per day spent on Spotify by the average user
  • 30%+ of streams are “algotorial”
  • 60% users converted from free channels into premium subscribers
  • 3+ million artists
  • 20,000 new pieces of content daily
  • 22,000 “top tier” artists in 2017 
  • 10 billion discoveries per month
  • BTS had 4.5 million monthly listeners then (14.4 million as of July 2020)
  • Premium subscribers use Spotify on average 80 minutes per day

I loved this graph:

Pasted Graphic 0

Some data from a shorter video about the business model:

  • “Radio is a $28 billion global market ripe for disruption”
  • 61% of MAUs are 18-34 year olds
  • R&D: 10% of revenues in 2017
  • “As of January 2018, Spotify is the number one app on time spent by Millennials”

This is funny: Beatles and Taylor Swift: This was then, this is now:

Pasted Graphic 1More interesting graphs:


Suggesting a Feature for Suggested Songs

If I could choose one feature to improve on music streaming apps, it is a recommendation of next tracks.

Tidal used to have a very nice implementation in the mobile app: in the main player screen, swiping right would bring a list of suggested tracks based on the song currently playing, each with a button to add to queue.

I think it is a great way to help users choose next songs, discover new or obscure stuff etc.

Unfortunately Tidal now uses swipe right to forward next track (even though there is a huge button for that on the screen) and the suggested tracks are hidden in the “…” button, where “Go to Track Radio” is the fourth option.

This option brings a list of suggestions, but the only action button is Play this song based “station”. To add a track to the queue you have to press its “…” and there “Play Next” is the fourth option.

It is a shame, we lose the option to easily and frequently discover new stuff without interrupting a listening session. Starting a song based station is something I do very rarely, while adding stuff to the queue I do all the time.

It is comprehensible that the services want to make it everything as easy and automatic as possible for passive, casual listeners but the relationship with a music service is a long term one. As more and more users become increasingly familiar with the services, they will want a bit more of power and flexibility for shaping their listening habits. “Suggested tracks to add to the queue” is a tool that could become very popular if done right, IMHO.