4 THINGS EVERYONE
GETS WRONG ABOUT
4 THINGS EVERYONE
GETS WRONG ABOUT
Playlisting, Instagram Ads, TikTok videos…
So many strategies have sprung up around the fact that everyone wants to blow up on Spotify…
There are countless costly tactics being advertised to musicians that promise more streams and playlist placements...
Unfortunately, these strategies often boil down to overnight success tricks that are too good to be true...
...because they aren’t (true).
For those who have tried to grow their Spotify following, this isn’t news. They’ve either been lied to or know a fellow musician who has been lied to…
They’ve bought placements on playlists named after popular movies. Or they have 50,000 follower bots from Argentina that pay out $0.0009 per stream…
We’ve all been duped in this new streaming economy.
My name is Kyle Lemaire and I am the Founder & CEO of the music marketing company, Indepreneur. Through Indepreneur, I’ve had a hand in growing sustainable businesses for artists big, small, and inbetween; from million-follower YouTubers to famous on-screen talent to local independent musicians.
I don’t say any of this to brag - just to show that I’ve had full-contact experience with the best (and the worst) music marketing strategies...
In this article, I’ll “give away the sauce” on 4 categories of myths and misconceptions about the inner workings of the Spotify platform...
I’ll also explain the growth strategies that I’ve seen actually help artists on the platform, and the dangerous foolishness that artists need to be aware of in order to avoid costly scams and stagnant releases...
1. How The Spotify “Algorithm” Actually Works
Did you know that there is no central Spotify “Algorithm”?
Spotify actually employs a series of machine learning models. These can be thought of as algorithms that recommend music to users. This collection of processes make up the Spotify “Recommendation Engine”.
The Spotify Recommendation Engine is a technology that many artists, managers, and marketers don’t understand...
Everyone is so excited to explain their theory about how it works that they forget to actually do the research...
Artists should know how it works. Then, they can form their strategy and efforts around how Spotify makes recommendations to users.
If you aren’t getting recommended to new users on Spotify, whats the point of being on Spotify? It’s certainly not the pay-per-stream rate!
So let’s learn everything we can about how Spotify recommends music to users.
The Spotify Recommendation Engine is built using three machine learning methods:
The main machine-learning model driving Spotify's recommendations is called "Collaborative Filtering". This model begins with a big database (sort of like a spreadsheet) called a “matrix”.
The rows in the database represent Spotify Users in the form of User IDs. This means there are over 230 million rows!
The columns represent Songs on Spotify in the form of Song IDs. Which means there are over 60 million columns!
Using a math function called "matrix factorization", Spotify is able to calculate a "vector" for each Song ID and User ID...
Vectors basically show how the Songs and Users relate to each other.
By understanding the relationship between users and songs, Spotify can map the songs, artists, and users who are most relevant to each other. This allows Spotify to make recommendations based on the similarities.
For instance, if a song ID is very “close” in “vector space” to a song ID the user already streams, it could be recommended. Similarly, if a User ID is very “close” to another User ID, Spotify can use songs streamed by the first User ID as recommendations for the second User ID. Spotify can also recommend songs that are close in vector space to the songs you already stream.
The other two models powering Spotify's recommendation engine build on top of the “vector space” created in this first model. So, you may imagine that the position of each song and user ID in the latent vector space can be more accurate (or be “nudged” closer to their proper place) by adding other types of data to this equation.
Natural Language Processing (blogs, metadata, and playlists)
Natural Language Processing is an artificial intelligence process. LPN translates language in a way that a computer can understand. NLP models help with technologies like voice assistants and translation.
Using NLP technology, Spotify is able to comb thousands of authority sites and blogs with content about songs and artists.
Spotify is then able to determine which terms or words are relevant to each song or artist (including other songs and artists). So, if a blog wrote that a song sounds like Linkin Park, Spotify can understand that the song has some kind of relationship to Linkin Park songs.
After processing all the data it can find, Spotify assigns terms for each song ID and artist. This set of terms helps Spotify understand how songs and artists are related in ways that collaborative filtering can't, because it doesn’t depend on streaming data.
Eventually, Spotify realized that playlists aren’t too much different from a blog article when it comes to NLP. Playlists, like blogs, are a document containing words that are relevant to other words in the document.
Spotify now uses user-generated playlists as targets for the NLP model.
Spotify has over 2 billion playlists. Not all these playlists are good indicators of relationships between songs and artists.
Spotify only runs their NLP models on playlists they consider “good”.
Spotify's developers have stated that only playlists that have many listeners and frequent updates are considered "good" playlists.
Because of the way this particular model works, it is clear that blog PR, biographies, metadata, and playlist placement campaigns which secure you placement on "good " playlists all contribute to the likelihood that you will be recommended to Spotify users.
Audio Data (Signal Processing and Analysis)
The first two learning models, Collaborative Filtering and Natural Language Processing, need streaming or text data about a song in order to properly position the song in latent vector space. Without a vector, or a “position” in vector space, Spotify can’t understand whether a given user will like a given song.
Because of this data requirement, any newly released song on Spotify is unlikely to be recommended if the song has not yet been streamed or written about online.
Spotify's development team refers to this problem as the "Cold Start" problem. They have been working on solving this issue for many years.
As a partial solution to this issue, Spotify began analyzing the audio data (“spectrogram”) of new tracks.
Spotify uses a machine learning process to decode the characteristics of each song. These characteristics include the key, tempo, rhythm, time signature, timbre, and over a dozen others. Spotify uses this data to add an audio profile to each Song ID. Then, Spotify draws similarities between this audio data and the data of other Song IDs.
Through this audio profile, Spotify can accurately position a song ID in latent vector space even if it has not been streamed or written about online.
Audio data helps Spotify know which users' Release Radar should include the track. When you submit your song from the Spotify for Artists dashboard for consideration in playlists, Spotify performs the audio analysis and includes the track in its Recommendations Pipeline.
NOTE: You must submit your song at least 7 days before release to qualify for audio analysis on the first week of release.
2. How To Work With, Not Against, The “Algorithm”
Once you are aware of these three models, you can actually tune your strategy for Spotify to what you actually know about the Recommendation Engine:
Directive 1. Optimize Relevant Terms (NLP Model)
Ensure that Spotify has accurate metadata for each song in your catalog. Check with your distributor.
Ensure that your biographies and track descriptions listed on music websites contain an accurate description of your sound.
List the artists whose fans would like your music. Then find their biggest blog placements online. Use any terms that appear in multiple articles for each artist to describe your own music in biographies and online profiles. This will help Spotify understand that your music is related to those artists.
Directive 2. Maximize Streaming Data For All Songs (Collaborative Filtering Model)
Ensure that all of your fans existing on other platforms (or not currently listening to you on Spotify) are aware of and encouraged to follow your Spotify profile.
Use campaigns that cover multiple channels (retargeting campaigns, email, Messenger, social media, etc.) to not only direct traffic to your Artist profile for the first time but also form a habit of listenership with your audience. For instance, you may schedule a set of weekly releases that comprise a narrative, rewarding fans for following along.
Directive 3. Increase Streaming Data Through Ads (Collaborative Filtering Model)
Once you have maximized your existing listener base, test Facebook and Instagram ads. Allocate a reasonable monthly budget you can sustain month-to-month. Find the best performing audiences, creative, objectives, offers, and destinations that deliver new listeners and followers. (For more help on this, see our Spotify Field Guide)
Establish a way to follow back up with listeners before sending them to Spotify (i.e. Facebook Pixel or Custom Audience, email opt-in, etc.). Ideally, you will only use warm audiences (people who know you already) to send to Spotify. However, cold audiences (people who don’t know you) can still perform well.
Directive 4. Increase Relevant Terms For Major Releases (NLP Model)
For tentpole releases, such as a Studio Album, Collaboration, or Single, use playlist placement and blog PR campaigns, according to your available budget. This may increase your chances of impacting Spotify's Natural Language Processing model.
Directive 5. ALWAYS Submit Songs to Spotify (Audio Model)
Whenever releasing songs to Spotify, always upload to your distributor at least two weeks before your actual Release Date. Then Submit The Song for playlisting at LEAST 7 days prior to your release date by:
- Navigating to artists.spotify.com
- Clicking "Music" > Upcoming > Submit This Song (next to the track)
- Filling out the submission form using accurate data
3. Playlists, Playlisting, and Playlist Promo (Scammers)
If you don’t understand Spotify’s playlists and how they affect your success, it is easy to be scammed by a promotions company. The more informed you are about Spotify playlists as an artist, the easier it will be to use them to your advantage. And, the less you will fall for silly nonsense.
Let’s dive deeper into some misconceptions about Spotify’s playlists…
There are many algorthimically generated playlists, including those generated based on occasions and moods. There are only two which can be directly impacted by your marketing efforts: Release Radar and Discover Weekly.
A unique Release Radar playlist is generated and updated every Friday for every Spotify user. The Release Radar playlist update comes with an email notification and front-row placement on your Spotify homepage. This playlist is one of the few instances where Spotify will give your listeners an email notification about your tracks.
Release Radar has two criteria upon which the recommended songs placed on the playlist are based:
1. Release Radar songs must be relatively new (2-3 months old)
2. Release Radar songs must be candidates for recommendation to the user who they are recommended to (meaning, they are close to the user in latent vector space based on existing listening data, natural language processing, or audio data)
Release Radar is the only place where the audio data about your song plays a primary role.
Besides users who your song is recommended to based on the song's audio data, your monthly listeners are also likely candidates for early Release Radar placement. This is because, while your song does not yet have a place in latent vector space, your listeners are already shown to enjoy releases from your Artist ID. So, they are natural choices for recommendation when you have a new release.
Release Radar distribution is cumulative. If users save your song after hearing it in their Release Radar, you are more likely to reach more users on the next Release Radar. As far as we can tell, your song's Release Radar performance will peak in the 3rd (sometimes 4th) week of being out. By the 12th week, your song is very unlikely to receive any Release Radar distribution.
A unique Discover Weekly playlist is generated and updated every Monday for every Spotify user. Discover Weekly has two criteria for recommended song placement:
1. Discover Weekly songs have never been played by the user to whom they are recommended
2. Discover Weekly songs must be close to the user in latent vector space. This is based on existing listening data, natural language processing, or audio data.
If these two criteria are met, songs are included in the Recommendations Pipeline for Discover Weekly. At the time of this writing (and, as far as we can understand), Discover Weekly trains its selection algorithm every week with a process that involves playlists and positive/negative feedback from users.
Discover Weekly distribution is the result of four important factors:
- popularity ranking
- playlist placement
- blog coverage/metadata
- streaming data
The more popular your song becomes, the more streaming data it will accumulate. Streaming data will help Spotify place your song accurately in "latent vector space".
Placement in latent vector space (as a result of users streaming your song) increases your likelihood of being recommended to the right listeners. If Spotify is correct in recommending your song, it's likely that the recommendation will continue to more users over time.
Playlist placement and blog coverage/online metadata both help to create an NLP profile for your song. The profile is based on relevant terms, artist names, and song titles. These terms help Spotify accurately calculate your song's place in latent vector space. Playlist placement will help your song accumulate more streaming data. This will also help to accurately calculate your song's place.
User Generated Playlists
Since the beginning of Spotify, user-generated playlists have contributed to the value of the platform. If it were not for user curation, Spotify would have a much tougher time competing with Apple Music.
However, user-generated playlists have also reintroduced an era of pay-to-play scams in the music industry which are now technically "pay-to-place" scams. To put it simply: if you want to get your music into playlists with top visibility, you're probably going to pay for it. Many user-generated playlists are owned by playlist networks. These networks offer pay-for-placements and playlist campaigns that span multiple playlist placements.
Another benefit Spotify receives from its user-generated playlist curation is a very large data set - over 2 billion playlists. Spotify can train machine-learning models with data to make better recommendations to users. This occurs at two levels (that we know about):
1. Spotify uses user-generated playlists to train Discover Weekly's model
Spotify uses real user-generated playlists to train machine-learning models to guess missing songs in random playlists. This same model is then used to make recommendations for songs to users based on their recent listening habits.
2. Spotify uses user-generated playlists to find similarity between artists and songs
Spotify employs a natural-language processing engine to scour the internet for articles about songs and artists in the Spotify library. It then analyzes articles and other text for relevant terms associated with each song and then stores the relevant terms. In doing this, Spotify is able to find similarities between songs based on language. Spotify uses this same model to analyze it's own playlists (as text documents) to find similarities between two or more songs or artists (or both).
However, Spotify first ensures that any playlists used in method 1 and 2 are "good" playlists.
As far as we understand, Spotify's main criteria for "bad" playlists are those which have very few listeners and are not frequently updated. Placement on "good" user-generated playlists is necessary to maximize your discovery potential.
If you want to maximize your placement on user-generated playlists, you must gain the skill of researching, validating, reaching out to, and paying for placement on those playlists.
What you should look out for are user-generated playlists that are popularized using sketchy tactics and bot traffic. These playlists represent corrupted data that is not useful to Spotify's recommendation engine. They are also primarily used by free users in countries where advertising CPMs are very low. Thus, the payout for streams generated from these playlists can be shockingly low (often 1/5th the payout of a premium user stream).
By researching good user-generated playlists that have an obvious positive impact on its featured artists, you can drastically increase the streaming data associated with your song, pick up new listeners and followers, and plan a 1-to-2 year trajectory towards the long tail of profitability on the platform.
Unlike user-generated playlists and the two main algorithmically-generated playlists, editorial and algotorial playlists do not directly impact the performance of your songs in Spotify's recommendation engine.
It is imperative for artists to understand that there is not much strategy that can be employed to get on these playlists. This is another arena where musicians get scammed into thinking that a service provider can make them explode on Spotify.
Playlist Placement Scams
Many playlist plugging companies offer “guaranteed” placement in exchange for a fee from an artist or band.
There are pitfalls and dangers to partnering with these companies.
Many of them actually own the playlists they will place you on - and they most likely sell placement to non-clients, as well.
So, you may be able to save money by researching their playlists and paying for placement yourself.
However, some playlist promotion companies use a combination of owned and partner playlists to achieve a desired campaign goal.
When you enter the world of playlist plugging services, you run the risk of being scammed by direct bot traffic, indirect bot traffic, and just plain old getting your money taken like it’s 5th grade recess. Artists must practice careful diligence when choosing playlist promotion companies to help you.
4. Short vs Long Term Spotify Strategies
The last category of misconceptions you need to be cautious of is the idea of Spotify growth in the short term.
Every single artist we work with who has a large Spotify following spent years growing it. There is no 3, 6, or even 9 month path to Spotify success.
Strategies related to increasing your Spotify listenership and likelihood of recommendation are very unlikely to provide a financial return of any kind at any point.
Efforts to increase listenership and followers should be largely viewed as a long term strategy, as opposed to a shorter term cash-grab.
In short, Spotify growth is a marathon, not a sprint.
Has someone told you that they can help you blow up your streams or followers in the short term? You should run from that person.
Spotify may still feel like a minefield without a clear perfect strategy for growth.
While there may not be a perfect strategy, there is a roadmap to help artists navigate the waters of Spotify and understand the nuances of how the platform works and evolves.
Just like anything - if you do the work and learn, you can increase your likelihood of success.
There are many ins and outs of Spotify as a monetization platform for artists. It took our entire team many months to research the platform for our Field Guide.
The Spotify Field Guide was created as a resource that can end the confusion about what works on Spotify once and for all.
With the Spotify Field Guide, artists and musicians can learn the ins-and-outs of every feature of the Spotify platform. This information is plucked from presentations of Spotify developers and machine-learning engineers.
This roadmap aims to empower artists to unlock the data inside Spotify (and feed it more of the right data), build a long term strategy, and leverage the Recommendation Engine (instead of trying to “hack” the algorithm”).
For independent artists who are seeking to:
- Leverage Spotify’s features to gain an edge as an artist on the platform...
- Grow their fan base and revenue by activating the Recommendation Engine...
- Track their Spotify growth and stop wasting money on unsuccessful strategies, and…
- Gain followers, monthly listeners, and streaming revenue predictably and reliably
Just our way of saying “good job reading this whole thing”
The Spotify Field Guide
The Spotify Field Guide
The #1 Source for Accurate, Bullshit-Free Tutorials
on Spotify Marketing and Growth Strategies
The #1 Source for Accurate, Hype-Free Tutorials on Spotify Marketing and Growth Strategies
What is 'The Spotify Field Guide'?
An End To The Confusion About What Works On Spotify...
Lately, it seems like everybody on the internet is publishing information about Spotify. Even Spotify's business team is woefully out of sync with their developers and how the platform actually works. This leads to MASSIVE confusion about how to grow your Spotify following - we give you the straight facts from the mouths of the engineers who built the platform...
Thousands of Dollars Saved On Counter-Productive Bullshit...
(Almost) every strategy out there for growing your Spotify is complete and utter fantasy. We've sourced examples and case studies from artists and companies with massive Spotify results from Green Light, high-payout countries so you know you're on the right track...
Accurate Breakdown of The "Algorithm" (Straight from Spotify Developers)...
There's nobody who knows how Spotify's Recommendation Engine works more than the engineers who built it - which is why we have poured over 50+ hours of developer conferences, white papers, and articles - and put the facts into one place. FINALLY learn how the Spotify "algorithm" ACTUALLY works!
Step-by-Step Course With...
Over 4+ hours of video training, bonus modules, and downloadable resources dedicated to pointing you towards the right long term Spotify strategy...
And, Most Importantly...
HONESTY. Throughout this training, we recommend against misguided costly strategies. Skip the hype-y tricks and get real, in-the-trenches advice about how to approach Spotify growth - no "one weird hack", no "overnight stars", no b*******.
Learn The Platform
Unlock Deeper Data
Build Long-Term Strategy
Get Recommended More
This stuff just works... see what our customers have to say!
The Spotify Field Guide
Frequently Asked Questions
Q: What is an Indepreneur "Training"?
A: It's halfway between a tutorial (step-by-step instructions on how to do something) and a course (series of information that teaches you a skill).
Our trainings are built to get stuff done - fast. At the same time, they are meant to leave you with a valuable skill that you can use again and again throughout your career.
Q: Why $95? ____ charges over $200...
A: If you look at how our competitors price their information, you may be inclined to think that we're either generous or that we're hiding something...
Well, here's a couple reasons for our pricing model that should ease your mind:
1. This company was started by musicians and for musicians - not by black hat marketers trying to make a million dollars. We understand where you are at - and we think it's criminal what some companies are charging to basically gas you up and then help you set up a social media page. We want to change all that.
2. On the other side of the coin, charging anything, even if its less than $100, does weed out those who are less-than-serious. We only want serious musicians who will help us change the way this industry works.
We also believe that, when you see how much actionable info we provide for this price, maybe (juuuuust maybe) you'll join us as an INDIEPRO member with full access to our entire Training Vault.
But that's it...
No fine print... no "hidden trials"... no funny stuff. Just the information you need and the results you want.
Q: How do I know this training will work for [insert genre]?
A: Here's how it works at Indepreneur:
Each year, we spend over 1000 tech-hours and over $100,000 testing different digital marketing strategies on music careers. We do this with:
- Our Testing Grounds Artists - an assortment of collaborating musicians that helped to start Indepreneur and who give us full reign to try out our crazy ideas
- Our Agency clients - Indies who hire us to set up their marketing campaigns and for whom we get to try out strategies that have worked for our Testing Grounds artists
When something works (and I mean really works) we document the strategy in a step-by-step "training" so our best campaigns can be repeated and rolled out across all of our artist accounts.
These trainings are not like some online courses you may have taken from other music marketers - they're almost entirely tech trainings.
We're not here to give you some overpriced pep talk - our trainings equip you with the very best of whats working right now. All killer, no filler!
Q: Is there a guarantee?
A: Absolutely - ALL of our trainings have a 60-day, no-questions-asked guarantee that works like this:
You either love it, or you get your money back. Period.
That means, if you aren't happy for any reason, just let us know and we'll give you a full refund on the spot. No forms to fill out, no hoops to jump through, no files to upload.