Metadata is the birth certificate of your release. It includes all the information third parties need to find and maximize your release potential. Without proper metadata, your track is lost in the sea of music available to the public and that's the first reason why you should care about it. It's important on release day, but not only then!
Just like a birth certificate, metadata comes in handy quite often in a track’s life and should be as accurate as possible. So now that we get the concept, what is it really? When we say ‘metadata’, we are using an umbrella term which includes many subcategories.
ID numbers
The first metadata category you need to know is the ID numbers. They allow you to tag your song and your master with a unique code, making it identifiable and retrievable in any circumstances. There are three main ones you need to know about and keep in a corner of your mind after release.
First of all, your ISRC. This ID code refers to a unique master, meaning a five-track EP will have 5 ISRCs. Still on the master side, your UPC is your next essential ID code. It functions like a barcode and allows you to identify an entire release. Using the previous example, a five-track EP will only have one UPC that identifies it.
Now on the composition side, you will need your ISWC. This ID code functions just like an ISRC but allows you to identify a composition, not an audio recording. Now let’s say I am releasing a cover of “Let’s get it on” by Marvin Gaye, my cover and Marvin Gaye’s original recording will both have the same ISWC since the song is the same, but the ISRC will be different as it is two different masters.
Technical descriptions
The second category we will touch on is the technical descriptions associated with the release. This information is anything and everything relating to the hard facts about your music, such as tempo, instrumentation, musicians, language, etc.
They are purely informative and allow collaborators to be credited as well as your music to be searched based on specific criteria. There are several instances in which your track could be missing a growth opportunity if it does not include its technical descriptions! For example, if someone were to filter a playlist by BPM, or lyrics language and your track did not identify those correctly, you would automatically be counted out even if your track qualifies. Accurate metadata helps you stay in the run of the competitive music business.
Creative descriptions
The final and most fun category is the creative descriptions of your track. For that one, each distributor or library will have their own requirements. Some bigger distributors might ask for a detailed description of your song to pitch to DSPs, some libraries will want keywords that they can push for SEO and some smaller distributors will stick to genre and mood, so there is no one-rule-fits-all here.
Some of the creative descriptions you might encounter include keywords, track description, mood as well as genre and subgenre. Those are also crucial because, picture this: you wrote an amazing Soul song that you pitched for a library. Your track is very retro, has a strong bassline and a gospel choir, which is exactly what this new upcoming movie director is looking for.
Now, there are hundreds if not thousands of production music libraries, with each including hundreds if not thousands of tracks, but luckily, the music supervisor for this movie is in the library that picked up your track. The only thing that will allow them to find you at first screening is SEO! Using the right words and describing the right feel might get you the gig, so take your time and do it as thoroughly as possible.
Why should I care about music metadata?
By now, you've probably got the gist of it: metadata is what makes your song traceable, easy to find and gives it an identity to people and software that will read about it before listening to it. It’s your song’s passport, or its resume. However, when exactly does it matter and what opportunities does it give you? We went through a couple examples above, but we will dive into it a little deeper in this second part.
Payouts (streaming and PROs)
First and foremost, get paid and recognized for your work. Accurate metadata allows all players to be credited properly and hence have a claim on royalty share, mechanicals and public performance revenue. The music industry has a serious payout problem emphasized by the lack of (or limited access to) a centralized PRO entity in a continuously more globalized world.
This ecosystem is already tricky to navigate when all writers' and publishers' shares are correctly computed in the system, so imagine when they are not… A typo, wrong publisher or missing name would greatly impact someone’s ability to get paid for their work. Also, as credits are crucial in the industry, being credited as a guitarist or a producer in a track that later on gains popularity could change someone’s life, so keep that in mind when filling your metadata.
Track management
Now, a track does not die on release day; it’s quite the opposite. Ideally, we want it to thrive and have many opportunities to be reborn again and again. To do that properly, you need to be able to keep control on your master and your composition separately, which gives you freedom on how to manage your track.
We previously mentioned the ISRC, which identifies your master. Any time you might want to do something new with your master, you will need your ISRC. As long as there are no changes in the recording, that code will allow you to take down your song from platforms, change distributor, include a previously released single in an EP or in a compilation-type album…
If you ever want to switch distributors and keep your streams, having the same exact metadata and using the same ISRC is mandatory. Basically, it allows you to move your master around but also to identify a new master for the same song. If you were to do a remix or a remaster, your ISWC code associated with the composition would stay the same, but the ISRC would change. That means one song can have many ISRCs, allowing you to know how many existing masters there are of that one composition.
Playlists
Playlists are what every releasing artist strives for, and rightfully so! They are an amazing tool for getting your music out there and that is why you should make sure curators can find you. Genre and mood are the glue tying a playlist together, meaning that is probably what curators will search for when looking for the missing piece of the puzzle. Of course, there are many aspects that go into curation, and matching the subgenre a curator is looking for is not the only thing you should be focused on, but it is a start. Your metadata goes directly into DSPs making all that information available to curators, increasing your chances to be heard and featured in the playlist you want.
However, curator playlists are not the only types of playlists on DSP! Algorithmic playlists are also huge and can get you many streams and introduce you to potential new fans. Spotify uses machine learning and natural language processing to process all streaming data and tracks’ metadata, ultimately creating the playlists we all know and love. As this is purely technical and can’t be pitched, metadata is all the algorithm will hear. There are many ways to increase your chances of getting on algorithmic playlists, but your metadata is the foundation.
Music Libraries
Not all music is made to be released on DSPs, but that does not mean metadata is not an issue. If you are a composer looking into production music, then you should polish your metadata until it shines. Production music is a highly competitive environment, and different criteria ranging from share splits to recording quality to accessibility will impact the likelihood of your track to be picked up.
But let’s start with metadata and SEO. When someone is browsing a music library, they are most likely looking for something precise, but willing to get inspired. They might know the genre they want but not have a certain instrumentation in mind, or they could be certain of everything. Point is, metadata is the first thing that will bring them to your track, then the quality of the music is what makes them stay.
When you are pitching a track to a music library, think as a director. How do you picture your track? What keywords come to mind when listening to it? How can you describe its feel? It could be romantic, heroic, ethereal… Think about texture, vocals, story—all those elements are key to make sure you are found. Also, do not forget technical elements! A scene might have a certain tempo to it, meaning a director might want a certain BPM. It is by being thorough that you will increase your chances!
Tools
Your metadata also allows you to pop up on the music industry tools, such as Chartmetric or MusicBrainz, which allows professionals to find you and get information on your catalog. Chartmetric is a great A&R tool as it tracks your success and your charts with many different filters. If your R'n'B track has had significant growth in the last month, you can only benefit from it being apparent on such websites. This might not be the first use for metadata, but again, another reason why it’s important.
Smart Speaker / AI Assistants
Last but not least, your metadata gives your track a shot to be streamed whenever someone is using a smart speaker such as Alexa or Google Home to play music. A query such as ‘Alexa, play me a sad chill R&B song’ relies on the fact that a track is tagged as R&B, chill and sad. Many factors enter into defining that metadata, and a lot of it ends up being generated on the spot, according to the report “Everybody’s Talkin’ Smart Speakers & their impact on music consumption,” which dives into the subject deeper.
However, tagging your music as specifically as possible and with great attention to detail puts you ahead of the curve and gives you a say on what commands bring out your music. In the report, we see that “In January 2018, NPR and Edison Research’s ‘Smart Audio Report’ claimed that one in six Americans now owned a smart speaker – 39 million people,” making it a very considerable niche to guarantee yourself a space in.
What is music metadata’s future?
Now that we’ve established its necessity in today's music industry, how will metadata be impacted by technology’s progress? Technical descriptions of music are already less dependent on human input and expertise, as supported by Nadim Kawwa’s article “Can We Guess Musical Instruments With Machine Learning?” where he demonstrates that it is possible to guess a musical instrument using simple machine learning algorithms written in Python with up to 85% accuracy. The models are not perfect and still require quality control, but if you are working with large sets of music metadata, those models can be game changers.
Spotify and other DSPs also have been using machine learning, which allows them to offer recommendations based on your listening habits, for example. To simplify, by processing existing factors of a track, those models are able to place the track on a Cartesian map with other musical works. Genre classification models also already exist, as Kill the DJ mentions in the article: “The genres linked to an artist are influenced by a myriad of nuanced factors, including "tempo," "duration," "color," "modernity," and "femininity." The algorithm then evaluates the artist's similarity to every genre within Spotify's system.”
But how about creative descriptions? Captioning models such as BLIP are used to describe images, and some of that technology is getting to music too, as studied by Ilaria Manco, Emmanouil Benetos, Elio Quinton and György Fazekas in their paper "MusCaps: Generating Captions for Music Audio." Considering how imminent those changes are, getting ahead of it allows you to keep a competitive advantage through increased efficiency in metadata computation.
As innovative and exciting as this is, those models still have some limitations, as pointed out by Stephen Brade, a music technology researcher who will start his PhD at MIT next fall. He says that, “Machine Learning is very dependent on the quality of your initial dataset meaning the less access to accurate metadata you have, the less success you will have with those models.”
Another crucial point he made is how brittle those models are in the face of musical innovation: “If a song is very different compared to what already exists, chances are they would not be properly classified." Computing your own metadata is also a way to have a say on what you are creating without having to compare it to what already exists, allowing music to keep evolving based on artists’ perceptions and without constraints. Introducing your music in your own words also allows those models to build on databases influenced by human sensitivity. Finally, including the way you experience a song might be valuable when you are appealing to another person’s sensitivity. It is your perception that might catch a music supervisor's attention as it is through their human experience that they can relate to your words.
According to Stephen Brade, the common point in all this does seem to be that music is moving away from set labels and more towards latent spaces where a track’s description is directly impacted by its similarity to other tracks. We can wonder if the process we know today as providing metadata for your track might move towards matching your track with others based on your perception of subjective musical factors.
Music metadata: Wrapping it all up
Metadata has three main categories, all important for different reasons, and being thorough in all of them is the first step to success. Some elements are consistent and will always be asked whenever you need to fill out a metadata form, while some elements will be dependent on how you are exploiting your track and which third party you are using to do so.
Point is, the more you fill out and the more detailed you are, the more control you have over your work. From getting paid to getting visibility, your track relies on its metadata to maximize its potential, which is why it should be computed properly. The rapid changes in technology and their impact on the music industry imply a shift in traditional music metadata, but considering the subjectivity of music, describing your track in your own words is still the most reliable way for it to be perceived as you intend.
