While tinkering on a side project, I found out that Discogs dumps all of its release, artist, label and master release data. Under a cc0 licence too! The dump as of September 2026 comes to about 65.3GB of XML (when uncompressed). I wrote a little tool to parse the XML and format it into Parquet (from 65.3GB of XML to just under 10GB of parquet, woof) so I could have a look at the data with DuckDB.
I found some interesting stuff! And... some odd stuff. But i'll get to that at the end.
Some caveats apply, first! The data is not necessarily an accurate representation of the world but is more a representation of the world discogs users care about. It might seem like not an important differentiation, but does frame some of the graphs in a different way. For ex, whether electronic overtook rock in 2009 is an accurate reflection of the world or is more a reflection of the kind of genres discogs users care about cataloguing is an open question. Same deal with what countries are represented, things like that. The audience of discogs and the things they care about skew what gets catalogued.
Also, since its a community driven catalogue the data is not necessarily complete. I ran into this a lot when trying to map trends in recent years! You'd get something like File releases beginning a meteoric rise in the 2010s only to plunge in the last 2-3 years. I don't think this is everyone deciding to swear off bandcamp, I think its more that it takes time for users to backfill data and that the cataloguing of recent years is simply less complete than older years. To this end, some of the charts will end a few years ago. This is why.
With that in mind, let's have a look at some of the data!
Countries
There's lots of interesting information when it comes to what countries are associated with releases in discogs. In a basic sense, we can see the countries with the most number of releases logged:
Cool that that the USSR still makes the top 25 despite its... current existential difficulty. Similarly, we can also take a look at the countries with the fewest records.
Some fairly remote places! One stood out to me though - Saar. A country I had never heard of. With a German song title. Wait? Saar as in Saarland? Turns out that Leila Negra's Mach Nicht So Traurige Augen did in fact get a release in the Saar Protectorate, a french protectorate that existed since 1947 and was integrated into West Germany in 1957. Neat.
Just raw numbers of releases logged are not, by itself, super interesting and is also quite inflected by population numbers (especially in the case of the US). A more interesting result would be the number of records released per capita...
with Jamaica and Iceland being ahead of the pack! I had half expected the former but not the latter, but the tiny population probably has something to do with it. Still, they're productive up in Reykjavik.
On that note, it'd be cool to see what musical style is most over-represented in certain countries. This is what that looks like, with the countries ranked by how much stronger their representation of the style is versus the world baseline:
Perhaps not so surprising that the styles tend to be very local ones. Nice on Turkey for keeping the Europop flame alive, though.
Speaking of countries though... what about ones that are no longer with us? Can we see the fall of the Berlin wall or the dissolution of the USSR?
You can! You can also see the drawn out demise of Yugoslavia too. You can see the drop beginning in the early 90s from republics leaving, the brief period of Serbia & Montenegro and then Serbia afterwards, with records released under Yugoslavia stopping around 2003.
Release Details
Outside of countries, there's also some pretty interesting details about releases themselves in the data. For example, take the record duration of tracks:
There are spikes in track durations in groups of 5. However, the spikes are much lower than I would have expected! It seems people are diligent about recording track lengths pretty granularly.
What about musical genres? In the data we can see the battle between electronic and rock play out from the mid 90s onwards, with electronic overtaking rock in 2009.
But also super interesting is the rise and fall of different media formats.
I personally think flexi-disc will make a comeback. File releaese are skyrocketting, and the vinyl comeback in the 2010s is well shown. One thing that caught my eye, though, is the cassette comeback over the same time period. Although the raw numbers for cassette are still much lower than file, vinyl or CD... I wondered if cassette's comeback, relative to its lowest point, was as strong as vinyls.
Turns out, in relative terms, the cassette comeback is even stronger! Of course this is a factor of cassette's lowest trough being so low, but I was still surprised by this. Curious that cassette's lowest point and the start of its recovery started before vinyl's too.
It was also interesting to look at formats by country, to see which countries have more releases on particular formats against the world average.
Interesting that you can see different formats mapping to different economics and politics. 8-track being more of a Canadian thing than a US thing ultimately is kinda funny.
This made me wonder... which musical styles have the greatest proportion of releases in the modern era (from, like, 2018 onwards) in cassette. I'd have guessed noise, black metal...
Power electronics! Dungeon synth! I was happy to see, as a black metal fan, black metal still being represented. Which made me wonder....
Black Metal Tangent
What the format journey for black metal actually looked like.
Quite fun to see Black Metal start as very cassette heavy, then CDs take over, then everything kind of converging towards the end. Fun also to see the prevalence of CDr, too! I imagine pretty small bedroom recording releases.
Black metal releases per capita also tells a, perhaps, not so surprising story:
Quite cool to see Greece ahead of Germany, though. Denmark being the odd Scandinavian out, though.
Weirdness in outliers
Finally... something odd I found in the data. I was counting label credits across releases, to find the most prolific label on Discogs. Turns out, its the placeholder Not On Label, with about 746k releases, almost twice that of Columbia. If you include artist specific variants (like Not On Label (Depeche Mode), its about 1.6m releases, almost 8.4% of everything in the data dump.
I thought, hey, looking at the top artist specific "Not On Label" variants could be a fun map of who gets bootlegged. And, lo:
| label | releases |
|---------------------------------------------|---------:|
| Not On Label (Nosferatu 1922 Self-released) | 2,029 |
| Not On Label (The Rolling Stones) | 1,637 |
| Not On Label (Depeche Mode) | 1,437 |
| Not On Label (Iron Maiden) | 1,416 |All the stars are here! Rolling Stones, Depeche Mode, Iron Maiden and... uh? "Nosferatu 1922 Self-released"? What the fuck is that? Well, turns out Nosferatu 1922 is a self-releasing electronic artist with, uh, over 2000 releases spread over just 93 dates, including 175 released on a single day (2018-03-29).
Well, okay. Whatever. Let's look at something else! What other weird outliers can we find. I wonder what the the release with the longest tracklist is. Turns out, that belongs to Lord Poozix's Grim and Frostbitten Penguins, clocking in at 6,666 tracks. Someone having a lot of fun.
Ok, ok. Whats the artist with the most aliases? Well, that turns out to be Merzlux, with 2103 aliases.
Well, these are all not randomly disconnected facts. Nosferatu 1922, Poozix and Merzlux are all mutually aliased. A traversal of the alias graph gives a connected component of 2151 artists. These include:
Every one of the 1532 artists with 1k+ aliases in the database.
1455 artists with the pattern of "<Name> Franklin, <n>".
This all seems to be the work of someone named Alex Usiel Ischenkau, who seems to be involved in 'spamwave, metawave and harsh noise wall'. Not sure what to make entirely of this! It doesn't seem to be vandalism, more of a Project, but I don't really feel like digging in to several thousand tracks of self professed harsh noise spamwave.
Run-Out Groove
Playing around with the discogs dump (and parsing and translating it into a more queryable format) has been pretty fun, and was a nice way to also play around with DuckDB which I hadn't really touched before. There was lots of stuff in the data that I didn't get round to putting in here for the sake of brevity, such as the fact that it seems Europe produces way more vinyl for the world than I had realised, or that George Peckham has eteched "A PORKY PRIME CUT" into 21,583 releases.
I'd love to share the parquet data but I'm not sure how! Once i've cleaned up my XML -> Parquet parser and translator I'll publish that too, in case you want to throw the data in there yourself and play around with it.