How to organise a DJ music library that will still make sense in five years

Put every track you own in one folder on one drive and never move it. Let the tags do the organising rather than the folder tree: genre, BPM, key, energy and a comment in every file, written into the file itself so that whatever software you are using in 2031 can read them. Keep the lossless master where you have one. Back the folder up to a second drive and to somewhere off-site, and treat your DJ software’s database as a view of the collection rather than the collection. That is the whole answer to how to organise a DJ music library, and it is the same answer whether you have 400 tracks or 40,000.

The reason it needs a longer explanation is that every one of those steps has a way of going wrong that only shows up a year later, usually the night before a gig. The rest of this is the detail.

One folder, one drive, chosen once

Rekordbox, Traktor and Serato all store the absolute path of every track. Move the folder, rename the drive, or plug the same disk into a machine that gives it a different letter, and every track goes grey. All three have a relocate function and none of them is fun, so the first decision is where the library lives, and you make it once.

For most people that is an external SSD. It survives a laptop change, it can be plugged into a second machine, and it keeps the music off the system drive. Give the drive a short name with no spaces and keep the top-level path short, something like DJ/Library, because long paths with punctuation in them are where Windows and macOS start disagreeing about what a filename is.

Inside that folder, do not build a genre tree. Genres drift, tracks sit between them, and a folder can only hold a track once. The structure that ages well is boring: one level of subfolders by year of purchase, or nothing at all. Its only job is to say where the file is. Everything else is a tag.

A second folder, outside the library, is the inbox: downloads that have not yet been renamed, tagged or checked. Nothing moves from the inbox into the library until it has been processed, and the inbox is empty at the end of every filing session.

Filenames that sort themselves

Shops all name files differently, and promos arrive as whatever the label’s intern typed. Pick one pattern and rename everything to it before it enters the library, because renaming after import is the same as moving.

Artist - Title (Mix Name).ext is the pattern most DJs settle on, and it is the one that reads best on a CDJ screen and in a file browser. Keep the mix name, because the original and the extended mix are different files with different intros. Avoid characters that one operating system or another refuses: slashes, colons, question marks, asterisks and quotation marks. Feature credits go in the title, not the artist field, or your artist list fills up with one-off collaborations.

Tag editors such as Mp3tag can rename a whole folder from its tags in one go. That is the direction to work in: fix the tags, then generate the filename from them, not the other way round.

Tags carry the meaning

The fields that matter, in order of how much they will save you later:

• Artist, title, mix. Correct, consistently capitalised, no “feat.” in the artist field.
• Genre. Your genres, not the shop’s. Beatport has dozens and they change. Write down ten to twenty categories that match how you actually play, and use those. If a track sits between two, pick the one you would reach for it under.
• BPM. Written by your analysis software. Check the half-time and double-time mistakes on anything below 100 or above 160.
• Key. Analysed once, in one notation, in every file. If half your library is in Camelot and half in Open Key, a filter on 8A misses every track tagged 1m. The Camelot wheel article covers the notations and where detection goes wrong.
• Year and label. Cheap to fill in, and the only way to find “everything on that label from 2019” three years from now.
• Energy. A number from 1 to 5 or 1 to 10 that says how hard the track hits, in your judgement. No analyser does this well. It goes in the comment field or the grouping field, somewhere that is written into the file.
• Comment. A short code you can read in the booth: the energy number, the role of the track (opener, peak, tool, closer), a word about vocals or breakdowns, and when you filed it. Something like “E7 peak vox / long intro / 2026-03” tells you more on a dark stage than a star rating.

The rule under all of this: prefer fields that are written into the file. Rekordbox’s My Tags and colours live in its database, not in the audio file, so they vanish when you switch software or lose the database. Comments, genre, key, BPM and grouping travel with the file wherever it goes. Unfamiliar terms are in the glossary.

Formats and the lossless master

Store the best copy you were sold. A WAV, AIFF or FLAC is the master, and you can make a 320 kbps MP3 from it any time; the reverse is not true. The one format point that affects organisation is that WAV has historically been the least reliable of the three for tags, so if you are buying lossless for a Pioneer setup, AIFF carries tags more predictably. FLAC support on club players depends on the model, so check before you rely on it. The full comparison is in the formats article. If you hold both a lossless master and an MP3 of the same track, decide which one plays and keep the other outside the library folder, or it shows up as a duplicate on every search.

What Rekordbox and Traktor do behind your back

Every piece of DJ software keeps a private database of analysis, cue points, beatgrids, playlists and history, and none of them agree about what belongs in the file.

Rekordbox keeps almost everything in its own database, and whether your edits reach the file tags depends on the version and the settings, so the safe assumption is that they do not. A key you corrected, a colour you set, a My Tag you added: all in the database. Its library backup, made inside the app, is the only thing that captures cue points and grids; a copy of the music folder alone will not restore them. Exported USBs carry their own analysis files, which is why editing a track on the stick and on the laptop produces two different versions.

Traktor stores its collection as an XML file, collection.nml, in its own folder, and by default it writes most tag edits to the file, including key in Open Key notation. Cue points and grids stay in the collection unless you turn on the option to write them to the file, which not everyone does. When Traktor imports a folder, it can overwrite tags with its own analysis, so run it on a copy the first time and see what it changes.

Serato keeps its crates in a Serato folder on the same drive as the music and writes cue points into the files themselves, which is why Serato cues survive a reinstall better than most.

The practical consequences: back up the software’s database folder as well as the music, do your tag editing in one place and let the others read it, and never let two programs analyse the same file with write access at the same time.

A backup that has been tested

The standard is three copies, on two kinds of storage, with one somewhere else: the working drive, a second drive synced after every filing session, and a cloud copy or a drive at someone else’s house. The music folder, the software database and the analysis data all go in, because losing the database means re-cueing every track you have ever played.

Sync tools that mirror deletions will happily mirror a mistake, so either keep a versioned backup or let the second drive lag a week behind. Once a year, restore the whole library to a spare machine and see if it opens. A backup you have never restored is a hope. The USB stick you play from is not a backup either; it is a subset prepared for a player, and the thing most likely to be left in a booth.

Duplicates and the annual clean

Duplicates get in through the front door: a track bought twice on two releases, a Bandcamp purchase downloaded in FLAC and again in MP3, a promo later bought as the same file. Left alone they clutter search results and split your cue points between two files.

ID Crate’s library indexer reads the tags of a folder in the browser, without uploading the files, and its duplicate detector lists the tracks that appear more than once so you can decide which copy stays. The same index gives you a key browser for seeing where the library is thin, and a comparison against your Spotify account that shows the tracks you keep playing there but do not own. Whichever tool you use, do a clean once a year: duplicates out, orphaned files found or removed, genres checked against your list, and everything that has been in the inbox for more than a month either filed or deleted.

Playlists are views, not places

Once the tags are right, playlists stop being work. A smart playlist that says 122 to 126 BPM, house, energy 3 or above, bought in the last six months, regenerates itself every time you file new music, and a gig crate is a playlist you prune once from that list. Build a monthly “new” playlist as you file, keep a few standing crates for the kinds of gigs you play, and let smart rules do the rest. Playlists reference the library rather than duplicating it: a track that belongs in five crates lives in the folder once and appears in five lists.

The five-year test

A library is organised well when you can answer three questions in under a minute: where is the file, what do I know about it, and what happens if this drive dies tonight. Folder, tags, backup. Everything else is decoration, and decoration does not survive a change of software. Tags in the file and a tested copy somewhere else do.