Ask ten writers what information a song should carry and you will get ten different answers, most of them shaped by whatever went wrong for them last. Someone who has lost a publishing split will tell you credits. Someone who has been asked for a radio edit at short notice will tell you versions. Both are right, and neither list is complete.
What follows is the set that consistently earns its place, in rough order of how much trouble its absence causes. The test throughout is simple: does not having this cost you real time or real money later. If it does not, it is decoration, and decoration is what makes people abandon systems.
The identity: what this song is
A stable title, spelled the same way every time, and the date it was started. That sounds too obvious to list until you have three folders called Golden Hour, Golden hour and GoldenHour_NEW and no idea which came first. Pick a spelling and never vary it. Working titles are fine, and changing a working title later is fine, as long as the old one is recorded somewhere rather than simply overwritten.
The start date matters more than people expect, because it is the one piece of information that can never be reconstructed later. File modification dates lie: they change when you touch a file, when you copy a drive, when you restore a backup. If you care when a song was written, and you may one day care a great deal, write it down at the time.
The people: who did what
This is the field whose absence costs actual money, and it is the one most often skipped, because on the day of the session everybody in the room knows who did what and it feels absurd to write down. Two years later, one person has changed their name, another has a different manager, and nobody can remember whether the second verse was a co-write.
- Writers, with their real legal names, not just the name you call them
- Producers, and whether that was full production or additional
- Performers on the recording, particularly session players
- Anyone who contributed a topline, a hook or a lyric idea, even in passing
- Agreed splits if they were discussed, and the date they were agreed
The critical detail is that these should be records of people, not strings of text. A name typed into a notes field can tell you who played on one song. A real record of a person can tell you every song they played on, which is the question you will actually want answered when they call you three years later.
The sound: tempo, key and what it feels like
Tempo and key are worth having on every song, and worth almost nothing typed in by hand, because the audio already contains both and reading them from the file is more reliable than remembering. The reason to have them is not documentation, it is retrieval: when a brief asks for something around 90 beats per minute in a minor key, you want that to be a filter rather than an afternoon.
Genre and mood are softer and more argued about, and still worth capturing, because they are how people actually describe what they want. Nobody asks for a song in E flat. They ask for something warm, or something that sounds like late summer. Whatever vocabulary you use, use it consistently: an inconsistent tag is worse than no tag, because it makes the filter lie.
The state: where this song is up to
Which version is current, and what stage the song is at. This is the one that decays fastest, because it changes every time you touch the song, and it is the one most likely to be wrong in any system that relies on you updating it by hand.
Keep it coarse. Idea, demo, produced, mixed, mastered, released is plenty for most people. Fine-grained status fields feel more accurate and end up less accurate, because nobody maintains twelve states honestly.
The history: where this song has been
Who you have sent it to, when, and what came back. This is almost never captured, and it is the field that most reliably embarrasses people: pitching the same song to the same person twice, or failing to follow up on genuine interest because it was six months ago and it was in an email.
It also compounds in a way the others do not. One record of one pitch is admin. Three years of them is a map of who responds to what, which is worth considerably more than the sum of its rows.
If you are starting from nothing, capture credits first. Everything else on this list can be reconstructed from the audio or from memory. Who agreed to what, on what day, cannot.
What does not earn its place
Plugin chains, sample rates, which mic you used on the vocal, the exact signal path. All genuinely interesting, all things people put in filenames, none of them things you will ever search your catalogue for. If it matters for one song, it belongs in a note attached to that song. It does not belong in a field you now have to fill in on every song forever.
The same goes for anything you cannot keep accurate. A field that is right half the time is worse than an empty one, because an empty field prompts you to check and a wrong one does not.
Where this information should live
Almost none of it should live in the filename. A filename can carry identity and version and that is genuinely about it. Everything else is too long, goes stale, and cannot be searched across a catalogue in any useful way.
The other common home is a spreadsheet or a database built alongside the audio, which works well on the day it is built and drifts from then on, because the music keeps changing and the spreadsheet only changes when somebody remembers. That drift is not a discipline failure, it is what two separate copies of the truth always do.
The version of this that holds up is one where the information lives with the audio rather than beside it, and where as much of it as possible is read from the files instead of typed. Tempo, key and version relationships can all be derived. Credits and history cannot, which is exactly why they are worth the effort of recording properly.
That is the principle Locker is built on, and it is worth saying plainly that the principle matters more than the tool. A catalogue that fills itself in stays complete. One that depends on a habit ends up half empty, no matter whose software it is kept in.