Every lab has attempted this. Someone circulates a convention, it holds for about six weeks, and then the folder fills up with final_2.xlsx, Copy of data (1).csv and asdf.csv again.
It lapses for a reason that has nothing to do with discipline. Most schemes are designed to look tidy in a file browser rather than to survive being taken out of the folder that explains them — and files are always, eventually, taken out of that folder.
A lab manager or PI trying to make a convention stick, or a student setting one up before the data arrives. It assumes nothing about software and works on a shared drive, an external disk, or an instrument PC.
The one idea: the name has to travel
A file is rarely read where it was written. It gets attached to an email, dropped into a slide, copied to a collaborator's disk, pulled into an analysis folder, or recovered from a backup with its folder structure gone.
Every one of those moves strips away the context that the path was carrying. So the test for any scheme is simple:
That is why longer names beat deeper folders. A path like 2024/March/Priya/plate reader/final/ holds four useful facts, and loses all four the moment the file moves. Put those facts in the filename and they move with it.
The four rules
1. Date first, and always YYYY-MM-DD
Leading with the date makes chronological order and alphabetical order the same thing, in every tool, on every operating system, for ever. No sorting, no scripts.
The format matters more than people expect. 03-04-2024 is March in one country and April in another, and you will not remember which you meant. 2024-03-04 is unambiguous everywhere and sorts correctly as plain text.
2. Then what it is, in words a stranger would use
Not a project code only you hold the key to. plate-read beats PR; aggregation-assay beats exp4. Abbreviations feel efficient while you are the only reader, and cost real time later.
3. Then the thing that makes it unique
Usually the sample, condition, or plate — whatever distinguishes this file from the one beside it. This is the part that stops you needing _v2, because two files that differ genuinely will differ in their names.
4. Never spaces, never punctuation, never version words
Hyphens between words, underscores between fields. Spaces break command-line tools and get rewritten as %20 in links. Slashes, colons and asterisks are illegal on some systems and will fail silently on transfer.
And never final, final2, FINAL-real. Version words are a promise nobody keeps. If you genuinely need versions, use a date or v01, v02 — zero-padded, so ten sorts after nine.
YYYY-MM-DD_what-it-is_which-one_initials.ext
2024-03-14_plate-read_aggregation-10uM_rk.csv
2024-03-14_confocal_hela-ctrl-01_rk.czi
2024-04-02_qpcr_batch7-plate2_am.csv
Long? Yes — about fifty characters. That is the point. Each one survives being emailed, and sorts itself.
Why longer names beat deeper folders
Deep folder trees feel organised and fail in four specific ways.
- They encode one order only. A tree filed by person cannot also be browsed by date or instrument. A flat folder of well-named files can be sorted or searched any way you like.
- The context evaporates on move. Everything the path was telling you is gone the moment the file is attached to an email.
- They hide duplicates. The same file in three branches looks like three different files.
- They break on transfer. Long paths hit character limits, and instrument software often refuses to write into nested directories at all.
Two or three folder levels are plenty — project, then year, perhaps then instrument. Let the filename carry the rest.
A good name is a manual solution to a real problem: keeping a file connected to what it means. woodle.cloud keeps that connection as a property of the record instead, so a badly named file is still traceable to its run. The convention above is worth adopting either way — it costs nothing and it works on any disk.
Where your files live →Making it stick
A convention circulated by email will lapse. These four things are what separate schemes that survive from schemes that do not.
The cases that break every scheme
Two runs on the same day
Add a time or a run number: 2024-03-14-1430 or _run2. Decide which now, in writing, because this comes up in week one.
The instrument names files itself
Some software will not be argued with. Do not fight it and do not rename the original — the vendor file often refers to its own name internally, and renaming can break the ability to reopen it. Keep it as written and record the link somewhere.
Data that arrives from a collaborator
Never rename what you were sent. Keep the original, add a note of where it came from and when, and rename only copies you derive.
A file that belongs to two projects
Store it once, under the project that produced it. Copies fall out of sync, and by year three you will not know which was edited.
The version-number honesty check
If a folder contains analysis-final.xlsx and analysis-final-v2.xlsx, the naming is not the real problem. It means two files exist that nobody has decided between — and a naming convention cannot decide for you.
Names make a decision visible. They do not make it.
The one-line shape, the four rules and the checklist are free to copy into your own lab documentation. No attribution needed and no signup. If you change one thing after reading this, make it the date format.