Setting up a document workflow

Nobody designs the folder full of files called final. It accumulates because three small habits were never agreed.

The folder full of files called final is not a software problem. It is three conventions nobody agreed, and agreeing them takes twenty minutes.

A folder of documents named date-first, sorting chronologically.
A folder of documents named date-first, sorting chronologically.

One: naming

2026-09-14 harlow-bridge quote.pdf
2026-09-16 meridian invoice-2026-184.pdf

Date first, year-month-day. Sorts chronologically everywhere and reads identically regardless of country, which the alternatives do not.

Then who it concerns. Then what it is. Lowercase, hyphens within parts, spaces between them.

The point is not elegance. It is that search finds filenames far more reliably than document contents, and scans contain no text at all unless recognition has been run.

Two: one current version

Every organisation eventually has final, final2, final-v3 and final-really.

The rule that prevents it: one current file at a fixed name with no version in it, and superseded versions moved to an archive folder with their date prepended.

harlow-bridge-quote.pdf
archive/2026-09-11 harlow-bridge-quote.pdf

Anyone looking for the current one finds it without deciding anything, and the history is kept and out of the way.

Version in the name Fixed name plus archive
Which is current Ambiguous Obvious
History kept Cluttered Yes, separately
Links to it Break each version Stable
Requires software No No

Three: publish what is read outside

This is the one that produces the most visible friction.

A shared storage link puts a permission screen in front of a client and downloads on their phone. It carries your folder structure with it. It was never designed for that job.

Anything somebody outside reads goes to an address. Storage stays what it is good at, which is you finding things again.

Keeping those two separate removes most of the exchanges about resending documents and requesting access.

One current document beside an archive folder of dated versions.
One current document beside an archive folder of dated versions.

Recognition on scans

A scanned document is a stack of photographs and nothing in it is searchable.

Most operating systems and storage products can run text recognition, either automatically or on request. Doing it at filing time makes the document findable by anything written in it.

One minute, and it is the difference between a contract being findable in two years and not.

Write it down

Conventions agreed verbally last about three weeks.

One page: the naming pattern with two examples, the current-version rule, and the line about publishing rather than sharing storage links. Put it wherever people look, which is usually the onboarding material.

That is the whole implementation.

Closely related: How to organise PDF storage, and PDF storage and hosting for the adjacent problem.

Put it at an address

Agree a naming convention, keep one current file with dated archives, publish anything read outside, run recognition on scans, and write the convention down.

Then the folder full of files called final stops accumulating.

Questions people ask

What causes most document friction?

Three things: inconsistent names, several versions competing, and using storage as a delivery mechanism for people outside the organisation.

What naming convention works?

Date first in year-month-day form, then who it concerns, then what it is. It sorts correctly, reads the same everywhere, and is findable by search.

How do I stop the version problem?

One current file at a fixed name with no version in it, and superseded ones dated in an archive folder. The rule is what matters, not the tool.

Where does publishing fit?

Anything somebody outside reads goes to an address. Storage is for finding things again, and using it to deliver produces permission screens and downloads.

Does this need software?

No. All three are conventions. Software can enforce them and none of them require it.

Keep reading