Best practices

How to design a filing structure that survives five years

Filing structures rot because they encode the organisation chart, project names and the habits of whoever built them. Here is what ages well, and how deep a structure should actually go.

Sascha Ladewig
Managing director & founder
2 November 2026
6 min read
Error loading image

Every filing structure starts out making sense to the person who built it. Most of them stop making sense to everybody else within about eighteen months. Then somebody creates a folder called "New structure", the old one stays where it is because nobody dares delete it, and five years later the drive contains three competing attempts and a folder called "Sort later" with four thousand files in it.

This is not a discipline problem. Structures rot for a reason, and the reason is predictable enough that you can design around it.

Why structures rot

A filing structure rots when it encodes things that change faster than the documents it holds.

Your organisation chart changes. Departments merge, a team is renamed, a function moves from one manager to another. If your top level is "Sales / Operations / Admin", every reorganisation leaves orphans, and nobody ever goes back to move the old documents.

People leave. A folder named after a colleague is clear while they are there and archaeology afterwards.

Projects end. A project name is a perfect label for two years and meaningless in year six, when nobody remembers whether "Phoenix" was the warehouse move or the website.

And the working year is not the retention year. A structure organised by calendar year looks tidy and then forces you to walk twelve folders deep to follow one client relationship across a decade.

The one test worth applying

Here is a test that takes ten seconds and predicts almost everything. Hand a document to somebody who joined last week, after one explanation of the structure, and ask them where it goes. Then put it where you would have put it.

If you disagree, the structure is wrong, no matter how logical it looks to you. A filing structure that only works for the person who designed it is a personal habit, and personal habits do not survive holidays, let alone five years.

What ages well

Organise around things that exist outside your organisation and have a life of their own. A client. A property. A vehicle. A machine with a serial number. A contract. A case. These things do not get reorganised when your management changes, they have obvious names that two people will arrive at independently, and they usually have a defined beginning and end, which makes retention tractable later.

Underneath that object, the next level that ages well is the type of document: contract, invoice, correspondence, certificate, photograph, report. Document types are remarkably stable. The categories a builder used in 1995 are the categories a builder uses now.

What does not deserve a folder level is anything you could answer with a question about the document. Status, year, responsible person, amount, whether it has been paid, which of your five locations it concerns. Those are properties, not places, and the moment you turn a property into a folder you have committed to one route through the information and closed off the others.

Depth is a symptom

A structure more than three or four levels deep is almost always a structure that has run out of other ways to express things. Each extra level is a question somebody has decided to answer with navigation rather than with a field.

Count the levels in your worst folder path. If the answer is seven, read the segments aloud. You will usually find that three of them are properties of the document: a year, a status, a region. In a case based system those become fields on the case or the document, and the navigation collapses to something a new colleague can hold in their head. That shift is most of what people mean when they say a DMS made their filing simpler, and it is the part of the feature set that actually changes daily work.

Retention belongs in the design

This is the step almost everybody skips, and it is the cheapest one to get right at the start.

In Germany, booking vouchers must be kept for eight years and commercial books, inventories and annual accounts for ten, under § 147 of the Abgabenordnung and § 257 of the Handelsgesetzbuch, with the clock starting at the end of the year of the last entry. Personnel documents, warranty paperwork and project records follow different logic again. Those are German obligations and they apply to organisations keeping books in Germany.

The practical consequence for your structure is simple: do not mix documents with wildly different lifespans in the same container when you can avoid it. If a case file contains the tax relevant invoices and also the marketing photographs from the site visit, then in year nine you have a container you can neither keep nor clear without going through it document by document. Separating them at the point of filing costs nothing. Separating them in year nine costs a week. We work through the periods and the mechanics on records retention.

A skeleton worth copying

For most small and medium organisations, this shape holds up:

The top level is the kind of relationship or object: clients, suppliers, staff, properties, vehicles, internal. The second level is the individual one, with a stable identifier in the name as well as the name itself, because names change and numbers do not. The third level is the document type. That is it. Everything else that you are tempted to add becomes a field: the year, the status, the responsible colleague, the location, the retention class.

Three levels, and the fourth dimension handled by metadata rather than navigation. It looks too simple when you first write it down, which is exactly why it is still standing in year five.

The honest part

Do not design the perfect structure. You cannot, because you do not yet know which of your assumptions are wrong, and the ones that are wrong will only reveal themselves when forty people use it.

Design something defensible for the next two years, write down why you made each decision, and plan to revise it once. One deliberate revision with the reasoning available beats five accidental ones.

And if you are a three person office with two thousand documents, do not spend a week on this. Spend an hour. Pick the object, pick the document types, agree it out loud, and get on with the work. A taxonomy project in a small organisation is a way of avoiding the actual job, and we would rather tell you that than bill you for it. When the structure genuinely is the hard part, because there are several departments with incompatible habits and nobody empowered to decide between them, that is the point at which an outside view earns its fee, and it is most of what our consulting actually consists of.

Topics

Filing structureMetadataRecords retentionDocument management

More posts