Chart of accounts framework: four layers and six steps derived from reports

Most account lists are built forwards. A transaction turns up, nothing fits, an account is created, and the list grows one improvisation at a time. Built that way, a chart records the history of your confusion.

Build it backwards instead. Start from the reports you owe, work out what each one has to be able to say, and only then decide what has to be separated to make that possible. Everything you cannot trace back to an output does not belong on the list.

Figures below are invented to show the mechanism.

What to take away

The four layers

A chart goes wrong when a lower layer does a higher layer's job. Encoding sites into account names collapses the dimension layer into the account layer. Creating an account for every subtotal you might want collapses the group layer into the account layer.

Keeping the layers apart keeps the list short while the reporting stays rich.

Step one: list the outputs you owe

Write down every report that has to leave your books, and who reads it.

Typically that is a monthly result you read yourself, an annual set produced with an accountant, whatever a lender or funder has asked for in writing, and any operational report a manager relies on. Four or five outputs is normal. If you cannot name the reader of a report, it is not an output, it is a habit.

Step two: write the question each output answers

An output that answers no question is decoration. Force each one into a sentence beginning with a question word.

Notice that only one of those is about compliance. A chart designed solely around the annual set will answer that question well and the other three badly, which is the most common defect in a small business's list.

Step three: turn each question into a minimum split

For each question, name the smallest set of separations that could answer it. Minimum is the operative word: the temptation is to list everything that might be interesting.

"Which of the things we do is worth doing more of" needs revenue split by the things you do, and it needs the costs of delivering each of them split the same way. That is two splits, not twenty.

"Did anything move that I did not expect" needs your overheads grouped so that a movement is visible, which usually means four or five groups rather than forty lines.

Step four: decide account or dimension for each split

Apply one test to every split you have listed. Does it describe what was bought or earned, or does it describe who, where or which job?

What goes in an account. Who, where and which job go in a dimension. This single test removes the multiplication that makes charts unmanageable, because dimensions combine and accounts do not: three sites times twenty cost types is sixty accounts, or twenty accounts and one dimension with three values.

Where your system has no dimensions, you have a real constraint. Handle it by picking the one split that matters most and encoding only that, and by keeping the encoding in a suffix so the account still sorts with its own kind. Do not encode two dimensions at once; that is where the list stops being maintainable.

Step five: fix the spine before the detail

Four decisions carry most of the value. Make them deliberately, write them down, and then the rest of the list is comparatively unimportant.

The first of those is worth arithmetic, because its effect is larger than people expect.

Take a hypothetical month with revenue of 100. Wages total 40, of which 25 is the time of people working on delivering what was sold and 15 is everyone else. Materials are 20.

Profit is identical. The information is not. In the left column, a job that starts consuming more labor shows up nowhere until it has eaten the whole operating result. In the right column, the margin moves the month it happens.

This is what the profit and loss statement is for. It only works if the boundary was set in the chart.

Step six: write the definitions as you go

Every account gets one line saying what belongs in it, and where two accounts could both take a transaction, the definition says which wins. Write these at the moment you create the account, while the reason is still in your head.

This is the artifact that makes the structure reproducible when somebody else does the coding, and it is the first of the documents worth keeping beside the ledger.

Keeping it stable while the business moves

A structure derived from outputs changes when the outputs change, which is far less often than when transactions change. That is the point of building it this way.

When a genuinely new activity starts, ask if any existing output needs a new answer. Usually that means a new value in a dimension; sometimes a new revenue line with its matching direct cost line; only occasionally a new group.

Adding a lone account because a new supplier appeared is the pattern this whole sequence exists to prevent.

Balance sheet accounts follow a different rhythm. They change when the business acquires a new kind of obligation or asset, not when it sells something new. Each must reconcile to an outside record or it will quietly rot.

The balance sheet guide covers what those accounts claim. The type rules and numbering underneath all of this are in the chart of accounts overview.

Common questions