The compliance case for full traceability

Traceability gets presented as a feature: click a figure, see the transactions. Useful, unremarkable.

Its real value shows up on the day somebody asks a question nobody designed for. A tax authority querying a treatment. A bank asking how a covenant figure was derived. A buyer's due diligence team asking what a balance consists of. An auditor asking why a class of transaction was coded a particular way for four months in 2025.

None of those questions can be prepared for individually. What you can do is keep records in a state where any of them is answerable.

What traceability means precisely

Two directions, and both are required.

Downward. From any figure to the transactions composing it, from a transaction to the source document, from the document to the treatment applied and its basis. The chain from a number in the accounts to the piece of paper it came from.

Complete. Every item accounted for, whether or not it was processed automatically. Nothing absorbed into a balancing figure, nothing written off to make something agree, no residue nobody can decompose.

The second is the one that gets lost. A system can offer excellent drill-down and still permit unmatched items to be cleared to a suspense account — at which point the reconciliation always agrees and has stopped being a control.

We have written about the operational version of this in what full traceability actually means. The compliance argument is separate and worth stating on its own.

Why it matters more with automation

Under manual processing, understanding was distributed among people. If a question arose, someone remembered — imperfectly, but often enough.

Automated processing removes that. Nobody looked at the transaction, so nobody remembers it. The record is the only source of truth about what happened, and if the record is incomplete the knowledge does not exist anywhere.

That is a genuine change in risk profile, and it is the strongest argument for insisting on traceability before scaling automation rather than after.

The questions traceability answers

"Why was this treated this way?" From a tax authority or auditor, usually about a category rather than a single item. Answerable if the basis was recorded, not if only the outcome was.

"What is this balance made of?" From an auditor, a buyer or a lender. A balance that will not decompose is the most common trigger for expanded scrutiny, because it invites the question of what else is unexplained.

"How long has this been happening?" After an error is found. Answerable only if rule versions and configuration changes were recorded with dates. Without that, the scope of a correction is guesswork — and you have to disclose that it is guesswork.

"Show me the supporting document." The most common request, and the one automation handles best, provided documents are attached to transactions rather than filed separately.

"Who was responsible?" Answerable if ownership and review were recorded at the time.

Where traceability breaks

Four common failures, all of them design choices rather than accidents:

Automatic write-off of small differences. Convenient, and it destroys the completeness property. If items below a threshold are cleared without record, there is a category of activity that leaves no trail — and its size is unknown by construction.

Corrections made by deletion. Amending by removing and re-entering rewrites history. The correction should be visible, with the original preserved.

Documents stored separately from transactions. A reference to a filename in a shared folder is not a durable link. Folders get reorganised.

Configuration changes made without record. The most damaging, because it prevents you establishing when behaviour changed.

The test worth running

Take a figure from your most recent management accounts — a cost line, mid-sized, unremarkable.

Try to get from it to a source document without asking anyone for help. Time it.

Then ask: is there anything in that account that arrived via a write-off, an adjustment, or a balancing entry that would not lead anywhere?

Most businesses discover the drill-down works and the completeness does not. The path from figure to document is fine for items that matched; the items that did not match went somewhere else, and that somewhere is usually a place nobody examines.

The position worth holding

Traceability is not primarily about satisfying auditors. It is about retaining the ability to answer questions you have not been asked yet, in a system where nobody has personal knowledge of individual transactions because nobody looked at them.

That capability is cheap to maintain if it is designed in and effectively impossible to reconstruct afterwards, which makes it one of the few genuinely non-negotiable requirements when choosing how to automate.

Common questions

What is traceability in accounting?

It has two parts: the ability to move from any figure in the accounts down to the transactions composing it, then to the source document and the basis for its treatment; and completeness, meaning every item is accounted for whether or not it processed automatically, with nothing absorbed into a balancing figure or written off to make something agree. Systems commonly deliver the first while permitting practices that break the second.

Why does traceability matter more with automated accounting?

Because nobody looked at the transactions, so nobody remembers them. Under manual processing, understanding was distributed among the people who handled the work and could be recalled imperfectly when questions arose. With automation the record is the only source of truth, so anything not recorded is knowledge that exists nowhere.

What breaks traceability?

Automatic write-off of small differences, which creates a category of activity leaving no trail and whose size is unknown by construction; corrections made by deleting and re-entering rather than reversing visibly; documents stored separately from transactions with only a filename reference; and configuration changes made without record, which prevents establishing when processing behaviour changed.

How do I test my own traceability?

Take an unremarkable cost line from your most recent management accounts and try to reach a source document without asking anyone for help. Then ask whether anything in that account arrived through a write-off, adjustment or balancing entry that would not lead anywhere. Most businesses find the drill-down works while completeness does not, because unmatched items went somewhere nobody examines.

Related: what full traceability actually means · explainability in accounting automation · record retention when records are generated