Everyone in a large organisation depends on the general ledger, and almost no one outside finance technology knows what it actually is. It is not a spreadsheet. In a large institution it is a layered system, and each layer exists to solve a specific problem. Here is the shape of it, without the jargon.
1. The chart of accounts — the vocabulary
Before anything can be recorded, the system needs an agreed language for what is being recorded: accounts, cost centres, legal entities, currencies. This is master and reference data, and it is the foundation. Get it wrong and every number downstream inherits the error. Most of the quiet, unglamorous work of a GL platform is keeping this vocabulary clean and consistent.
2. Ingestion — bringing the transactions in
Real economic activity happens in source systems — lending, trading, payments, payroll. The ledger does not originate these; it receives them. An ingestion layer pulls transactions from many sources, each with its own format and timing, and translates them into the ledger’s shared vocabulary.
3. Accumulation — turning events into balances
A raw feed of transactions is not yet accounting. The accumulation layer aggregates those individual events into balances: by account, by entity, by period. This is where a stream of activity becomes a position you can report on.
4. The ledger and the trial balance — the source of truth
At the centre sits the GL database itself, capable of producing a trial balance — the check that, across every account, debits equal credits. When the trial balance ties out, the books are internally consistent. When it does not, something upstream is wrong, and the system has to help you find it.
5. Validation and controls — catching errors automatically
You cannot rely on people to eyeball millions of entries. Automated validation controls run continuously: balancing checks, tolerance thresholds, reconciliation against source. The aim is to surface a problem before it reaches a financial statement, not after.
6. The compliance framework — proving it, not just doing it
In a regulated institution it is not enough to be correct; you must be able to demonstrate that you were correct. A compliance layer captures the audit trail, the approvals, and the evidence that controls ran and passed.
7. The reporting warehouse — a separate stage for the audience
Finally, reporting does not run against the live ledger. A separate downstream data warehouse holds the numbers in a shape optimised for analysis and disclosure. Keeping reporting off the transactional system protects the source of truth and lets each side be tuned for its own job.
Why the separation is the whole point
The recurring theme is separation of concerns. Master data is kept apart from transactions; transactions apart from balances; the source of truth apart from the reporting layer. Each boundary is a place where errors can be caught and where one part can change without breaking the others.
That is also why these systems are so hard to migrate. You are not moving one thing — you are moving seven interlocking layers, each with its own assumptions, and the trial balance has to keep tying out the entire way across.