There is a particular kind of meeting that only happens in companies with four systems: the one where two people have different numbers and the first forty minutes are spent establishing which export each of them ran, and when.
An export is a snapshot with the timestamp removed
The problem is not that exports are inaccurate. It is that they are accurate as of a moment that is not written down anywhere the reader can see. Two people pulling the same report four hours apart get two defensible numbers, and there is no artefact that explains the difference — so the difference gets investigated as if it were an error.
Multiply that by cards, payables, banking and the ledger and you have four snapshots that have to be shown to agree before anyone can say what the company spent.
What “one ledger” buys
Not elegance. It buys the ability to answer a question without a reconciliation first.
- A number has one source, so disagreement is about interpretation rather than provenance.
- Evidence hangs off the transaction rather than off a folder that mirrors it.
- The close stops with judgement calls, because the mechanical agreement already holds.
The migration everybody fears
It is usually smaller than expected, and the reason is that most of the four systems are not sources of truth at all — they are places where a copy was landed for someone’s convenience. The genuine sources are typically two: the bank and the ledger. Everything else is a view that stopped being called a view.
Run one month in parallel, reconcile both books once, side by side. After that there is one book and no export to prove it — see the first close is checked twice for what that month actually involves.
When four systems is the right answer
Genuinely different regulatory entities with genuinely different ledgers are not a failure of architecture. The test is whether the consolidation is a mapping (fine, mechanical, automatable) or a negotiation (not fine, and no tool fixes it).
