Your ERP was built for less volume than you process
Your ERP records what it is given, correctly. At high volume, what it is given arrives in pieces. Why the check has to happen before the ledger, not at close.

It is the second week of the close, and someone is explaining a difference the ERP never flagged. Again. Yet the ERP did nothing wrong. It recorded every entry it received, in order and without error. The trouble started before those entries reached it.
Most finance leaders we speak with know a version of this. The system is stable, the team is capable, and the close still takes weeks. The people are not the problem, and neither is the software. It comes from an assumption almost nobody says out loud.
The assumption inside every ERP
SAP, Oracle, NetSuite, Exact and Dynamics were all built around the same idea. Transaction data arrives at the accounting engine already correct. An order exists, an invoice matches it, a payment is applied once. The ERP only has to record it faithfully.
At 5,000 transactions a month, that holds. A controller can review the flow by hand, errors surface early and balances reconcile in a day. At 150,000 transactions a month, the picture changes. Two or three payment providers, a marketplace feed, an order system and manual corrections now feed the same ledger. Nobody can watch that flow anymore. Checks turn into sampling, and samples miss the difference that matters.
What actually arrives
Data reaches the ERP in pieces.
- A payment shows up in the PSP export on one day and in the settlement file three days later.
- An order ships in two parts, so one sale becomes two lines without a shared reference.
- A refund is recorded before the charge it reverses, because the two come from separate systems on different schedules.
- A duplicate arrives once via the API and again in the batch file.
None of this is unusual. It is how any business behaves when it handles real money across several partners. The ERP records all of it, exactly as it was built to. What it cannot see is the payment that never arrived, or the settlement split across three files. Nor does it see the refund that belongs to a charge from last quarter. Those gaps only show up when someone goes looking, usually at close.
The close pays for it
When the ERP accepts fragments, the matching moves downstream. PSP report to bank, settlements to orders, each order to the ledger. It happens in spreadsheets and hand-kept lookups, and in the memory of the person who has done it for years. Teams add buffer days, hire a reconciliation specialist and accept a close that takes weeks.
The cost rarely appears as a line item. Hunting differences takes days every month, days that do not go to analysis or forecasting. A difference found late also costs more to fix. By then the books are closed, the payment has cleared and the customer has been told. Auditors notice it too. If reconciliation after the fact is your main control, they test harder and want more documentation, even if balances are right.
Why more people do not help
Every new payment provider, marketplace or currency needs a new mapping, a new script and someone who understands the rules. The first one is manageable. By the fifth, the close starts to slip.
Volume grows quickly, while headcount rises in steps. So the team gets bigger, the workload per person stays high, and the next provider starts the cycle again. A bigger team often produces a longer close, not a faster one. The problem sits upstream, where headcount does not reach.
Check before the ERP, not after
The fix sits between the sources and the ERP. Before a transaction is recorded, it is checked against where it came from.
- Does the payment exist at the provider?
- Does the order system agree?
- Is the customer attached?
- Is the settlement status what the provider reports?
- Has this transaction come in before?
Only when every answer is yes does the entry move on. That is not reconciliation. Reconciliation compares two sides after the fact. This check happens before the ERP, so the ERP only receives data that has already been established as true.
Then bank, provider, order system and ledger all agree with each other. Differences still occur, but they arrive as exceptions with a reason. The close confirms what the system already knows, instead of rebuilding what happened three weeks earlier.
Trace one balance
Pick one month-end balance and trace it back to source without asking anyone. Note where the trail stops. That point is the real limit of your architecture, and it usually sits upstream of the ERP, not inside it.
Want a second pair of eyes? The Actuals scan takes 15 minutes and shows where your data first stops matching before it reaches the ledger.
In practiceWhat breaks first when your volume doubles?


