Five manual reconciliation errors that grow with volume

Manual reconciliation fails in five predictable ways, and each one grows with volume. What they are, what they cost and how to catch them before the close.

The Actuals teamActuals · 4 min read

An invoice for €4,691 is keyed in as €4,619. At a few hundred transactions a month, someone would spot it. At tens of thousands a day, it is one line among many, and it waits for month end with all the others.

Manual reconciliation does not fail at random. It fails in five recognisable ways, and each of them grows with volume.

The five errors

  • Transposition. Two digits get swapped, like the €4,691 that became €4,619. The difference is 72, which is divisible by 9, as it always is when two digits trade places. That makes the error detectable, but only if someone is looking.
  • Omission. A PSP deducts its fee before the settlement reaches your bank, and the fee is never booked separately. Over time, those small gaps add up to material differences in margin and cash position.
  • Duplication. An automated bank feed imports a settlement, and a team member also enters it from the remittance advice. Where several people process the same payment batches, this is one of the fastest-growing error types.
  • Timing misclassification. A settlement that has not cleared yet is treated as an error and corrected. Now there is a real discrepancy where none existed. It happens most in close week, when every open item has to go.
  • Format and mapping. A PSP uses a 16-character alphanumeric reference, while your order system uses a numeric ID. If the two do not map cleanly, matching fails at scale and each unmatched item is investigated by hand.

Why volume multiplies them

At low volume, a skilled accountant catches a transposition, books the missing fee or spots the duplicate. At high volume, four things change.

More sources. Several order systems, PSPs, bank accounts and an ERP, each with its own structure and timing. In a row-by-row spreadsheet, every extra source adds chances for a miss.

Limited hours. When volume exceeds what people can check, teams start sampling. Sampling catches trends but misses individual exceptions, especially the small ones that add up.

Late detection. An error found at month end has lost its context. Source documents are harder to find, and the pressure to close encourages plugging the gap instead of tracing the cause.

No upstream control. If reconciliation only happens at the ERP, the wrong data is already in reports, accruals and intercompany balances. Fixing it at that point is reconstruction, not control.

What it costs

The hours are the visible part: time spent investigating mismatches that could have been flagged on the day they occurred.

The less visible part is money. Unreconciled settlements often hide fees, chargebacks or deductions that reduce net revenue. The longer they sit, the harder they are to recover.

Then there is risk. The ACFE’s Occupational Fraud 2026 Report found that more than half of the fraud cases it studied involved a lack of internal controls, or an override of existing ones. Reconciliation is one of those controls. Fraud detected within the first six months had a median loss of $40,000. Schemes that ran for more than five years caused median losses above $1.1 million. Discrepancies that stay labelled as "pending investigation" for period after period are a recognised warning sign.

Catching them before the close

  1. Normalise your sources. Transaction IDs, amounts, timestamps and currencies follow one standard across all systems. This is where most reconciliation efforts fail first, because each system exports data differently.
  2. Match each transaction, not the totals. A duplicate and an omission of similar value cancel each other out in a total, and both stay hidden.
  3. Classify exceptions straight away. A timing difference that will clear in the next settlement cycle, a known exception such as a cancelled order, or a genuine error. Only the last one needs investigation.
  4. Fix the cause, not the total. A "miscellaneous" adjustment balances this close and guarantees the same error next time. Correct the upstream entry and record what you did.
  5. Do it daily. Each session has fewer items, the context is still fresh, and corrections land before they spread into other reports.

Four controls that keep it honest

  • Segregation of duties. Whoever reconciles does not also initiate payments.
  • A second pair of eyes. An exception marked as explained needs approval from someone else before it is written off.
  • A fixed schedule. Daily for high-volume transaction flows, weekly at minimum, so backlogs cannot build up.
  • An audit trail for every correction. What the error was, why it happened, who approved the fix and when. Over time, the patterns in that trail show which problems to fix at the source.

Where automation fits

Automation does not replace your team’s judgment. It takes over the mechanical matching, so your people spend their time on the few items that need a person.

Actuals sits between your operational systems and your ERP. It normalises the data from your PSPs, banks and platforms and matches every transaction individually. The exceptions go to your team to resolve. The ERP receives only reconciled, auditable journals.

"We use Actuals to track all transactions from initiation to our general ledger, ensuring they match with supplier invoices and remain balanced. A paradigm shift to be in control," says Patrick Roozeman, CEO of MultiTankcard.

Where to start

Look at when your team finds most of its errors. If the answer is month end, start with step 5. An error that is one day old still has its context, and nothing has been built on top of it yet.

The Actuals scan takes 15 minutes and shows where your data first stops matching.

Sources

  • ACFE, Occupational Fraud 2026 Report
In practiceWhat breaks first when your volume doubles?

Written by

The Actuals teamActuals