It is 9:40 in the treasury back office. The USD statement from the correspondent bank landed overnight, the EUR one is still missing, and the mirror ledger shows a balance that is off by an amount nobody recognises. Someone opens last month's workbook, copies the tab and starts ticking lines by hand. The forex dealers want a position by eleven.
A nostro reconciliation process is not one comparison but four, and the work is explaining the gap, not finding it. We think that framing matters more than the tool, though nostro reconciliation software for banks makes it a daily habit rather than a month-end scramble. Below is the checklist we would run, a worked example, and the honest case for staying in a spreadsheet.
The process is four comparisons, not one
A nostro account is your bank's foreign-currency account held with a correspondent bank abroad. You keep a mirror ledger of it in your core or GL. The correspondent sends its own statement, usually as a SWIFT MT950. The two should agree, and on most days they do not.
Every unmatched line falls into one of four buckets. Naming them is the first step in any nostro reconciliation process:
| Open item | What it usually means |
|---|---|
| Debited in our ledger, not on the statement | A receipt we booked that has not reached the correspondent's statement yet |
| Credited in our ledger, not on the statement | An outward payment we booked that the correspondent has not processed yet |
| Debited on the statement, not in our ledger | Charges, a claim or a payment we have not booked |
| Credited on the statement, not in our ledger | An inward remittance nobody has applied yet |
Teams that skip this step end up with one long list of "unmatched" lines. That list is hard to own, because each bucket belongs to a different person: payments, remittances, or finance.
Most nostro breaks have five ordinary causes
Nostro breaks feel mysterious at 9:40. By lunch they are usually one of a short list:
- Value date differences. We booked today; the correspondent books tomorrow. The item clears itself in a day or two.
- Correspondent charges. Fees deducted on the statement that nobody booked in the ledger.
- Unapplied inward remittances. Money arrived, but the customer or purpose is unclear, so it sits unbooked.
- Missing or late statements. No MT950 means no comparison, so the whole account looks open.
- Reference and amount mismatches. A truncated narration, a duplicate posting or a keying error.
The first two are timing and housekeeping. The last three need a person. A good process separates them early, so analysts spend the morning on remittances and errors, not on items that will clear on their own.
The gap is always the sum of its open items
An illustrative example, with round numbers. Say a bank's USD nostro mirror ledger closes at 1,200,000 and the correspondent's MT950 closes at 1,249,940. The gap is 49,940. Four open items sit behind it:
| Open item | Amount (USD) | Effect on statement vs ledger |
|---|---|---|
| Outward payment booked by us, not yet on the statement | 15,000 | +15,000 |
| Inward remittance on the statement, not yet applied by us | 40,000 | +40,000 |
| Correspondent charge on the statement, not booked by us | 60 | −60 |
| Receipt booked by us, not yet received | 5,000 | −5,000 |
The arithmetic: 1,200,000 + 15,000 + 40,000 − 60 − 5,000 = 1,249,940. The ledger, adjusted for open items, equals the statement. That is a reconciled account, even with four lines still open.
Notice what the example does not say. It does not say the account is clean. The 40,000 inward remittance is the item to chase today, because a customer is waiting for money that has already arrived. The 60 charge needs booking. The 15,000 should clear tomorrow. The 5,000 needs a question to the counterparty if it does not arrive soon.
A nostro account can balance to the cent and still hide a customer's money in an open item nobody owns.
Ageing tells you more than the match rate
A high match rate is pleasant to report. It is also the least useful number on a nostro recon, because the risky items are the few that stay open. What we would track instead, each day, for each account:
- Open items by bucket, with an owner against each one.
- Age of each item in days since it first appeared.
- Which items crossed your internal escalation threshold today.
- Statements expected but not received.
Picture the morning after this is in place. The dealers get their position before ten. The back office opens a short list of named items, each with an age and an owner. The 60 charge was booked yesterday. The inward remittance is with the operations desk, who have the customer on the phone. Nobody is copying last month's tab.
How long an item may stay open, and when it must be escalated or written back, is set by your bank's policy and the applicable RBI guidance. Confirm the thresholds with your compliance team rather than borrowing someone else's. Our guide to classifying and ageing reconciliation breaks covers the workflow in more detail.
The strongest objection: low volumes do not need software
A smaller bank or NBFC might hold four nostro accounts with a few hundred entries a day. The treasury team knows every correspondent and most of the counterparties. A careful analyst with a spreadsheet can close that in an hour. Why change?
That is a fair point, and for some teams the answer is: do not change the tool yet. Change the process. Name the four buckets, age every item, record an owner, and write the matching rules down. If the spreadsheet does all of that and the analyst can take leave without the recon stopping, it is doing its job.
The case for software starts when one of three things is true. The rules live in one person's head. Open items age without anyone noticing. Or the checker signs the close without being able to see how a match was made. Those are control problems, not volume problems, and they grow quietly.
Tolerance rules need a ceiling and a checker
Tolerance rules are where nostro recons drift. A small amount tolerance absorbs correspondent charges. A date window absorbs value date differences. Both are sensible. The trouble is that a tolerance widened once to clear an awkward month rarely narrows again.
Three habits keep tolerances honest. Set a written ceiling per account and currency. Require a second person to approve any change. And report what each tolerance matched, so the checker can see whether it is absorbing charges or hiding errors. Our guide to exact, tolerance and fuzzy matching rules explains the rule types.
This is how we built AutoRecon's nostro recon: the mirror ledger matched daily against the MT950 on rules you write, including tolerance rules for amount and timing differences, with every open item classified, aged and given an owner until it clears. Rules are versioned and go through maker-checker, and recurring breaks come back as suggested rules. More material is in the AutoRecon guides.
What to do this week
- List every nostro account, its currency, correspondent and when its MT950 normally lands.
- Sort today's open items into the four buckets and put an owner against each one.
- Add an age column and flag anything past your internal escalation threshold.
- Write down every tolerance in use, with its ceiling and who approved it.
- Book the charges and chase the unapplied inward remittances first.
None of this needs a project to start. It needs one morning, a clear list and the habit of explaining every gap before the dealers ask.