Reconciliation Automation in Bank Operations Done Right

Automate Reconciliation in Bank Operations

18% of accountants make errors every day. Reconciliation automation with clear matching rules and owned exceptions keeps bank books in line.

Alexander Kiel

Business Analyst · Banking & Wealth Management

11.12.2025 · 8 MIN READ · PROCESS OPTIMISATION

ON THIS PAGE

1. List Every Reconciliation You Run

2. Clean the Data Before Matching

3. Define Automated Reconciliation Rules

4. Route Each Break to an Owner

5. Retire Spreadsheets With Care

6. Track Match Rates and Root Causes

Sample case: BNY Mellon

In short

FAQ

Reconciliation is where errors in bank operations come to light. Much of it is still manual. A Gartner survey of controllership staff found that 18% of accountants make financial errors at least daily. Reconciliation automation aims to catch those errors early and leave people only the breaks that need judgement.

The data is changing as well. On 22 November 2025 the coexistence of MT and ISO 20022 ended for cross-border payments. Swift observed 97% of payment instructions sent using ISO 20022. The Basel Committee expects regular verification and reconciliation of transactions and accounts as part of sound operational risk management.

Yet many banks still match statements against ledgers in spreadsheets and clear breaks by email. Does your bank know today how many open breaks are older than a week?

Illustrative example of 10,000 daily items in a nostro reconciliation: 9,200 match exactly, 500 match by split or tolerance rules and 300 become exceptions routed to owners by cause

STEP 1

List Every Reconciliation You Run

Start with an inventory. List every reconciliation in your operations: nostro accounts, securities positions, custody holdings, general ledger against sub-ledgers and derivative portfolios. For each one, record the two sources, the frequency, the daily volume, the number of open breaks, the hours spent and the owner. The list shows where automation pays first.

Some frequencies are set by regulation. Under EMIR, financial counterparties must reconcile portfolios every working day when 500 or more OTC derivative contracts are outstanding and once a week for 51 to 499 contracts, according to the Dutch AFM. My article on as-is process mapping shows how to record each step and handoff before you change it.

✓

List each reconciliation with its two sources and frequency

✓

Record volumes, open breaks and hours spent per month

✓

Mark reconciliations required by regulation or by audit

How many reconciliations run in your operations each day and who owns the full list?

Keep in mind: Shadow reconciliations kept in personal spreadsheets rarely appear in any inventory. Ask each team what they check by hand.

STEP 2

Clean the Data Before Matching

A matching engine is only as good as its inputs. Before you write rules, align the formats on both sides: dates, currencies, signs, account identifiers, ISINs and counterparty codes. Agree which field carries the reference that links a payment to its booking. Make sure every source system fills it.

Structured payment data helps. Swift calls the rich, structured data of ISO 20022 foundational to the future of payments. Remittance fields that were free text in MT messages now arrive in defined elements. The Basel Committee asks that risk data should be reconciled with bank's sources, including accounting data where appropriate. My article on the data quality framework shows how to name a golden source for each field.

✓

Align date, currency, sign and identifier formats on both sides

✓

Agree one reference field that links each payment to its booking

✓

Fix missing references in the source system, not in the tool

Which field links a payment in your bank to its booking in the general ledger?

Keep in mind: Mapping tables built inside a reconciliation tool become a second source of truth. Keep reference data in one owned system.

STEP 3

Define Automated Reconciliation Rules

Write the matching logic in layers. Start with exact matches on reference, amount and value date. Add one-to-many and many-to-one rules for netted or split payments, then rules with tolerances for small amount or date differences. Document each rule, its tolerance and the person who approved it. Test it against a month of real data before it goes live.

Automation reduces errors but brings its own risk. The Basel Committee notes that automated processes are less prone to error than manual processes, while warning that they introduce risks that must be addressed through sound technology governance. A tolerance set too wide closes breaks that should have been investigated.

✓

Start with exact matches on reference, amount and value date

✓

Add split, netted and tolerance rules one layer at a time

✓

Test every new rule against a month of real data

Who in your bank approves a new matching rule or a wider tolerance before it goes live?

Keep in mind: Every tolerance is a decision to accept a difference. Review tolerances with Finance and Risk at least once a year.

STEP 4

Route Each Break to an Owner

Unmatched items need a home. Classify each break by cause, such as missing booking, wrong amount, timing difference or duplicate. Route it to the team that can fix it. Give every break an age, set escalation points and report open breaks by age and value to management each week.

Regulation sets clear deadlines for some breaks. Under the EMIR risk mitigation rules, counterparties must have specific procedures in place for disputes that are not resolved within five business days. Banks must notify the Dutch central bank of disputes representing a minimum value of EUR 15 million, which have not been resolved within fifteen business days.

✓

Classify each break by cause when it is created

✓

Route breaks to the team that can correct the source

✓

Report open breaks by age and value every week

How many breaks in your bank are older than five business days and who chases them?

Keep in mind: A break closed with a manual adjustment hides the cause. Require a reason code and a reference for every write-off.

STEP 5

Retire Spreadsheets With Care

Spreadsheets often hold the knowledge of how a reconciliation really works: special cases, manual lookups and adjustments nobody wrote down. Before you switch one off, document what it does, run the old and new process in parallel for a full month-end and compare the results line by line.

Supervisors treat spreadsheets as a risk. BCBS 239 expects effective mitigants where a bank relies on manual processes and desktop applications (eg spreadsheets, databases). Robots on top of a broken process do not fix it: EY reported that as many as 30 to 50% of initial RPA projects fail, partly from targeting RPA at the wrong processes. My article on how to automate reports in Excel shows how to make a spreadsheet step traceable while it remains.

✓

Document every manual step and lookup in the spreadsheet

✓

Run old and new process in parallel through month-end

✓

Compare results line by line before switching off

Which reconciliation in your team depends on a spreadsheet only one person fully understands?

Keep in mind: Parallel runs feel like double work. They are the only proof that the new process catches what the old one caught.

STEP 6

Track Match Rates and Root Causes

Measure the result. Track the automatic match rate per reconciliation, the number of new breaks per day, the average age of open breaks and the share of breaks by cause. When one cause keeps returning, fix it at the source system or in the upstream process instead of clearing the same break every morning. In a 2025 reconciliation project I worked on, most breaks traced back to reference data in a source system that had been coded too generally, with no exceptions built in.

Errors are frequent without such feedback. In the Gartner survey, 59% of accountants make several errors per month. The same research found that companies with high technology acceptance see a 75% reduction in financial errors, which depends on staff trusting and using the tools rather than working around them.

✓

Report the automatic match rate for each reconciliation monthly

✓

Rank break causes and fix the top three at source

✓

Review rules when the match rate drops after a change

What share of items in your largest reconciliation matches automatically without any manual touch?

Keep in mind: A high match rate can hide wide tolerances. Read the match rate together with the volume of tolerance matches.

Will Automation Take Away All Our Jobs? | David Autor (Professor of Economics @ MIT)

SAMPLE CASE

BNY Mellon

Two UK entities of the Bank of New York Mellon held client safe custody assets that peaked at about £1.3 trillion at its London branch. They served 6,089 UK-based clients. The firms ran global platforms that did not record which group entity each client had contracted with, so the books could not show who owned which asset.

Without entity-specific records, the firms could not carry out entity-specific external reconciliations of client assets. Safe custody assets were commingled with firm assets across 13 proprietary accounts. The failings ran from 1 November 2007 to 12 August 2013 and were not identified by the firms' own compliance monitoring.

In April 2015 the Financial Conduct Authority fined the two firms £126 million. Without the settlement discount, the penalty would have been £180 million.

Key takeaway: Reconciliation can only be automated as far as the records allow. When accounts do not show which entity or client owns an asset, no matching engine can produce a reliable result.

"Had the Firms become insolvent, the total value of safe custody assets at risk would have been significant."

Georgina Philippou (Acting Director of Enforcement and Market Oversight @ Financial Conduct Authority)

In Short

Automating reconciliation in bank operations rests on six habits: a full inventory of reconciliations, clean and aligned data, layered matching rules with approved tolerances, an owner for every break, a careful exit from spreadsheets and steady tracking of match rates and root causes.

The result is fewer manual hours, breaks that are found and closed within days, recurring errors fixed at the source and a control set that auditors and supervisors can follow.

Frequently Asked Questions

What is reconciliation automation?

Reconciliation automation is the use of software to compare two sets of records, such as a bank statement and a ledger. It matches them by rules instead of by hand. Matched items close automatically. Unmatched items become exceptions that are routed to a person for investigation. The aim is to reduce manual work and find real errors sooner.

What is a nostro reconciliation?

A nostro reconciliation compares the bank's own record of an account it holds with a correspondent bank against the statement that correspondent sends. Each payment and receipt should appear on both sides. Differences point to missing bookings, wrong amounts, fees or timing gaps. Banks usually run nostro reconciliations daily, because open items can tie up liquidity and hide fraud.

Which reconciliations should a bank automate first?

Start with reconciliations that combine high volume, frequent breaks and many manual hours, such as those for nostro accounts, cash and securities positions. Reconciliations required by regulation or audit also rank high. Low-volume reconciliations that need complex judgement often add little value when automated. A short inventory with volumes, break counts and hours spent usually makes the order clear.