Blog

Choosing Reconciliation Automation: What Controllers Should Test

Choosing Reconciliation Automation: What Controllers Should Test

Written by

Naman Mathur

Published on

Key Takeaways

  • Run five tests on your own last close before you sign, because a reconciliation automation demo on vendor data proves little.

  • Exceptions should route by value and account risk, with thresholds anchored to performance materiality and every auto-clear logged.

  • Watch the sign-off curve instead of the match rate: a 95% auto-match means little if the remainder still waits for working day five.

  • Most implementation pitfalls start before go-live: no exception policy, unstandardised templates, stale enterprise resource planning (ERP) exports and a spreadsheet tracker kept "just in case".

Most reconciliation automation shortlists are decided on match rate. That is the wrong number to compare. The close is lost on the remainder: the unmatched items and the exceptions nobody owns by working day three.

The harder question is which reconciliation automation software holds up on your data. Our answer is to test it on your own last close before you sign. For a vendor roundup, see account reconciliation software compared; this article is a test plan.

Six questions follow: what to test, exception routing, bank reconciliation time, status visibility, close speed and implementation pitfalls.

What Should a Controller Look for When Choosing a Reconciliation Automation Tool?

Look for evidence on your own data across five tests. Reconciliation automation is software that pulls transactions from your ledger, banks and subledgers, matches them, and routes the rest to a named reviewer. A good reconciliation automation tool does this continuously during the month and records every decision in an audit trail.

Our enterprise guide to reconciliation automation judges tools on five criteria: ERP integration depth, matching depth, exception workflow, controls and scalability. Below, we turn those criteria into tests you can run in a pilot, using last quarter's data and your own reviewers.

Test 1: Load a Late Journal and Time the Refresh

Post a journal to a reconciled account after the vendor's data load, then time how long the reconciliation takes to reflect it. With live ERP sync, the balance moves on the next refresh and the reconciliation reopens. With a batch export, you get a balanced reconciliation against a ledger that no longer exists.

Then widen the test, because your subledgers rarely sit inside the ERP. Cash arrives through Stripe or Adyen, spend runs through Coupa or Pleo, and payroll has its own provider. Ask the vendor to pull one of those feeds directly and tie it to the general ledger (GL) control account.

Test 2: Run Your Busiest Bank Account for Three Months

Give the vendor three months of statements for your highest-volume bank account and ask for a match rate with no new rules written. Then look at what is left: split payments, grouped receipts, foreign exchange (FX) differences and rounding. Ask why each item failed, because the answer tells you whether the engine learns or only applies rules.

In our implementations, rules-based matching plateaus at 70–75% and needs constant tuning, while learned matching reaches 95%+ out of the box. The gap sits in exactly those messy items, which is where your team's hours go.

Include every statement format your banks send. Most European groups receive a mix of camt.053, the ISO 20022 bank-to-customer statement, and MT940, the legacy Swift message it replaces.

Swift's Cross-Border Payments and Reporting Plus (CBPR+) plan is the reference here. The Swift CBPR+ roadmap statements show coexistence for statements planned to end with Standards Release 2028 (SR 2028), due November 2028, subject to revision. Ask your own banks when they will switch, and check the tool parses both formats.

Test 3: Follow One Exception From Detection to Sign-Off

Pick a real unmatched item from last month and follow it through the tool. It should get an owner, a probable cause, an ageing clock and a due date as soon as it appears. If it passes its due date, it should escalate without anyone sending an email.

Then check the resolution. Can the preparer draft the correcting journal from the exception, route it for approval and attach the evidence? If the answer involves a spreadsheet export and an email chase, the exception workflow has failed, whatever the dashboard shows.

Test 4: Try to Break Segregation of Duties

Ask a preparer to review their own reconciliation and see whether the tool stops them. Then change a balance on a signed-off account and check that the reconciliation reopens, with the original sign-off kept in the log.

This evidence now matters at board level in the UK. Under the Financial Reporting Council (FRC) UK Corporate Governance Code, Provision 29 applies for financial years beginning on or after 1 January 2026. Boards of companies under the Code must declare whether their material internal controls are effective, and each board defines which controls are material.

For many groups, balance sheet reconciliations will fall inside that scope. A tool that lets a preparer approve their own work hands your audit committee a gap to explain.

Test 5: Ask for the Basis of an Auto-Cleared Match

Pick five auto-cleared matches and ask the vendor to show why each one cleared, including the fields matched and the tolerance and model version applied. Then rerun the same data and check you get the same answer.

Repeatability is the point, and our AI reliability test of three general-purpose artificial intelligence (AI) agents on eight datasets shows why. The agents gave different answers across ten identical runs in 22 of 24 cases. Some of those false matches still balanced to the cent, so a reviewer checking totals could have signed them off.

An AI match counts as audit evidence only when it gives the same result twice and leaves a trail a reviewer can follow. Make that a pass-or-fail criterion in your scoring.

How Do Automated Reconciliation Tools Handle Exceptions Above a Materiality Threshold?

They route exceptions by value. Items above a threshold go to a senior reviewer with the evidence attached, and items below go to the preparer's queue. Differences inside a set tolerance can auto-clear, provided the tool logs the basis.

The auditor's materiality under the International Standards on Auditing (ISAs) is not a ready-made reconciliation threshold. ISA (UK) 320 performance materiality is set at less than materiality for the financial statements as a whole to reduce aggregation risk. Our view is that management's thresholds should borrow that logic, set by judgement per account rather than as a mechanical percentage.

Low-value items still need treatment. ISA (UK) 450 clearly trivial states that "'Clearly trivial' is not another expression for 'not material.'" Apply the same principle to auto-clear. Let small differences clear only when the tool tracks their aggregate across periods and flags items by nature, such as related-party balances.

Tier

Typical trigger

Route

Evidence kept

Tolerance auto-clear

Difference within tolerance on a low-risk account

Clears automatically

Match basis, tolerance applied, running aggregate

Preparer queue

Below the reviewer threshold

Account preparer

Probable cause, correcting journal draft

Reviewer

Above the reviewer threshold, or any high-risk account

Senior reviewer

Evidence pack, approval timestamp

Controller escalation

Above the escalation threshold, overdue, or flagged by nature

Financial controller

Escalation log, resolution, sign-off

Source: Stacks implementation practice, 2026.

In a demo, ask the vendor to set two thresholds by value and one by account risk, then push test items through each. If routing depends on someone reading a report, the tool does not route.

How Does Automated Matching Reduce the Time Spent on Bank Reconciliations?

Automated bank reconciliation removes the export-and-reformat step and the line-by-line tick-and-tie, so the remaining hours go on exceptions. Early adopters of our Bank Rec Agent cut manual matching by up to 90%.

The saving comes from learned patterns. In one Bank Rec Agent example, the agent learned that batch payments identified by ERP references typically settle one to four days after posting.

Step

Manual process

Automated matching

Load data

Download statement, export GL, reformat columns

Bank feed and ERP sync load daily

Match transactions

Tick-and-tie line by line in a spreadsheet

One-to-many and many-to-many matches clear automatically

Investigate breaks

Search emails and portals for remittance detail

Exception arrives with probable cause and source documents

Post adjustments

Key the journal into the ERP by hand

Correcting journal drafted for approval

Review

Reviewer re-checks the spreadsheet

Reviewer approves exceptions with the full trail

Source: Stacks implementation practice, 2026.

Measure it yourself before you believe any vendor figure, including ours. Time last month's bank reconciliation by step, then time the same steps in the pilot month.

How Can Finance Teams Get Real-Time Visibility Into Reconciliation Status Across the Close?

Use one status view fed by live ERP data, where each account's reconciliation status changes as matches clear and exceptions age. Visibility is only real when the data underneath is live. A dashboard built on Tuesday's export shows you Tuesday.

Look for four things per account: owner and status, value outstanding, ageing of open items, and whether it reopened after sign-off. A controller should see which entities are behind on day two without asking anyone. Volt's finance team, split between Poland and the UK, uses that view to monitor close progress in real time and spot bottlenecks.

The demo test is simple. Change a ledger balance on an account marked complete and watch the status view. If it stays green, the view is reporting on old data.

What's the Relationship Between Reconciliation Automation and Overall Close Speed?

Reconciliation automation shortens the month-end close when it moves reconciliation work into the month, so working day one starts with exceptions instead of a backlog. Matching alone does not do that. A tool that matches 95% of items on day three still leaves the close waiting on the remainder.

That shift is what drives the results: early adopters of our Bank Rec Agent closed the month 50% faster. Volt, a payments company running NetSuite, cut its monthly close by three days (33%) within three closes of going live, by January 2025. As Graham B., Senior Vice President (SVP) of Finance, puts it in Volt's close transformation: "We've already reduced the time to close by 3 days since implementation."

Your own numbers matter more than ours. Record these four before the pilot and again after two live closes:

  • Auto-match rate by account

  • Exception ageing at each working day

  • Share of reconciliations signed off by the target working day

  • Number of reconciliations reopened after sign-off

The third number is your sign-off curve, and it is the one your close calendar feels. For the wider close, including journals and review, see cutting the close without headcount.

What Are Common Pitfalls When Implementing Reconciliation Automation Software?

The common pitfalls are decisions skipped before go-live, and the mistake we see most often is starting with the messiest accounts. Begin with your cleanest, highest-volume accounts, so gains show in the first close and trust builds before the hard cases.

Two more come from skipping preparation. Teams skip template standardisation, so every account migrates with its own layout. Others go live without an agreed exception policy, which leaves thresholds and escalation rules to be invented in the middle of a close.

Stale data does the most quiet damage. A tool fed by stale exports reconciles to a ledger that has moved on, so start with planning your ERP integration before scoping anything else. Keeping the old spreadsheet tracker "just in case" splits the status between two places, and reviewers will trust neither.

After launch, the risk shifts to patience. Go-live is the start of tuning, and the first two or three cycles are where match rates and thresholds settle. Scope creep into multi-entity operations too early is the other trap, and running reconciliations across entities needs its own plan once one entity is stable.

Why Standard Advice on Reconciliation Automation Fails

The standard playbook says to buy the reconciliation tool with the highest match rate and the longest integration list. Both numbers are easy to compare in a demo, which is why they dominate shortlists.

The playbook fails twice. A 95% match rate on 20,000 lines still leaves 1,000 items, and the tool's value depends on what happens to them. And a higher auto-clear rate is not automatically better: a match the tool can't explain or repeat is a finding waiting for your auditor, not a saving. An integration list has the same blind spot, because it shows data can get in, not who owns an exception or how it ages.

The fair counterargument: a single entity with one bank and low volumes can reconcile well inside Xero, Sage or Exact. The case for a dedicated tool starts with multiple entities, several banks, payment provider subledgers or a board that must declare on its controls. Score vendors on the five tests, and give the exception workflow at least as much weight as match rate.

Why Choose Stacks for Reconciliation Automation

Stacks connects to your ERP and pulls live ledger data, so reconciliations run continuously through the month instead of from a month-end export. Stacks matching learns your patterns and improves each period. At Volt, Stacks matched over 95% of bank transactions in the first close and 97% by the third, replacing three to four days of manual matching in Excel.

Stacks reconciles bank accounts, payment service provider (PSP) settlements and intercompany balances on the same live ERP connection.

Unmatched items become owned, prioritised work items with a probable cause and a drafted correcting journal, evidence attached. Approvals and timestamps are captured as the work happens, so the audit trail is assembled before your auditor asks for it. Your team reviews the exceptions and provides the sign-off.

Andy M., Head of Finance at Cleo, describes the result. "With Stacks, we're saving a noticeable amount of time on the financial close, and as a result we can produce the management accounts faster each month." See how it works in reconciliation automation for enterprise.

FAQs for Choosing Reconciliation Automation

Can Reconciliation Automation Replace Manual Review?

No. It replaces the manual matching, so reviewers spend their time on exceptions and sign-off, where judgement is needed.

What Type of Control Is a Reconciliation?

Most control frameworks treat a reconciliation as a detective control, because it finds errors after they are posted. In our experience, groups rely on balance sheet reconciliations as key controls, so the sign-off evidence matters as much as the match.

What Auto-Match Rate Should We Expect?

Expect 95% or more on your main bank accounts within the first few closes. Volt matched over 95% in its first close on Stacks and 97% by its third.

Which Accounts Should We Automate First?

Start with your cleanest, highest-volume accounts, usually the main operating bank accounts. Leave complex accruals and intercompany accounts until the team trusts the tool.

How Long Should a Pilot Run?

Test on three months of historical data, then run at least two live closes before you judge the results.

See How Stacks Maps to Your Close Process

Book a 30-minute walkthrough. We'll discuss your close process, show you which tasks run automatically, and outline the next step from there.

GET STARTED

"Stacks has transformed how our finance team operates... it's saved us time and reduced frustration."

Graham B., SVP of Finance at Volt

Trusted by fast-growing companies including: Datavant, Future, Cleo, Motorway, Nivoda, Yoto and Epidemic Sound.

See how Stacks maps to your close process

Book a 30-minute walkthrough. We’ll discuss your close process, show you which tasks run automatically, and outline the next step from there.

"Stacks has transformed how our finance team operates... it's saved us time and reduced frustration."

Graham B.

SVP of Finance at Volt

Trusted by fast-growing companies including:

See how Stacks maps to your close process

Book a 30-minute walkthrough. We’ll discuss your close process, show you which tasks run automatically, and outline the next step from there.

"Stacks has transformed how our finance team operates... it's saved us time and reduced frustration."

Graham B.

SVP of Finance at Volt

Trusted by fast-growing companies including:

See how Stacks maps to your close process

Book a 30-minute walkthrough. We’ll discuss your close process, show you which tasks run automatically, and outline the next step from there.

"Stacks has transformed how our finance team operates... it's saved us time and reduced frustration."

Graham B.

SVP of Finance at Volt

Trusted by fast-growing companies including: