innovaq

Back to Finance

AI · 4 min read

Bank reconciliation: method, rules and automation

Matching every transaction to its document is the only way to know what has really been paid. A practical method for doing it every day and leaving only the exceptions to your team.

A necessary, unloved task

Bank reconciliation means matching every transaction on the account with the document that supports it: a paid invoice, a supplier payment, a salary, a bank fee. It is the only way to know for certain which invoices have actually been paid and to make sure the books reflect what is really in the account.

It is also one of the most repetitive jobs in finance. When transactions are numerous and payment descriptions unclear, manual matching takes time and care, and tends to be put off. The result is misaligned balances, unrecorded receipts and reminders sent to customers who have already paid.

The most common causes of mismatches

Most of the difficulty comes down to a few recurring cases. The customer who pays several invoices with a single transfer, or one invoice in several installments. The transfer that arrives net of fees or of a credit note not yet recorded. The payment description that quotes an order number instead of the invoice number, or nothing at all.

Then there are transactions with no supporting document, such as bank charges, tax debits or transfers between accounts of the same company, which still have to be recorded and classified. Knowing your own recurring cases is the first step toward handling them methodically instead of treating each one as an exception.

The quality of the underlying data matters too. Up-to-date customer master data, invoices that clearly state the reference to quote in the payment description, and credit notes recorded promptly eliminate many doubtful cases at the source. Part of the reconciliation work is done before the transaction even reaches the bank.

Illustrative example: the breakdown varies from one company to another.

A method that starts with the easy cases

A good reconciliation process starts with importing transactions regularly, ideally every day, straight from the bank feeds. Working through a few days of transactions is far easier than tackling a whole month at once.

Next comes automatic matching of the obvious cases, where amount, counterparty and reference line up unambiguously. Then the doubtful cases are reviewed, with the operator choosing between several candidate documents or splitting a transaction across multiple invoices. Last come the exceptions: transactions with no matching document that need a specific entry or a check with the customer or supplier.

The value of this method lies in focusing attention where it is needed. If the obvious cases close themselves, the team’s time goes to doubtful cases and exceptions, which is where human judgment really makes a difference.

  1. 01 Daily import
  2. 02 Automatic matching
  3. 03 Review of doubtful cases
  4. 04 Handling exceptions

Rules that get better over time

Many transactions recur with the same characteristics: the same supplier debiting every month, the same quarterly fee, the same customer always paying with a particular reference. Defining rules for these cases, by bank and by counterparty, steadily reduces manual work.

Rules do, however, need regular review. A customer changing how they pay or a newly opened account can make well-functioning rules obsolete, and a wrong rule is more insidious than a missing match, because it produces errors nobody looks at.

That is why each rule should have an owner and a last-reviewed date, and why the rules applied should always be visible next to the match. Whoever checks the work must be able to see at a glance why a transaction was linked to a given document.

A good rule has

  • A defined bank and counterparty
  • An owner
  • A last-reviewed date
  • Its reason shown next to the match

The role of AI

In Finance, AI suggests matches between transactions and documents based on amount, date, counterparty and payment description, even when the reference is partial or the payment covers several invoices. Every suggestion shows the reason for the match and a confidence level, so the operator knows how far to trust it.

User confirmations and corrections improve future suggestions, and recurring cases can be turned into explicit rules with a single click. Reconciled receipts immediately update the receivables schedule and report, so you never chase customers who have already paid. The goal is not to remove human oversight, but to ensure the team only works on the exceptions.

The goal is not to remove human control, but to let the team work on the exceptions only.

Keep reading

Innovaq is a company in its launch phase.