The Friday Bank Detail Change: A 45-Day Review of Supplier Payment Controls
Photo: Foreign and Commonwealth Office, OGL v1.0, via Wikimedia Commons
A supplier emails on a Friday afternoon to say its bank details have changed. The message arrives from a familiar address, quotes a recent invoice number, and asks for the next payment to be redirected. The finance team follows the policy: the change is logged, a second approver reviews it, and the payment run proceeds. Every check is ticked. The money still leaves for the wrong account.
This is not a story about absent controls. It is a story about controls that were designed when the business was smaller, the payment run was slower, and the finance team knew every supplier by voice. UK Finance has consistently reported that authorised push payment fraud exploits authorised, expected behaviour rather than technical breaches. The control environment looks adequate on paper because it was adequate once.
Why the control passes the check but fails the pressure
Most mid-market finance functions have a supplier change policy. Fewer have tested whether that policy survives the last week of the quarter, when payment volumes spike, approvers are travelling, and the person who usually makes the callback is on leave. The gap is rarely a missing signature. It is a control that assumes time, attention, and unchanging scale.
A 45-day review is long enough to test the control under realistic conditions and short enough to hold management attention. It should be scoped to supplier master data, payment instruction changes, and the authority limits attached to both. It is not an IT security audit. It is a finance operations review with a board-ready output.
Days 1 to 10: map who can change bank details
Start with the supplier master file, not the policy. Identify every route by which a bank detail can be created or amended: the finance system, the purchase-to-pay platform, the banking portal, and the shared mailbox where supplier emails arrive. For each route, record who can initiate, who can approve, and what evidence is retained.
Pay particular attention to inherited access. A former purchase ledger clerk may still sit in an approval group. A temporary contractor may have been given standing rights to amend supplier records during a system migration. The question is not whether the policy says dual authorisation. The question is whether the system enforces it for every route, including the manual workaround used when the workflow is down.
Produce a one-page map showing the routes, the roles, and the evidence trail. If the map cannot be produced from system records, that is the first finding.
Days 11 to 20: test the callback under realistic pressure
A callback is only a control if it is made to a known number, not a number supplied in the change request. Test this with a controlled exercise. Take a sample of recent supplier changes and ask the finance team to demonstrate how the callback number was verified. In many firms, the number came from the same email chain that requested the change. That is not verification. It is repetition.
The test should include a simulated request arriving late on a Friday, with a payment run due the following Monday and the usual approver unavailable. Observe what happens. Does the team escalate, or does it rely on the relationship? Does it use the number on the original supplier contract, or the number in the signature block? The exercise is not designed to catch people out. It is designed to reveal where the control depends on individual judgement rather than a defined process.
Record the results by route and by payment value. A callback that holds for a 2,000 pound payment but collapses for a 200,000 pound payment is not a control. It is a habit.
Days 21 to 30: reconcile dual authority with actual payment sizes
Dual authority is often set by role, not by value. Two approvers may be required for any payment above a threshold that was set several years ago, when the business was half its current size. Inflation and growth have quietly moved the effective risk. A limit that once captured most material payments may now sit below the routine run rate.
Extract the last twelve months of payment values and compare them with the authority matrix. Identify the proportion of total value that passes through single-authority routes. If a small number of large payments routinely clear below the dual-authority threshold, the control is decorative. Recalibrate limits to the current payment distribution, and make the recalibration an annual board item rather than a system setting that nobody revisits.
Days 31 to 38: close the gap between policy and quarter-end reality
Finance policies are usually written for a calm Tuesday. Quarter-end is not a calm Tuesday. During the final week of a quarter, payment volumes rise, approvers are under pressure, and exceptions are granted verbally. The review should sample transactions from the last week of each of the previous four quarters and test whether the same controls were applied.
Look for payments released under manual override, supplier changes approved without a documented callback, and approvers acting outside their delegated limit. The finding is not necessarily misconduct. It is evidence that the control design assumes a pace the business no longer operates at. The remedy may be a separate quarter-end payment protocol with named deputies, a hard cut-off for supplier changes, and a requirement that any override is logged and reviewed the following week.
Days 39 to 45: the one-page control register
The output of the review should be a single page the board can scrutinise without a finance glossary. It should list each control, the risk it addresses, the evidence that proves it operated, the owner, and the last test date. A sample structure is below.
| Control | Risk addressed | Evidence | Owner | Last tested |
|---|---|---|---|---|
| Supplier bank detail change requires dual approval | Impersonation and unauthorised amendment | System audit trail and approval log | Financial Controller | Month 1 |
| Callback made to independently verified number | Callback to fraudster-controlled number | Callback record with source of number | Purchase Ledger Manager | Month 1 |
| Dual authority threshold aligned to payment distribution | Large payments clearing single authority | Authority matrix and payment value analysis | Finance Director | Quarter 1 |
| Quarter-end override log reviewed weekly | Verbal exceptions and manual releases | Override log with reviewer sign-off | CFO | Quarter 1 |
The register is deliberately short. A board does not need twenty controls. It needs the five that would have stopped the Friday afternoon email, and evidence that they operated when the pressure was real.
The blunt cost comparison
A single successful supplier impersonation fraud in a UK mid-market firm can easily exceed the annual cost of prevention. The loss is not only the payment. It includes recovery costs, legal advice, temporary liquidity pressure, and the management time spent rebuilding supplier confidence. The annual cost of a disciplined control review, a callback log, and a recalibrated authority matrix is typically a fraction of that figure, even when external support is used.
The comparison is not between fraud and no fraud. It is between a controlled loss and an uncontrolled one. The invoice that passes every check is the one that tells you the checks were designed for a different business. The 45-day review is how you find out before the Friday email arrives.