Instant Payments Turn Treasury Into a Night Shift, and Fraud Into a Timing Problem.
Twenty-four-hour irrevocability changes the authoritative cash position from an end-of-day report to a sequence of executable balances, cut-offs and prefunding decisions. The operational discipline is to connect payment permission to the liquidity that can be used at that exact moment.
The rule
For an irrevocable payment, cash that arrives tomorrow is not liquidity today.
One small phrase in treasury reporting has done a great deal of harm while looking perfectly innocent.
Cash available.
Late in the afternoon it means the balance in your general ledger.
At 22:15 on a Saturday it might mean the money in a bank account.
At the moment an instant-payment rail asks whether your institution may put value on the system, it means something much narrower and much more useful.
Settlement value.
In the required currency.
At the required account.
Free of prior claims.
Before the relevant cut-off.
With somebody authorised to use it.
Not synonyms.
Different facts wearing the same name badge.
Real-time payments have not made your treasury more strategic.
They have made it more nocturnal.
The old arrangement allowed a civilised separation between the customer’s instruction, the bank’s batch, the funding movement, the reconciliation run, and eventually somebody in Treasury deciding what the day had meant.
An irrevocable payment accepted around the clock has rather fewer manners.
It asks whether the liquidity exists now.
If your answer is in the group, broadly, the rail will respond with the sort of silence usually associated with a poor credit committee.
So the important operational change is not speed.
It is the replacement of one end-of-day cash position with a sequence of executable positions.
Each is local.
Currency-specific.
Reserved or unreserved.
Bounded by a clock.
Your fraud problem joins at the same point.
A control that reaches a brilliant conclusion after finality has written an excellent incident report and nothing else.
The Dashboard Balance Is Not the Payment Capacity
Start with the three balances every dashboard would rather merge.
Ledger cash is the accounting record of what your organisation believes it has.
The bank-reported available balance is the bank’s view, after value dates, overdrafts, restrictions and transactions in transit.
Usable settlement balance is the amount actually available to meet this rail’s obligation, in this currency, through this settlement account, at this moment.
The first is authoritative for accounting.
The second comes from your bank.
The third belongs to your settlement bank, your central bank, or the rail’s prefunding arrangement.
Your forecast is authoritative about none of them.
It is a controlled opinion about what each should become.
Treating that forecast as cash is not optimism. It is a category error wearing a tie.
The ordinary example contains the whole problem.
An account holding $1.4 million at 08:00, against a same-day $1.2 million outgoing wire, has at most $200,000 available for a new decision before other files and charges.
Now add an instant rail.
A reserved outbound instruction.
A conversion window that closed before the customer asked to pay. An internal transfer requiring an approver who is currently at dinner.
Your group may still be rich.
Your settlement account is not.
Which is why you should stop asking what the cash position is.
The question is too pleased with itself.
Ask instead:
What settlement value can this legal entity release, on this rail, in this currency, at this time, after every prior reservation and before every funding cut-off?
TIPS makes the separation unusually tidy.
It settles in central-bank money, supports dedicated accounts, and provides final, irrevocable settlement around the clock.
Its platform has no transaction upper limit.
Participant limits and available settlement liquidity become the actual boundary.
FedNow publishes a US$10 million network limit for customer credit transfers and payment returns, and a participant may set a lower internal limit.
Neither statement tells you whether your institution can send the next payment. That answer lives in your available settlement value, not in the brochure.
Build the Record That Can Be Released
Do not store Cash = available beside Payment = approved.
That is not a control record.
It is an invitation to a meeting where everybody arrives with a different extract.
For every payment-capacity decision, keep the nine-field intraday-liquidity record.
Payment rail. Currency. Available balance. Reserved balance. Credit line. Collateral availability. Funding cut-off. Settlement account. Escalation owner.
The rail identifies the finality rule, the message path and the operating clock.
Currency stops an impressive dollar balance being volunteered against a euro obligation with no conversion route.
Available and reserved balances separate what exists from what is already earmarked.
Credit line and collateral availability separate a headline facility from funding you can actually draw or monetise.
The funding cut-off says when the option stops being an option.
The settlement account says where the money has to be.
The escalation owner is the person who can make an exception, without anyone pretending a shared mailbox has signed the mandate.
Not an onerous addition to your cash forecast.
It is the part that makes the forecast executable.
Treasury already knows how to build the longer view.
Daily buckets for the next 15 business days, weekly buckets through week 13, monthly buckets through month 24.
A working design, not a universal commandment.
The near view needs a smaller unit, because a missed flow there prevents a payment.
A locked period might forbid automated changes to same-day forecasts after 11:00 local, except under an approved exception.
Not because eleven o’clock has acquired theological significance.
Because a forecast needs a point at which people stop editing it and start funding it.
Each flow needs value date, amount, currency, source system, owning business unit, and confidence class: contracted, approved but not contracted, or forecast.
A sales forecast in week ten is not the same funding object as a contracted bond coupon in week ten.
Nor does typing 63.7% beside it improve its moral character.
The forecast supplies a probability.
The payment rail requires a decision.
The Rail Has Its Own Opinion About Completion
The payment is a sequence, and it should stay one in your record.
A structured instruction carries debtor account, creditor account or alias, amount, currency, end-to-end identifier and authentication record.
That unique identifier does not decorate a confirmation screen.
It distinguishes a retry from a second instruction when somebody sees an uncertain response and, as people do, presses the button again.
Before release, your sending provider checks account status, customer funds, authorisation, scheme limit, fraud and compliance controls, and settlement capacity.
A passing customer balance test is not a settlement-capacity test.
You can reserve or debit the payer and send the message. The receiving side still validates it and chooses its response.
RTP documents three receiving responses: Accepted, Accepted without Posting, and Rejected.
The first two cause settlement.
The third releases your reservation and settles nothing.
Dull distinctions, until somebody labels all three completed in a data warehouse and then has to explain whether a customer was debited, a receiver was funded, or a payee balance was usable.
At which point the taxonomy becomes quite popular.
Final settlement discharges the inter-provider obligation under the system’s rules.
Customer posting is the receiving provider’s action that makes the customer’s balance usable.
Close together.
Not the same event.
TIPS validates, conditionally reserves, obtains the receiver’s acceptance, then settles and confirms on a positive reply.
Its published processing figure is under 100 milliseconds.
Excellent engineering.
Not a reason to let your accounting model become vague.
The EU rules are equally exact about the customer-facing outcome.
Regulation (EU) 2024/886 requires the receiving provider to make funds available and confirm completion within 10 seconds of receiving the order.
If confirmation does not arrive in that window, the payer’s provider restores the account as though the payment had never happened.
So the rail hands you three operationally distinct outcomes.
A rejection.
A settlement followed by posting.
A timeout with restoration.
A competent ledger preserves all three.
A cheerful app can call them whatever it likes afterwards.
Fraud Is a Release Decision With a Clock Attached
In a batch environment a suspicious payment sits in a queue while a human finds the file, rings a number and forms views about the beneficiary.
On an instant rail, fraud control sits before finality and has to finish inside the execution window.
That makes fraud neither less serious nor more mystical.
It makes it a decision under a clock.
Your inputs are familiar.
Device and credential signals.
Beneficiary information.
Transaction velocity.
Account history.
Negative lists.
Sanctions controls.
Customer-specific limits.
FedNow supports negative lists and account-activity thresholds based on count or cumulative value in a chosen period, and automatically populates three cumulative-value criteria for send-and-receive participants.
Useful pre-authorisation controls.
They no more relieve you of a fraud policy than a smoke alarm relieves a building of stairs.
The awkward case is degradation.
FedNow procedures say that where a fraud-tool check exceeds its time threshold, the message may continue processing and a warning is recorded.
That is the decision a committee would rather defer, which is exactly why you write it before Saturday night.
Do you hold the payment and lose the speed promise?
Do you release and accept the recovery problem?
Who can override the control?
What record proves why?
There is no third option in which the payment stays instant and everybody gets another afternoon to think.
The same discipline applies afterwards.
RTP guidance says a credit transfer is irrevocable once sent to the system, and final when the creditor institution accepts it.
A return request is a distinct credit transfer or scheme procedure.
It is not a time machine.
RTP’s compliance bulletin requires unauthorised RTP-native payments to be reported through a designated system message no later than two business days after your fraud determination.
Reporting creates actionable information. It does not unsend money.
So the audit trail keeps the original instruction, the authentication result, the control decisions, the timestamps, the settlement response, the customer notice and the later return request.
Funding Is Real Only If It Arrives Before the Payment
This is where your treasury earns its evening.
As a settlement account approaches its low-watermark, you may draw a committed revolver, borrow overnight, sell an eligible investment, repo securities, use an overdraft, make an internal transfer, or run a cash-pool sweep.
Each is a different source.
Different currency.
Different counterparty.
Different notice period.
Different collateral condition.
Different account route.
Different cut-off.
Calling all seven liquidity saves space on a slide and removes the one distinction that matters.
An undrawn £40 million line carrying £12 million of letters of credit and a £5 million overdraft sublimit may yield £28 million, or £23 million, or less, depending on the facility agreement.
An unused facility with expired signatory evidence, or unmet information undertakings, is unavailable until cured.
A group holding a sterling deficit at a payroll bank and a surplus at a pool header can move the money only where the accounts, payment window, documentation and authority are already in place.
The cash-pool screen is not a settlement instruction, though it frequently has the self-confidence of one.
Test the contingent route outside normal hours.
Bank contact.
Facility condition.
Drawdown mechanics.
Collateral release.
Internal transfer.
Approval sequence.
Bank acknowledgement.
Final settlement account.
Sensible treasury procedure keeps a call tree and one emergency payment route, tested as a controlled exercise.
The question is not whether your revolver exists.
It is whether you can turn it into the necessary currency, in the necessary account, before the rail has to decide.
That timetable has to be stricter than your corporate forecasting calendar.
An investment maturing three days after a payment of the same size is not a match, unless settlement timing and fallback liquidity have been checked.
A treasury waiting for end-of-day statements finds its shortfall after the cut-off for the only practical funding route.
The liquidity report remains correct in a certain philosophical sense.
It has simply stopped being useful.
Continuous processing does not abolish business dates either.
FedNow continues through its funds-transfer business-date rollover at approximately 7:01 p.m.
Eastern Time.
The route you need to fund an outbound payment may keep a different clock again.
FX conversion, collateral substitution, bank transfer, internal pool sweep and facility drawdown all take a keen interest in calendars.
A twenty-four-hour rail is not a twenty-four-hour guarantee that every supporting market is open, reachable and charming.
The Ladder, Because Someone Will Say the Group Has Plenty of Cash
Payment status?
Ask the rail operator.
Usable settlement balance?
Ask the settlement bank or the central bank.
Internal cash forecast?
Ask Treasury.
Credit availability?
Ask the facility agent.
Collateral value?
Ask the custodian and the haircut schedule.
Fraud recall?
Read the rail rules.
Emergency release?
Demand a named treasury signatory.
For an irrevocable payment, cash that arrives tomorrow is not liquidity today.
What the paper establishes
- A customer ledger balance, a bank-reported available balance, and usable settlement liquidity are different facts with different authorities.
- Treasury must forecast payment liquidity by bank account, currency, intraday time bucket, cut-off, and confidence rather than as one consolidated cash total.
- Every instant-payment release needs a record of the payment rail, currency, available balance, reserved balance, credit line, collateral availability, funding cut-off, settlement account, and escalation owner.
- Payment finality, customer posting, timeout restoration, return requests, and fraud reporting are separate events governed by rail rules and provider controls.
- Fraud controls must decide before finality or follow a documented degradation rule; a later report or return request does not reverse the original payment.
- Contingent funding counts only after available drawings, collateral, documentation, signatory authority, payment route, and cut-off have been tested outside ordinary business hours.
- A manual hold or emergency release requires a named treasury signatory and retained evidence of the authority exercised.
Questions
- Why is a positive cash balance not enough to send an instant payment?
- A positive ledger or consolidated cash balance does not prove that the sending institution has usable settlement liquidity in the required currency and account. The payment rail may require prefunded value, a dedicated central-bank-money balance, or a settlement-bank position before acceptance. Existing reservations, payment cut-offs, restrictions, collateral terms, and available drawings can reduce the amount that can support an irrevocable payment.
- What does treasury need to know before an instant payment is released?
- Treasury needs the payment rail, currency, available balance, reserved balance, credit line, collateral availability, funding cut-off, settlement account, and escalation owner. The record links a customer instruction to the liquidity that can be used at the relevant moment. It prevents a general cash forecast, a customer account balance, or an undrawn facility headline from being treated as proof of payment capacity.
- When does an RTP payment become impossible to cancel?
- RTP guidance states that a credit transfer is irrevocable once it is sent to the system and final when the creditor financial institution accepts it. A later request for return of funds is a separate scheme process and does not cancel the original transfer. The provider therefore completes authentication, fraud, sanctions, beneficiary, balance, and settlement-capacity controls before release rather than after settlement.
- Is a completed instant payment the same thing as the payee seeing the money?
- No. Final settlement discharges the inter-provider obligation under the rail’s rules, while customer posting makes the payee’s account balance usable under the receiving provider’s terms. RTP can record Accepted without Posting as well as Accepted and Rejected. The separate statuses preserve the distinction between the receiver’s settlement decision, its customer-ledger action, and the sender’s final confirmation.
- What happens when an instant payment times out?
- The outcome depends on the rail rule and the point reached in processing. Under Regulation (EU) 2024/886, the payee’s provider must make funds available and confirm completion within 10 seconds of receiving the order from the payer’s provider. If the payer’s provider receives no confirmation in that period, it restores the payer’s account as if the transaction had not occurred.
- Why does fraud become a timing problem in instant payments?
- An instant-payment fraud control must decide before finality, inside the rail’s execution window. Device and credential signals, beneficiary information, transaction velocity, account history, negative lists, sanctions controls, and customer-specific limits help decide whether release is safe enough. A later investigation, fraud report, or return request can support recovery and remediation, but it does not provide a leisurely cancellation period for the original payment.
- Can a fraud control be too slow to protect an instant payment?
- Yes. A fraud control that does not return a decision within its allowed time creates a documented choice between stopping service and allowing processing to continue. FedNow procedures state that a message may continue processing when a fraud-tool check exceeds its time threshold, with a warning recorded. The provider needs a risk appetite and escalation rule for that degradation before an incident occurs.
- How should treasury forecast cash for a twenty-four-hour payment rail?
- Treasury forecasts the near-term position by bank account, currency, and the smallest time bucket in which a missed flow can prevent payment. Each expected flow carries a value date, amount, currency, source system, owning business unit, and confidence class. A consolidated group balance is a second view because one entity’s surplus may be unavailable to another entity’s settlement account.
- Does an undrawn facility always count as instant-payment liquidity?
- No. A facility counts only to the extent that current drawings, letters of credit, guarantees, covenant restrictions, collateral conditions, notice requirements, documentation, signatory authority, and the payment route permit a draw in time. An unused facility with expired signatory evidence or unmet information undertakings is unavailable until cured. Treasury tests drawdown mechanics and the emergency payment route rather than relying on the commitment headline.
- Why do payment cut-offs still matter when the rail operates all day?
- A payment rail can operate continuously while the funding, conversion, collateral, internal-transfer, or settlement-bank route needed to support a payment has its own cut-off. FedNow processing continues through a funds-transfer business-date rollover at approximately 7:01 p.m. Eastern Time. Treasury therefore records the rail clock and the separate cut-offs for every route that turns broader cash into usable settlement value.
- Who can stop or release an emergency payment hold?
- A payment hold or emergency release requires a named treasury signatory with authority under the relevant mandate, account controls, and payment arrangement. The retained evidence identifies the person, authentication, scope of authority, instruction details, time, and any required approval. A manual intervention without a named authority turns a fraud or liquidity control into an untraceable decision.