IPS, Saudi Arabia's instant payment system, moves customer transfers between banks in seconds, at any hour and on any day of the year. Most banks connected to it as a payments project. The harder work sits in operations, where many processes still assume the bank closes each evening and opens again in the morning.
A bank built around the business day
Core banking platforms, operating procedures and support rotas were mostly designed around a business day. The core runs its end-of-day batch overnight. Changes go in during a quiet weekend window. Reconciliation happens the next morning, and the payments team works office hours with an on-call number for emergencies.
SARIE fits that model. It has business days, settlement windows and cut-off times, and operations teams plan around them. IPS does not close. A customer can send money at 02:00 on a Friday, on the first day of Eid or in the middle of your core's end-of-day run, and expects the recipient to see it within seconds.
Each of those habits needs a new answer. The sections below cover the ones behind most incidents and complaints.
Keep posting while the core runs end of day
The first conflict is with the core banking batch. During end of day, many cores either stop accepting postings or accept them into the next business date. An instant payment cannot wait for the morning, so the bank needs a defined way to check funds, hold them and credit the payee while the batch runs.
Three patterns are common, and each has a cost:
- Online posting through end of day. The simplest answer, where the core supports it. Prove it under peak IPS load, during the heaviest part of the batch, before you rely on it.
- Memo posting, applied later. Payments update an available balance outside the core and post once the batch finishes. This works, but it needs strict reconciliation and a named owner for anything that fails to apply.
- A stand-in balance. A copy of balances outside the core authorizes payments within set limits. Service stays up, but the bank now holds a second view of each customer's money, and the two must agree.
Whichever pattern you use, write down which business date an instant payment belongs to when it arrives at 23:59, at 00:01 and during the batch. When the core, the payment hub and the general ledger disagree about value dates, the resulting breaks can take days to explain.
Design for the timeout first
Instant payments run on short time budgets. When a step does not answer in time, the bank does not know whether the payment went through. Most double debits and stuck holds come from the wrong reaction to that uncertainty.
- Give every payment request an idempotency key, so a retry from the channel or the hub returns the original result and never creates a second payment.
- Hold funds before sending, post once on a confirmed status, and release the hold on a rejection.
- When the status is unknown, ask before acting. ISO 20022 includes a payment status request (pacs.028) for this purpose, where the scheme supports it. Do not reverse or resend on a guess.
- Give returns (pacs.004) and cancellation requests (camt.056 and camt.029) their own defined paths and owners, because they also arrive at any hour.
This is the layer PayConnect covers between the payment hub and the core: verify and hold, post exactly once on confirmation, and reconcile every item across core, hub and rail. Payments initiated by licensed third parties need the same controls, as we set out in what open banking asks of your integration layer.
Reconcile during the day
Next-morning reconciliation was acceptable when payments moved in a few windows a day. With IPS, a break found at 08:00 may already be ten hours old, and the customer may already have called. Core postings, payment hub messages and scheme settlement reports need to be compared as payments flow, with a cut-off view at the points that matter to finance.
Two changes make the biggest difference. First, match continuously and report the exceptions: every unmatched item gets an owner and an age, and the oldest are reviewed first. Second, agree with finance how instant payments made over a weekend or a public holiday are booked, and run that case in testing before it happens in production.
Fund and watch liquidity out of hours
Instant payments keep moving when treasury desks are closed. The bank's position for IPS has to be funded and monitored through evenings, weekends and holidays, under the scheme's settlement rules. Salary days and the days before Eid bring peaks in outgoing volume that are easy to predict and costly to miss.
Treasury needs a view of the position it can trust at any hour, alerts well before a threshold is reached, and a named person with authority to act. If the plan for a Friday-night shortfall is to call someone at home, write it down, test it, and decide whether it is good enough.
Fraud screening inside the time budget
Instant payments are hard to recover once the payee has been credited. Authorized push payment fraud, where a customer is persuaded to send money to a fraudster, moves to whichever rail is fastest. Screening has to finish inside the payment's time budget, and it has to work on a Friday night when nobody is watching the review queue.
- Score outgoing payments in real time on payee, amount, device and behavior signals, with rules that can be changed quickly under change control.
- Decide in advance which payments are held for review, who reviews them out of hours, and what the customer is told while they wait.
- Watch incoming payments as well. Mule accounts receive funds and move them on within minutes.
- Give the contact center one view of each payment's status, so a customer who calls at midnight gets a real answer.
An operations floor that stays open
The largest change is in people and process. When a customer service runs around the clock, someone has to own it around the clock. Some banks staff a full night shift; others pair a smaller shift with on-call engineers. Either way, a few questions need clear answers:
- Who sees an IPS slowdown first, and how quickly? An alert in one of a dozen consoles is easy to miss at 03:00.
- Which runbook applies, and who can approve a replay, a release of held payments or a failover in the middle of the night?
- Is the incident described to executives in customers affected and money involved, or in component names?
- How do the next shift and the morning management meeting learn what happened overnight?
OpsCommand is designed for this work: one wall for the services customers use, end-of-day orchestration measured against the morning deadline, incident runbooks, maker-checker on operational actions, and an executive report ready at the start of the day.
Release changes while the service runs
Weekend maintenance windows used to be the safe place for change. For a service that never closes, every change happens in front of customers. Releases on the payment path need deployment patterns that keep the service running, a change calendar that respects the scheme's rules on planned unavailability, and a rollback that has been tested.
Regression testing has to cover the edges of the day: payments during end of day, across the date change, over a weekend and on a public holiday, plus late or missing responses from each dependency. A core banking migration now has to plan for instant payments arriving while the core is down; see migration rehearsals and go / no-go gates.
Measure what customers notice
Component uptime says little about whether customers could pay. More useful measures for an instant payment service are:
- end-to-end success rate and time to confirmation, by channel and by hour;
- timeouts and unknown-status payments, and how long each took to resolve;
- rejects and returns by reason;
- unmatched reconciliation items by age;
- customer-minutes of failed or delayed payments each month.
Report them every morning, and give the night the same weight as the day.
What to do next
- Map the business-day assumptions. List every process that assumes the bank closes: core end of day, batch interfaces, maintenance windows, reconciliation, treasury, fraud review and support. Mark the ones on the instant payment path.
- Settle the end-of-day pattern. Decide and document how the core handles instant payments during the batch, and how value dates are assigned across midnight, weekends and holidays.
- Walk the timeout path. Review timeout, retry and status-inquiry handling from channel to scheme, with idempotency at every step.
- Reconcile continuously. Compare core, hub and scheme settlement as payments flow, with an owner and an age for every exception.
- Name the out-of-hours owners. Liquidity, fraud review and incidents each need a person, a runbook and approval rights at any hour.
- Report customer-level measures. Agree them with the business and review them every morning.
If you want an outside view of how your operations hold up when IPS volumes peak outside business hours, book an executive briefing. We will look at your end-of-day, reconciliation and incident arrangements and give you a written view of the gaps that matter most.