For most of banking history, liquidity was managed at end of day. Payments moved in batches, systems closed at night and at weekends, and a treasurer who knew the closing position could plan the next morning's funding. That model is ending. Instant payments run around the clock, customers can move balances from a phone in seconds, and 2023 showed that a loss of confidence can empty a bank faster than any end-of-day process can respond.
The speed of a modern run
The Federal Reserve's review of the March 2023 failure of a US regional bank describes the scale plainly. On 9 March the bank lost more than $40 billion of deposits, and management expected to lose more than $100 billion on 10 March. Together that was roughly 85% of its deposit base. By comparison, the review estimates that one large US failure in 2008 involved about $10 billion of outflows over eight days, and another $19 billion over 16 days1. The withdrawals were highly correlated among a concentrated network of depositors and were fuelled by social media1.
The 2023 run was an order of magnitude faster
Deposit outflows in selected US bank failures, as estimated by the Federal Reserve
| Failure | Outflow | Period |
|---|---|---|
| US regional bank, March 2023 (actual) | Over $40 billion | 1 day |
| US regional bank, March 2023 (expected next day) | Over $100 billion | 1 further day |
| Large US failure, 2008 | About $19 billion | 16 days |
| Large US failure, 2008 | About $10 billion | 8 days |
Note: 2023 actual and expected outflows together represented roughly 85% of the bank's deposit base.
The review also points to operational gaps. The bank had not tested its ability to borrow at the discount window in 2022, had limited collateral pledged there and could not move securities collateral quickly from its custodian or its Federal Home Loan Bank to the discount window1. Management had also changed stress-test assumptions, including for intraday liquidity, in ways that reduced the modelled shortfall1. None of this was a data architecture failure alone. But a treasurer without a live, trusted view of positions and collateral cannot act at the speed a digital run requires.
Instant payments make every hour a business hour
Regulation is removing the old buffers. Under the EU Instant Payments Regulation, euro-area payment service providers have had to receive instant euro credit transfers since 9 January 2025 and to send them, with verification of payee, since 9 October 2025. Providers in non-euro member states follow in January and July 20273. Instant transfers must be available 24 hours a day, every day, and cost no more than a standard transfer3. Funds must reach the payee within ten seconds4. Before the mandate, SCT Inst's share of SEPA credit transfers ranged from about 5% to more than 50% across euro-area countries4. The mandate aims to make instant the default.
In the US, adoption is voluntary but accelerating. The FedNow Service settled 8.4 million payments worth $853 billion in 2025. In the second quarter of 2026 it settled almost 5 million payments, up from 2.7 million the quarter before2.
US instant payment volumes are compounding
FedNow Service settled payments per quarter, thousands (k)
Note: Settled customer credit transfers; figures rounded to the nearest thousand.
Source: Federal Reserve Financial Services, “FedNow Service volume and value statistics” (2026)
Instant payments also raise the operational stakes. In the first year of DORA incident reporting, 16% of major ICT incidents in the EU financial sector occurred in the payments sector6. An outage in a 24/7 rail is a liquidity event as well as a service event.
Instant payments also change deposit behaviour at the margin. When moving money is free, instant and available at any hour, the friction that once slowed outflows disappears. That does not make every bank more vulnerable, but it does mean deposit stability assumptions built on pre-instant behaviour deserve fresh scrutiny, especially for concentrated or highly connected depositor bases like the one described in the 2023 review1.
From monthly monitoring to live management
The Basel Committee's intraday liquidity monitoring tools (BCBS 248) were published in 2013, with monthly reporting starting in 2015. They set out seven quantitative tools covering peak intraday liquidity usage, available intraday liquidity, total payments and time-specific obligations5. They remain the right measures. What has changed is how often they matter. A monthly report tells a supervisor what happened. A live position tells a treasurer what to do next.
The underlying data problem is familiar. In many banks the positions that treasury sees are rebuilt from several systems after the fact: core banking, payment hubs, correspondent accounts, custody and collateral management. Each has its own cut-off, identifiers and latency. The result is a number that is accurate at end of day and unreliable at midday. Moving to event-level data, where each payment, settlement and collateral movement updates a shared position as it happens, is the change that makes the Basel metrics usable intraday rather than a month later.
In practice, a real-time treasury capability needs four things:
- Event-level position data. Payment, settlement and collateral events streamed from core, payments and custody systems into a single intraday ledger, not rebuilt from end-of-day extracts.
- Behavioural early warning. Outflow velocity, concentration of large depositors and unusual patterns across channels monitored continuously, with thresholds tied to the contingency funding plan.
- Tested contingent funding. Collateral pre-positioned and test transactions run, so that the funding sources in the plan actually work, as the 2023 review found they did not1.
- Around-the-clock operations. Alerting, decision rights and runbooks that function at 2 a.m. on a Sunday, because instant payment rails do.
The immediate agenda is practical. Map the data behind each BCBS 248 metric, find where it arrives late, and fix those feeds first. Rehearse a weekend outflow scenario end to end. Agents can then help with live monitoring, drafting alerts and escalating early, provided the data they watch are timely and the decisions stay with people who have the authority to act.