Batch Against Real Time

34.0 What this chapter gives you#

1. You will be able to explain why a £2,400 salary lands on Thursday while a £248,000 house completion lands before lunch, without reaching for “one system is old”.
2. You will be able to place any rail on two axes — continuous or cyclical, gross or net — and say which of the four boxes CHAPS, Bacs and Faster Payments occupy.
3. You will be able to separate clearing from settlement, and say precisely what happens in seconds in Faster Payments and what happens three times each business day.
4. You will be able to explain how netting discharged £6.05 trillion of Bacs payments in 2025 with average daily interbank settlement of £5.6 billion.
5. You will be able to say what prefunding is, why it converts credit risk into opportunity cost, and what a Net Sender Cap caps.
6. You will be able to read the Bank of England’s RTGS daily timetable and explain why Bacs settles at 9.30am and the card schemes at 11.15am and 12.15pm rather than the other way round.
7. You will be able to say why the £1.8 million average CHAPS payment matters less than the £4,586 median, and what that median reveals about why people use CHAPS.
8. You will be able to explain Herstatt risk from the events of 26 June 1974, and name the four tools that shrink, collateralise, cap or eliminate the gap it describes.
9. You will be able to argue that batch is not simply legacy, citing the recall window, file-level control and predictable funding that real-time rails cannot offer.
10. You will be able to explain why the Bank of England is extending settlement hours, and why that is an argument for making batch safer rather than for abolishing it.

Two payments leave the same British bank at nine o’clock on a Tuesday morning. One is a salary of £2,400 for a warehouse supervisor in Wolverhampton. The other is £248,000 for the completion of a house purchase in Hove. Both are in sterling, both electronic, both instructed by people at screens inside regulated institutions. The salary lands on Thursday. The house money lands before lunch.

The obvious explanation is that one system is old and one is new, and that given enough investment everything will eventually land before lunch. That explanation is wrong, and believing it will make the rest of this volume incomprehensible. The salary is slow on purpose; the house money is expensive on purpose. They are two answers to a question every payment system in every country must settle before it processes a single instruction: when a payment is made, do you act on it now, or write it down and act on a pile of them later?

That is the divide. It runs underneath Bacs and Faster Payments and CHAPS, underneath SEPA and the American ACH, underneath every card scheme in Volume III, and underneath every promise a fintech has made about instant settlement. It is not a divide between good and bad, or modern and legacy. It is a trade between cost, risk, liquidity and operational calm, and Britain runs several payment systems side by side because different payments want different points on that trade.

The plain version#

Imagine a street with two shops on it. Priya runs the bike shop; Tom runs the café opposite. They buy from each other all day long — Tom buying inner tubes and puncture kits for the bikes he lends to customers, Priya buying coffee and sandwiches for herself and her staff. There is no card machine and no bank in this story, just a cash box in each shop and a road between them.

There are two ways to run this street.

The first is that every time money is owed, somebody walks. Tom buys a £6.50 puncture kit, so he takes £6.50 out of his cash box, crosses the road and puts it in Priya’s. Twenty minutes later Priya buys two coffees for £5.40, so she walks back with £5.40. Every purchase is a walk. Nobody ever owes anybody anything, because the moment a debt exists it is already paid.

The second is that they keep a slate — a chalkboard by the till. Tom writes “£6.50 to Priya”. Priya writes “£5.40 to Tom”. Nobody walks. At six o’clock they compare slates, work out who is ahead, and one person makes one walk with the difference.

Now put real numbers on a real Tuesday. Tom buys from Priya eleven times, totalling £54.30. Priya buys from Tom nine times, totalling £49.80. Under the first system there are twenty walks across the road, moving £104.10 of coins. Under the second there is one walk at six o’clock, moving £4.50.

Four pounds fifty instead of a hundred and four pounds ten; one crossing instead of twenty. The slate made nobody richer or poorer — Tom and Priya end the day exactly where they would have either way. What it did was cancel the payments against each other before any coins moved. That cancelling is the single most useful trick in money, and the whole first half of this book has been building towards it.

Why the slate gets better as the street gets longer#

Add shops. Now there are twelve businesses on the street and by the end of the day they have made two hundred payments to each other. Under the walking system that is two hundred crossings. Under the slate system, all twelve slates go to the newsagent at six o’clock, who works out that seven of the twelve owe money overall and five are owed it, and that the totals come to £312. Seven people walk to the newsagent with money; five collect from her. Twelve walks, not two hundred.

That is roughly what Bacs does with British payroll, and the numbers are not small. In 2025, Bacs handled 6.86 billion payments worth £6.05 trillion. Because it settles once each working day, the amount that actually had to be moved between banks averaged £5.6 billion a day — somewhere near £1.4 trillion across a working year to discharge £6.05 trillion of payments. Roughly four pounds of payments for every pound that moves. The other three pounds cancelled themselves out on the slate.

What the slate costs you#

The slate is not free, and a bright twelve-year-old spots the problem before an adult does: what happens if a shop shuts before six o’clock?

Suppose Tom sells the café at four and disappears. Priya has been serving him all day on the slate. Under the walking system she would already have his coins in her box; under the slate system she has chalk. She spent the day handing over inner tubes in exchange for a number on a board, and the number is now worthless.

That gap — the hours between the goods changing hands and the money changing hands — is the price of the slate. Everybody on the street is trusting everybody else to still be solvent at six o’clock. It usually works. Occasionally, catastrophically, it does not, and that is why some payments do not use a slate at all.

The trolley and the runner#

The other half of the divide is about timing rather than cancelling, and the school post room is the better picture for it.

A school office has a trolley. Letters posted into the internal tray are taken round the building once a day at four o’clock. This is efficient: one person, one circuit, hundreds of letters. It also means a letter dropped in the tray at 4:05pm sits there for twenty-four hours, not because anybody is being slow but because the trolley has gone. There is a cut-off, and being one minute the wrong side of it costs a whole day.

The alternative is a runner, who takes each letter to its recipient immediately. That is fast for you, expensive per letter, and it means the school must keep a runner sitting available at all times, doing nothing, in case somebody has a letter.

Both exist in British money. The trolley is Bacs, which takes three working days on a fixed timetable and carries most salaries and Direct Debits. The runner is CHAPS, which will move a house purchase across London this afternoon and charge for it. And there is a third thing, Faster Payments, which looks like a runner to you and behaves like a trolley behind the scenes: the message about your money sprints across in seconds, while the coins between the banks go on a trolley three times a day.

The numbers that make the point#

Here is the whole argument in one comparison. In 2025, Bacs handled 6.86 billion payments; CHAPS handled 53.3 million, about one for every hundred and twenty-nine Bacs payments. But the money that had to be moved between banks to settle Bacs averaged £5.6 billion a day, while CHAPS moved £371.3 billion a day. The system with a hundred and twenty-nine times fewer payments moved sixty-six times more actual money.

That is not a paradox. It is the design. Bacs carries an enormous number of small, predictable, cancellable payments and squeezes them down before anything moves. CHAPS carries a small number of large, urgent, un-cancellable payments and moves each one on its own, immediately, in full.

You should now be able to say the following at dinner and be right. Some payment systems collect instructions and process them together, which is cheap and lets payments cancel each other out, but leaves a gap in which somebody could fail. Others act on each instruction the moment it arrives, closing the gap but requiring everybody to keep a great deal of money sitting ready at all times. Britain runs both, because British payments are not all the same kind of payment.

Where the plain version stops being true#

The street with the slate is a good picture of netting and a poor picture of a payment system. Four corrections, and each one matters for what follows.

The first correction is that “batch” and “slow” are different things, and so are “real time” and “instant”. The analogy makes it sound as though the trolley is slow because collecting things takes time. It is not. A Bacs file of four hundred thousand salary payments is validated, sorted and distributed by the central infrastructure in a tiny fraction of the three days it takes to reach people’s accounts. The three days are a rulebook, not a speed limit — a deliberate agreement about which day is input, which is processing and which is entry, so that every bank knows exactly when to expect what. Double the computing power tomorrow and the cycle would still be three days.

The reverse correction is sharper. Faster Payments credits the person you are paying in seconds, twenty-four hours a day. But the banks do not settle with each other in seconds. As of writing, the net obligations arising from Faster Payments settle in the Bank of England’s system three times each business day. What happens in seconds is the clearing — the exchange of the instruction and the crediting of the customer. What happens three times a day is the settlement — the movement of central bank money between the banks. Your money is available immediately; the banks square up later. Confusing those two clocks is the commonest error in payments writing.

The second correction is that the slate hides who is carrying the risk, and modern systems do not leave it lying around. In the analogy Priya simply absorbs the loss if Tom disappears. That is not how a designed payment system works. The customer receiving a Faster Payment gets usable funds immediately while the sending bank’s obligation is not discharged until the next settlement cycle, so the exposure is real — and it is managed rather than tolerated. Bacs, Faster Payments and the cheque-based Image Clearing System all use prefunding: each settling participant must hold cash in a segregated account at the Bank of England covering the largest net debit it could possibly reach before the next settlement, and if it fails, that ring-fenced cash completes the settlement. The Bank’s own description is unambiguous — prefunding eliminates settlement risk in the systems that use it, at the cost of tying up liquidity.

So the honest version of the slate is that Priya may trade with Tom on chalk only up to a limit, and Tom deposited that limit in cash with a trusted third party before the day began. The risk did not vanish; it became a liquidity cost.

The third correction is that there is one cash box in the analogy and two tiers of money in reality. When Bacs nets four hundred thousand salary payments down to a single figure, what finally moves is not “money” in the sense a customer means. It is a balance in a settlement account at the Bank of England — central bank money, the only kind of sterling that can extinguish an obligation between banks with finality. The salary in the supervisor’s account is commercial bank money: a liability of his bank to him. Netting compresses obligations denominated in commercial bank money; settlement discharges the residue in central bank money. A system that clears in one form of money and settles in another has a seam in it, and every argument in this volume about risk, finality and failure lives in that seam.

The fourth correction is that “batch versus real time” is not one axis but two, and there are four boxes, not two. The plain version blurs the question of when obligations are calculated with the question of how. A system can process each payment individually the moment it arrives, or in periodic cycles; independently, it can settle each payment at face value, or settle only the net difference. CHAPS is individual and gross. Bacs is cyclical and net. Faster Payments is individual for clearing and cyclical and net for settlement — a hybrid that did not exist in Britain before 2008 and now carries most of the country’s non-card retail payments. Once you see the two axes, the industry’s vocabulary stops sounding contradictory.

The technical version#

Two axes, four quadrants#

Clearing is the exchange, validation and reconciliation of payment instructions between participants, including calculation of the resulting obligations. Settlement is the discharge of those obligations by the transfer of funds — for systemically important sterling systems, in central bank money across accounts at the Bank of England. Chapter 7 established this distinction; here it becomes the coordinate system.

The first axis is timing: does the system act on each instruction as it arrives, or accumulate instructions and act on a defined cycle? The second is obligation calculation: does each payment settle at face value, or are payments offset so that only a net position settles?

The bottom-right box is where most payments in most countries live, because the arithmetic is cheapest there. The top-left box is where most value lives, because the risk is smallest.

What the Bank of England actually operates#

As of writing (August 2026), the Bank of England supports four settlement models across its real-time gross settlement service: real-time gross settlement, delivery versus payment, and prefunded and unfunded variants of deferred net settlement.

Real-time gross settlement is used by exactly one system: CHAPS. Delivery versus payment is used by the payment leg embedded in CREST, the securities settlement system operated by Euroclear UK & International, where sterling settlement takes place in high-frequency cycles through the day, cash settling against securities in real time against segregated liquidity.

Deferred net settlement is used by seven sterling systems, all settling their net obligations across accounts in RTGS: Bacs, Faster Payments, the cheque-based Image Clearing System, LINK, Mastercard Europe, PEXA and Visa Europe. Each determines its own access criteria and its own number and duration of settlement cycles. Sterling settlement for CLS, the foreign exchange settlement system, runs through CLS’s direct participation in CHAPS rather than as a separate model.

The relative sizes are worth stating precisely, because most people’s intuition is wrong by two orders of magnitude. In the second quarter of 2026, average daily settlement across RTGS totalled £902,018 million: CREST delivery-versus-payment £477,841 million, CHAPS £411,562 million, and the seven retail net settlement systems between them £12,615 million — Bacs £6,205 million, Visa Europe £2,356 million, Faster Payments £2,298 million, Mastercard £1,480 million, LINK £238 million, cheque imaging £37 million, PEXA £1 million. Every retail payment made in Britain by every non-card rail and both major card schemes, netted down, comes to about three per cent of what CHAPS settles the same day.

The RTGS daily timetable#

The clearest single artefact in British payments is the Bank of England’s published timetable of settlement events, the skeleton on which everything else in this volume hangs. Accurate at time of writing, the sterling settlement day runs as follows, with the caveat the Bank itself attaches: timings can vary in contingency.

Two things deserve underlining. First, the deferred systems are not scattered randomly; their cycles are placed so that money arriving from one is available to fund obligations in the next — Bacs at 9.30am, before LINK at 11am and the card schemes at 11.15am and 12.15pm. Second, CHAPS has two closing times: 5.40pm for customer payments carried in the ISO 20022 pacs.008 message, and 6pm for financial institution payments in pacs.009. The gap gives banks twenty minutes to move liquidity between themselves after the last customer payment.

Bacs: the shape of a three-working-day cycle#

Bacs is the archetype of deferred net settlement in Britain, running since 1968; over 183.6 billion transactions have been credited or debited to British accounts through it. It carries two products: Direct Debit, where the beneficiary pulls funds under a mandate, and Bacs Direct Credit, where the payer pushes them. In 2025 it processed 6.86 billion payments worth £6.05 trillion, including 5.0 billion Direct Debits, an annual high.

The cycle has three named days. Input day (day one) is the last day on which a service user or a Bacs approved bureau may submit a payment file for that cycle. Files must be transmitted to Bacs between 07.00 and 22.30, normally over Bacstel-IP, and may be submitted up to thirty days ahead of the intended payment date — which is why a payroll bureau can build December’s run in November. Processing day (day two) is when the central infrastructure sorts the file and delivers the resulting items to each receiving payment service provider, which processes each payment against its own customer accounts. Entry day (day three) is when payments are credited to recipients’ accounts and debited from the payer’s, and it is the day the interbank obligation is settled: Bacs net settlement takes place once each business day in RTGS, at 9.30am.

“Day” here means working day. A file submitted on a Thursday enters accounts the following Monday; a file submitted before a bank holiday weekend waits for the calendar. That is the source of most of the frustration ordinary people feel with Bacs, and of most of its value: because the timetable is fixed and published, a payroll department knows in October exactly which day in December the money becomes irreversible, and can plan its funding to the pound.

There is also a control feature real-time systems cannot offer. Because the file sits in the cycle for two days before entry, payments can be recalled after submission, provided the payer’s provider is notified before a specified cut-off. A submitted batch is not yet a payment made.

Faster Payments: real-time clearing, deferred net settlement#

The Faster Payment System, operated by Pay.UK, is the hybrid, and understanding it correctly is the point at which most people’s model of British payments either becomes accurate or stays wrong forever.

On the customer-facing side it is a real-time system: payments are cleared and the beneficiary credited within seconds, day and night, three hundred and sixty-five days a year. It supports payments of up to £1 million, and it is worth being exact about that limit. £1 million is the value limit applied by the central infrastructure. Individual banks and building societies apply their own, usually much lower, limits, which vary by channel — mobile app, online banking, telephone and branch may face four different ceilings at the same institution. When a customer says “my bank’s Faster Payments limit is £25,000”, they are describing their bank’s policy, not the rail.

On the interbank side it is a deferred net settlement system, settling three times each business day in RTGS, at 7am, 1pm and 5pm as of writing. In 2025 Faster Payments processed 5.55 billion transactions worth £4.84 trillion, while the net obligations settling in RTGS averaged £2,070 million a day. On a 365-day denominator that is around £13 billion of clearing a day compressed to around £2 billion of settlement, better than six to one. The comparison is indicative rather than exact, because clearing runs every calendar day while settlement runs only on business days, so weekend traffic is carried into Monday’s cycles.

The exposure between cycles is handled by prefunding. Each direct settling participant holds funds in a dedicated prefunding account at the Bank of England sufficient to cover the maximum net debit position it could reach before the next settlement. The cap on that position is the participant’s Net Sender Cap. If a participant fails intra-cycle, the ring-fenced cash completes the settlement and no other participant takes a credit loss.

That is an elegant answer to the problem the plain version raised, and it is not free: prefunding converts credit risk into opportunity cost, because cash in a segregated account is cash not being used for anything else. Pay.UK’s framework for setting these caps has historically been prescriptive, built around a Minimum Net Sender Cap and a Peak Contingency Value that participants had to hold at all times, and Pay.UK announced in 2026 that it is moving to a more flexible model. Treat the detail as live; the principle — a capped, collateralised maximum debit — is stable.

The cost of getting prefunding wrong is not theoretical. The Bank’s service records note that on 8 September 2021 the first Faster Payments settlement cycle of the day was delayed by forty-five minutes because a participant had insufficient funds in its settlement account. A system instant to customers can still be held up at the settlement layer by one institution’s liquidity management.

CHAPS: gross, real time, and what that costs#

CHAPS is the sterling high-value system, operated by the Bank of England since November 2017, with more than thirty direct participants and several thousand institutions accessing it indirectly through them. Payment obligations between settlement participants are settled individually, on a gross basis, in RTGS, throughout the business day, and settlement is final and irrevocable between direct participants at the moment it occurs. There is no minimum and no maximum payment value.

The system is usually open from 6am to 6pm, Monday to Friday, excluding bank and public holidays in England and Wales, with customer payments closing at 5.40pm. Until 17 June 2016 the settlement day was 6am to 4.20pm; it was extended by one hour and forty minutes from 20 June 2016, and in the following summer an average of 6.1 per cent of CHAPS daily values and 5.3 per cent of volumes settled in the additional hundred minutes.

This is where the phrase “early-afternoon cut-off” needs handling carefully, because it is repeated everywhere and it is not a system fact. The scheme cut-off for customer payments is 5.40pm. What is early-afternoon is the internal cut-off imposed by individual banks on their own customers, set earlier — often much earlier — to leave room for checks, funding and exception handling. The Bank states this explicitly: customer cut-off times set by providers may be earlier. A solicitor who says “we must instruct by two o’clock” is describing their bank, not CHAPS.

Direct participants may submit payments as urgent, settling immediately subject to available liquidity, or as non-urgent, accepting some delay in exchange for liquidity efficiency. Non-urgent settlement is usually available from 6am until 5.30pm, after which outstanding items are automatically promoted to urgent. Non-urgent payments run through a liquidity saving mechanism of matching cycles, offsetting payments so that a smaller amount of liquidity discharges a larger amount of value; on average they settle around eight minutes after submission. Even a real-time gross system, then, quietly borrows the netting idea — not to defer settlement but to reduce the liquidity needed to achieve it.

The scale figures for 2025 are these. CHAPS volumes grew 1.1 per cent to a record 53.3 million payments, an average of 210,482 per working day. Total value settled rose 7.3 per cent to £93.9 trillion, an average of £371.3 billion a day. The average payment value was £1.8 million; the median was £4,586. Ninety-four per cent of CHAPS payments were for £1 million or less. Financial institution payments in pacs.009 were 24 per cent of payments but 73 per cent of value. CHAPS represents 0.4 per cent of UK payment volumes but 91 per cent of sterling payment values, excluding flows internalised within providers, and it settles the annual GDP of the United Kingdom every nine working days.

The median matters. If you believe CHAPS is only for enormous wholesale transfers, the £1.8 million average confirms it; the £4,586 median tells the truth. The typical CHAPS payment goes through CHAPS not because it is large but because it must be certain and it must be today.

The arithmetic of netting, and the arithmetic of liquidity#

Under gross settlement, a participant must fund every outgoing payment at the moment it settles, from liquidity it holds. Its peak requirement over the day depends on the timing and size of its outflows relative to its inflows: pay out £3 billion before receiving anything and you need £3 billion available. CHAPS direct participants manage this with their own reserves, secured intraday liquidity provided by the Bank against eligible collateral, and the liquidity saving mechanism, which reduces the amount required by matching offsetting payments rather than deferring them.

Under deferred net settlement, a participant needs only enough liquidity to cover its net position at the settlement point, and if it is a net receiver, none at all. Against that, it must either accept credit exposure to other participants between cycles or prefund the maximum net debit it could reach, pinning down liquidity for the whole cycle regardless of its actual position. Set the two side by side using published figures.

Bacs and Faster Payments carry payments of almost identical average size — £880 against £870 — yet one takes three working days and the other takes seconds. The difference is not size but purpose: Bacs carries predictable, scheduled flows where certainty of date is worth more than speed, and Faster Payments carries unscheduled ones where speed is the point. And CHAPS, carrying 0.8 per cent of the payment count of Bacs, moves sixty-six times as much central bank money each day. Real-time gross settlement is not a tweak to netting; it is a different liquidity regime, affordable only because the payments that use it are few.

Settlement risk, and why anybody bothers#

The reason the whole industry does not simply net everything and settle once a month is a Thursday in June 1974.

Bankhaus Herstatt was a German bank active in foreign exchange. On 26 June 1974 its banking licence was withdrawn and it was ordered into liquidation during the German banking day. Counterparties had already paid Deutsche Marks to Herstatt in Frankfurt that morning; the corresponding dollar payments, due in New York where the business day had not yet ended, were never made. They had irrevocably delivered one leg and never received the other. The exposure created by settling two legs of a transaction at different times, in different time zones, with a gap in between, has been called Herstatt risk ever since, and it is the reason CLS exists — a system settling both legs of a foreign exchange transaction simultaneously, whose sterling settlement runs through direct participation in CHAPS, with pay-in and pay-out deadlines every hour between 7am and 11am on the timetable above.

Herstatt is the general case of Priya and Tom. Any gap between the creation of an obligation and its discharge is a window in which a participant can fail and somebody else discovers they were holding a promise rather than money. The toolkit for that window is small: shrink it by settling more often, collateralise it by prefunding, cap it with net sender caps, or eliminate it by settling gross, in real time, in central bank money.

Finality is the legal counterpart. A settlement that has occurred in RTGS cannot be unwound by a participant’s subsequent insolvency. In the United Kingdom that protection comes from the Financial Markets and Insolvency (Settlement Finality) Regulations 1999, which shield transfer orders in designated systems from insolvency law that would otherwise reverse them, together with the recognition regime for payment system operators under the Banking Act 2009. Without the legal certainty the mechanical certainty would be worthless: a liquidator could reach back and undo it.

The Bank of England’s current work on settlement hours is framed in exactly these terms. Its consultation paper of 18 May 2026 notes that longer RTGS hours would reduce settlement risk by enabling more frequent net settlement overnight and at weekends, observing that over an Easter weekend the gap between settlements runs from Thursday evening to Tuesday morning. Four and a half days of retail payments accumulating with no settlement in central bank money is the largest exposure window in the sterling system, and it exists because the calendar has holidays in it.

Cost per transaction, and what batch buys you#

The cost argument is usually asserted and rarely evidenced. It can be evidenced, starting with how the settlement agent charges. The Bank of England operates RTGS and CHAPS on a full cost recovery basis. The CHAPS tariff is an annual participation fee for each direct participant plus a per-item fee on debits. For the retail net settlement systems, the Bank charges an annual fee per settlement participant and no per-item fee at all. That is the economics of the divide, written into a published price list: a gross system consumes settlement resource per payment and is priced per payment, while a net system consumes it per cycle and its price does not vary with how many payments the cycle contained.

The same shape recurs at every layer. In a batch rail, the marginal cost of the four-hundred-thousandth record in a payroll file is a few dozen bytes and a fraction of a second on infrastructure that was going to run anyway. In a real-time rail it is a full request-and-response round trip with validation, fraud scoring, sanctions screening, an account credit, a customer notification and a durable audit record, executed inside a latency budget of single-digit seconds, on infrastructure sized for peak rather than average and available continuously.

Operational cost differs in kind rather than degree. A batch system has a natural quiet period and an obvious point at which yesterday is closed and reconciliation can run against a fixed, complete set of records; it can be patched and tested between cycles; and its usual failure mode is a late file, recoverable within the cycle without any customer seeing anything. A twenty-four-hour real-time system has none of that: no natural end of day, so reconciliation is cut artificially against a moving target; no maintenance window, so upgrades happen live; and a failure mode that is immediately visible, because the customer is standing there watching for the money. Chapter 53 argues that reconciliation is where fintechs actually die; the reason is that real time removes the seam reconciliation was designed to use.

Batch has three further properties the rhetoric of the last fifteen years has treated as legacy defects and which are in fact advantages. It gives a recall window: a Bacs submission can be withdrawn before the cut-off, whereas a completed Faster Payment cannot be recalled at all, recovery being a request for goodwill rather than a mechanism — which is why authorised push payment fraud is a Faster Payments problem far more than a Bacs problem. It gives file-level control: a failed file is rejected, corrected and resubmitted as a unit, whereas a systematic error on a real-time rail produces thousands of individually completed wrong payments. And it gives predictability of funding: a treasurer running payroll on Bacs knows the exact date and amount days in advance.

Why real time is not simply better#

Real time eliminates settlement risk in the gross model and shortens it in the hybrid. It gives customers what they want, and it enables use cases — refunds while the customer is still in the shop, gig-economy payouts on completion — that a three-day cycle cannot serve at all.

Against that, it costs more per payment at every layer from the settlement agent downwards. It demands far more liquidity, as intraday balances for gross settlement or prefunded collateral for the hybrid. It removes the recall window, converting fraud from a controllable operational problem into a customer-harm problem with statutory consequences. It removes the maintenance window and the reconciliation seam — the Bank’s own consultation weighs a twenty-two-hour settlement day, preserving a defined window for maintenance, upgrades and reconciliation, against a 23.5-hour day that maximises availability but reduces the opportunity to maintain overnight. And it makes availability a systemic obligation rather than a service target: a batch system down for two hours in the afternoon has usually harmed nobody, whereas a real-time system down for two hours has visibly failed thousands of people.

The Bank’s published availability record makes the last point concrete. On 18 July 2024 a 245-minute outage to CHAPS settlement was caused by an issue with Swift’s Y-Copy service globally; settlement was extended by around ninety minutes and everything submitted settled successfully. On 31 July 2024 a 91-minute disruption caused by an expired certificate in the Bank’s IT infrastructure delayed CHAPS and a number of retail settlements. On 14 August 2023 a technical issue kept RTGS closed from 6am until 12.20pm, everything nonetheless settling before the end of the day. These are the incident records of a well-run system, and they are instructive precisely because the deferred systems absorbed the disruption while the real-time one propagated it.

The spectrum, and where the rest of this volume sits#

Every rail in the chapters that follow can be placed on the two axes.

The last row is the one people get wrong most often. Open Banking is an interface standard that lets an authorised third party initiate a payment from your account with your consent. It does not clear, does not net, does not settle, and holds no accounts at the Bank of England. When a payment initiated through an Open Banking journey arrives in seconds, that is Faster Payments arriving in seconds. Chapter 44 develops this; the point to carry forward is that an API is not a settlement model, and no API design changes where a payment sits on the two axes above.

What is moving, as of writing#

The batch–real-time frontier is where most of the change is happening. The following are live as of August 2026 and should each be checked against the primary source before being relied upon.

The Bank of England migrated RTGS to a new core ledger and settlement engine on 28 April 2025, replacing the Enquiry Link user interface with BERTI. In February 2026 it announced that CHAPS settlement will open at 01:30 rather than 06:00, Monday to Friday, from September 2027, subject to final confirmation of timelines with affected direct participants. Its consultation paper of 18 May 2026 sets out the next steps towards near-24x7 settlement: an additional settlement day at the weekend, most likely Sunday, with certain UK bank holidays, but not before 2029; then twenty-two hours a day across Monday to Friday and one weekend day, not before 2031; and in the longer term either twenty-two hours every day with a daily maintenance window, or near-continuous settlement with a short daily downtime. Today there are twelve hours in which CHAPS does not settle, reducing to seven and a half in September 2027.

Notice what that sequence is. It is not the abolition of batch. It is the systematic shortening of the windows during which deferred systems accumulate unsettled obligations, by making the gross system available at more times. The Bank’s stated reason for prioritising weekend settlement is that it addresses the longest period during which settlement is unavailable and reduces the risk and liquidity costs borne by the net settlement systems. Real time is being extended to make batch safer.

In the euro area, Regulation (EU) 2024/886 — the Instant Payments Regulation — entered into force in April 2024, with obligations taking effect for euro-area credit institutions on 9 January 2025 and 9 October 2025, and later for electronic money institutions, payment institutions and providers in non-euro EEA countries. The European Payments Council’s 2025 SEPA Instant Credit Transfer rulebook, in effect since 5 October 2025, gives the payer’s provider, the payee’s provider and the clearing and settlement mechanism nine seconds in total to make funds available, within the ten-second maximum set by the amended SEPA Regulation, and removes the scheme-level maximum transaction amount.

In the United States, the Nacha rules raised the Same Day ACH per-payment limit to $1 million from 18 March 2022, across three daily processing windows; entries above the limit submitted into a same-day window are not rejected but processed for next-day settlement in the next available window. Nacha has announced a further increase to $10 million effective 17 September 2027. That is a batch system being pushed up the value curve rather than replaced by a real-time one — the clearest evidence available that the two designs are converging rather than one winning.

A note for implementers#

Three design consequences follow, and they are why this chapter comes before the rail-specific ones.

Value date is not processing date, and conflating them produces wrong statements, wrong interest and wrong reconciliation. In a batch rail the debit, credit and settlement all carry the same value date by rulebook design, and that date is not the date you submitted; in a hybrid rail the customer credit and the interbank settlement fall on different clocks and sometimes different days.

Cut-offs are business logic, not configuration. A submission at 22:31 on input day is not late; it is a submission into the next cycle, with a different entry day, value date and funding requirement. Systems that treat a cut-off as a warning rather than a state transition generate an expensive class of incident.

And real-time rails require idempotency while batch rails require reconciliation, neither substituting for the other. A real-time payment that times out may or may not have happened, and the only safe response is a retry that cannot double-pay. A batch file that was accepted may still contain items rejected downstream, and the only safe response is to reconcile the returned reports against what you submitted, item by item.

Everything from here is a specific answer to the question this chapter posed. Bacs answers with three days and one settlement. Faster Payments answers with seconds for the customer and three settlements for the banks. CHAPS answers by refusing to accumulate anything at all. SEPA, SWIFT correspondent banking, UPI and the instant rails elsewhere each answer differently again, shaped by their own regulators, their own liquidity conditions and their own tolerance for the gap between a promise and a payment.

34.98 Common wrong ideas#

Wrong: batch rails are slow because collecting and processing takes time. Right: a Bacs file of four hundred thousand salary payments is validated, sorted and distributed in a tiny fraction of the cycle; the three days are a rulebook, and doubling the computing power tomorrow would not change them.

Wrong: Faster Payments settles in seconds. Right: the beneficiary is credited in seconds, but the net interbank obligations settle in RTGS three times each business day; clearing and settlement run on different clocks.

Wrong: deferred net settlement leaves participants exposed to each other between cycles. Right: Bacs, Faster Payments and the Image Clearing System all prefund the maximum net debit in segregated accounts at the Bank of England, so the exposure becomes a liquidity cost rather than a credit loss.

Wrong: what settles between banks is the same money the customer sees. Right: netting compresses obligations denominated in commercial bank money, while settlement discharges the residue in central bank money — two kinds of money and two sets of books.

Wrong: CHAPS has an early-afternoon cut-off. Right: the scheme cut-off for customer payments is 5.40pm; the early-afternoon deadline is the individual bank’s internal cut-off imposed on its own customers.

Wrong: CHAPS is only for enormous wholesale transfers. Right: the average is £1.8 million but the median is £4,586 and 94 per cent of payments are for £1 million or less, so the typical CHAPS payment goes that way because it must be certain and today.

Wrong: a bank quoting a £25,000 Faster Payments limit is describing the rail. Right: the central infrastructure limit is £1 million, and bank and channel limits are policy layered on top of it.

Wrong: Open Banking is a payment rail. Right: it is an interface standard that initiates over Faster Payments; it does not clear, net or settle, and holds no account at the Bank of England.

Wrong: a real-time gross system never nets anything. Right: CHAPS non-urgent payments run through a liquidity saving mechanism of matching cycles that offsets payments so less liquidity discharges more value, settling on average about eight minutes after submission.

Wrong: the direction of travel is the abolition of batch. Right: the Bank is extending gross settlement hours to shorten the windows in which deferred systems accumulate unsettled obligations, and Nacha is pushing Same Day ACH up the value curve — the two designs are converging, not competing.

34.99 Chapter summary in 20 lines#

1. Every payment system must settle one question before it processes a single instruction: act on each payment now, or write it down and act on a pile of them later.
2. Acting later lets payments cancel against each other, which is the most useful trick in money: twelve businesses making two hundred payments can square up with twelve walks rather than two hundred.
3. Bacs is the industrial version of that trick, discharging £6.05 trillion of payments in 2025 with an average of £5.6 billion a day actually moving between banks.
4. The price of netting is the gap between goods changing hands and money changing hands, during which a participant can fail.
5. The other half of the divide is timing: a trolley that runs a fixed circuit with a cut-off, against a runner who leaves immediately and costs more per letter.
6. Bacs is the trolley, CHAPS is the runner, and Faster Payments looks like a runner to the customer while behaving like a trolley between the banks.
7. Batch is not the same as slow, because the three Bacs days are an agreement about input, processing and entry rather than a limit on computing speed.
8. Real time is not the same as instant, because Faster Payments clears in seconds and settles its net obligations in RTGS three times each business day.
9. Designed systems do not leave the settlement gap lying around: prefunding in segregated Bank of England accounts converts credit risk into an opportunity cost.
10. Clearing happens in commercial bank money and settlement in central bank money, and every argument in this volume about risk, finality and failure lives in that seam.
11. Timing and obligation calculation are two independent axes giving four boxes, and CHAPS is continuous and gross, Bacs is cyclical and net, and Faster Payments is real time for clearing and cyclical and net for settlement.
12. The Bank of England runs real-time gross settlement for CHAPS alone, delivery versus payment for CREST, and prefunded and unfunded deferred net settlement for seven sterling systems.
13. In the second quarter of 2026 those seven retail systems together settled £12,615 million a day against CHAPS’s £411,562 million, so every retail rail netted down is about three per cent of what CHAPS settles.
14. The published RTGS timetable is the skeleton of the sterling day, and its cycles are ordered so that money arriving from one settlement is available to fund obligations in the next.
15. Bacs and Faster Payments carry payments of almost identical average size, about £880 against £870, so the difference between them is purpose rather than value: scheduled certainty against unscheduled speed.
16. CHAPS carries 0.4 per cent of UK payment volumes and 91 per cent of sterling payment values, and settles the annual GDP of the United Kingdom every nine working days.
17. Herstatt’s failure on 26 June 1974 is the general case of the settlement gap, and the toolkit for it is small: settle more often, prefund it, cap it, or settle gross in central bank money.
18. Legal finality under the Settlement Finality Regulations 1999 is what stops a liquidator undoing what RTGS has done, without which the mechanical certainty would be worthless.
19. Gross settlement is priced per item and net settlement per cycle, and batch also buys a recall window, file-level control and predictable funding, which is why authorised push payment fraud is a Faster Payments problem far more than a Bacs one.
20. The frontier is moving through earlier CHAPS opening, weekend settlement and a $10 million Same Day ACH limit, but the direction is convergence: real time is being extended to make batch safer, not to replace it.

Sources: Bank of England (payment and settlement pages, summary of RTGS daily timetable, payment and settlement statistics, CHAPS pages, and the consultation paper “Extending RTGS and CHAPS settlement hours”, 18 May 2026); Pay.UK and Bacs (scheme pages, Bacs processing cycle, Net Sender Cap publications); European Payments Council (SCT Inst scheme and 2025 rulebook); Nacha and Federal Reserve Financial Services (Same Day ACH rules). All timings, limits and figures accurate at time of writing, August 2026.