Securities Settlement: T+1, CCP Novation and CSD Architecture (2026)

A stock trade feels instant. You press buy, the price prints, and the position appears in your account. Yet in the legal and operational sense, nothing has been exchanged yet. Ownership, cash and counterparty risk are still in flight, and the machinery that closes that gap is securities settlement: a chain of matching, guarantee, netting and simultaneous exchange that runs largely out of sight.

That machinery just got much less forgiving. The United States moved to a one-business-day cycle on 28 May 2024, and the EU, UK and Switzerland are aligned on 11 October 2027. Halving the window does not halve the work, because every manual step that once hid inside a second day now has to finish on the trade date.

This post is a systems analysis, not investment advice. You will leave with the trade lifecycle, what a central counterparty (CCP) actually does when it novates, worked netting arithmetic, the three delivery-versus-payment models, how depositories differ, and an architecture for post-trade platforms.

What this covers: lifecycle, T+1 mechanics, CCP novation and netting, DvP models, CSDs (DTC, Euroclear, Clearstream), fails and CSDR penalties, margin, DLT pilots, architecture, a runnable Python demo, and failure modes.

Context and Background

Trading and settlement are different businesses. A trading venue matches a buyer and a seller at a price. Settlement is the later process in which the seller delivers securities, the buyer pays cash, and both records of ownership change. The gap between the two is the settlement cycle, written T+n, where T is the trade date and n counts business days to settlement.

The gap exists for reasons that were once physical. Paper certificates had to be moved, cheques cleared and inventories reconciled. Over decades the industry dematerialised securities into book entries held at a central securities depository (CSD), and centralised counterparty risk in a CCP. The cycle shortened accordingly: the G30 recommended T+3 in 1989, the US adopted T+3 in 1995, T+2 in 2017, and T+1 in 2024. Canada and Mexico moved on the same date as the US.

Risk is the reason regulators keep pushing. While a trade is unsettled, the buyer is exposed to the seller failing to deliver and the seller to the buyer failing to pay. The longer the cycle, the more market movement can accumulate against a defaulter and the more margin the system must hold. The CPMI-IOSCO Principles for Financial Market Infrastructures (PFMI, 2012) turned this into an international standard for CCPs, CSDs and settlement systems, covering credit risk, liquidity risk, settlement finality and default management.

Three institutions recur in this story. DTCC, the Depository Trust & Clearing Corporation, is the US holding company for two core subsidiaries: NSCC, the National Securities Clearing Corporation, which is the CCP for US equities and corporate bonds, and DTC, the Depository Trust Company, which is the US CSD. In Europe, settlement is split between national CSDs, the two international CSDs (Euroclear Bank and Clearstream Banking), and TARGET2-Securities (T2S), the Eurosystem platform that settles in central bank money for participating CSDs. CCPs such as LCH, Eurex Clearing and EuroCCP sit upstream.

If you have read about collateral and margin management for derivatives, many of the same primitives apply here: initial margin, variation margin, default fund contributions and the waterfall. The difference is that cash securities settle in days rather than staying open for years, so the dominant design problem is throughput and timing, not long-dated exposure.

Official anchors for the facts used below are the SEC’s final rule release on shortening the settlement cycle (February 2023), DTCC’s published T+1 statistics, the UK Accelerated Settlement Taskforce and ESMA’s work on the EU transition, and ICMA’s tracker of CSDR settlement discipline. Where I could not reach a primary source this run, I say so.

The Post-Trade Reference Architecture

The post-trade chain has five jobs: agree what was traded, make someone guarantee it, shrink the obligations, exchange cash and securities safely, and keep every party’s books consistent afterwards. Each job belongs to a different system with a different owner, and the architecture is mostly about the interfaces between them.

Figure 1: The securities settlement lifecycle from execution to custody, with margin and netting feeding the settlement instruction.

The diagram reads left to right. Execution creates a trade, capture records it, allocation splits block trades across client accounts, confirmation and affirmation agree the details, the CCP novates and nets, margin is collected, and the CSD performs delivery versus payment. Reconciliation closes the loop against custody records.

Stage by stage: what each step produces

Execution and capture. A venue or broker-dealer matches an order and emits an execution report. The firm’s order and execution management systems capture price, quantity, instrument identifier (ISIN or CUSIP), trade date, and counterparty. This is the first immutable fact in the chain. Keeping it as an append-only event log, as discussed in the event sourcing and bitemporal data post, makes every later correction explainable.

Allocation and confirmation. Institutional orders are often executed in blocks and then split across many funds. Allocation tells the broker how many shares belong to each account. Confirmation is the broker’s statement of the trade terms to the client.

Affirmation. The investment manager or custodian agrees those terms and confirms that settlement instructions are correct. In the US, the DTC-operated affirmation and matching services let this be done electronically, and the SEC’s rules push it to the end of the trade date. Affirmed trades are the ones that settle cleanly, which is why affirmation rate is the number the industry watches.

Clearing. A trade submitted to a CCP is validated and, if accepted, novated. After novation the CCP is the counterparty to both sides. The next section works through this.

Settlement. The CSD debits securities from the deliverer and credits the receiver, while the payment leg moves in central or commercial bank money. Done correctly, either both legs happen or neither does.

Why this chain exists as separate systems

One could imagine a single ledger doing all of this. The industry did not build it that way, because each layer carries a different kind of trust. Trade capture is internal to the firm. Affirmation is a bilateral agreement between firm and client. Clearing is a legal guarantee underwritten by a risk-sharing waterfall. Settlement is a final transfer of title, protected by law. Collapsing them would collapse the legal boundaries too.

The practical consequence is that the weakest interface sets the speed. A trade can only settle as fast as the slowest of: instruction matching, inventory availability, funding, and the cut-off times of the CSD and payment system. The rest of this post keeps returning to that point.

Message standards on the chain

The chain also depends on standardised messages. ISO 15022 (SWIFT MT) has long carried settlement instructions and confirmations, and ISO 20022 equivalents (the sese and semt families) are replacing it in securities, as T2S already uses ISO 20022. If your payments background is in cards, the same migration logic discussed in the ISO 8583 versus ISO 20022 comparison applies: richer structured data in exchange for a harder transition. I have not verified a current industry-wide cut-over date for securities messaging, so treat any specific date you hear as something to check with your CSD.

T+1 Settlement: What Actually Changed

T+1 settlement means a trade settles one business day after the trade date instead of two. The headline is simple. The consequence is that the overnight window between trade date and settlement date, which used to absorb allocation, affirmation, funding and fix-ups, shrank to a single overnight batch.

The US transition

The SEC adopted amendments to Rule 15c6-1 in February 2023, shortening the standard settlement cycle for most broker-dealer transactions from T+2 to T+1, with compliance required from 28 May 2024. The same release added two supporting requirements that are easy to miss. Rule 15c6-2 requires broker-dealers to complete allocation, confirmation and affirmation as soon as technologically practicable and no later than the end of the trade date, where the trade would otherwise be subject to those steps. A new CCP rule, 17Ad-27, requires clearing agencies that provide a central matching service to have policies for straight-through processing. Together they move the work earlier, which is the real point of the rule.

The change was operationally concentrated. DTC set a 9:00 PM Eastern cut-off on the trade date for affirmation, so a firm whose client confirmations arrive at 8 PM has one hour to fix breaks. In practice that pulled processing in Asia-Pacific time zones and Europe into a tighter window as well, since cross-border investors trade US securities with local funding and foreign exchange (FX) legs that may not be open on the same schedule.

What DTCC measured

DTCC published early results after go-live, and they are worth quoting precisely because they come from the operator rather than from commentary. On 29 May 2024, the first full T+1 day, the overall affirmation rate by the 9 PM trade-date cut-off was 94.55%, compared with about 73% at the end of January. Prime brokers reached 98.6%, up from 81%. Central matching with investment manager auto-affirmation reached 97.5%, up from 92%. The slowest segment, custodians and investment managers affirming themselves, reached 84.29%, up from 51%.

Fail rates did not worsen. DTCC reported a continuous net settlement (CNS) fail rate of 1.90% on 29 May, against a 2.01% May average under T+2, and a DTC non-CNS fail rate of 2.92% against 3.24%. A fail rate of about 2% sounds small but is a large number of trades at US volumes, which is why fails and their causes get their own section below.

The NSCC clearing fund also fell. DTCC reported it down 29% from the past-quarter average of $12.8 billion to $9.1 billion. I come back to why that happens, with a clearly labelled illustrative calculation, in the margin section.

Europe, the UK and Switzerland

The European and UK timetable is later and, by design, synchronised. The UK Accelerated Settlement Taskforce recommended 11 October 2027 as the move date, and trade press reported that the EU and Switzerland were aligned on the same date. At the time of reporting the UK date was a taskforce recommendation rather than a final regulatory decision, and I did not verify the final legal instruments this run. The EU legislative amendment to the settlement-cycle provisions of the CSD Regulation (CSDR) and ESMA’s supporting roadmap are the sources to confirm before you plan against that date.

The alignment matters more than the date itself. A UK-only move would have broken the link with EU post-trade infrastructure, since much European settlement runs through T2S and linked CSDs, and depositary receipts, ETFs and cross-listed names would have settled on mismatched cycles. Aligning removes that class of break, but it also means the whole ecosystem, from asset managers to FX desks, faces one cliff-edge.

The original angle: T+1 is a liquidity-timing problem

Most coverage frames T+1 as a technology deadline. A more useful framing is that settlement is a set of time windows that must nest. There is a window for instruction matching, one for securities availability including lending recalls, one for funding and FX, and one for the CSD cut-off. T+2 allowed those windows to overlap loosely. T+1 forces them to nest, so the critical path is the longest chain of dependent steps, not the sum of work. Improving one slow step, such as FX funding, can matter more than speeding up everything else together.

CCP Novation, Netting and the Default Waterfall

A CCP interposes itself between the two original counterparties. This legal act is novation: the original bilateral contract is extinguished and replaced by two new contracts, one between the CCP and the buyer, the other between the CCP and the seller. The CCP becomes the seller to every buyer and the buyer to every seller.

Figure 2: CCP novation replaces one bilateral exposure with two exposures to the CCP, which are then netted per member and margined.

The reason to do this is to convert a web of bilateral credit relationships into a hub-and-spoke structure with one well-capitalised, rule-bound centre. The reason it works is that, once everyone faces the CCP, a member’s obligations in the same security can be added and subtracted regardless of who the original counterparty was.

Novation versus open offer

Not every CCP gets there the same way. Under the novation model, used in the US by NSCC at a defined point after trade comparison, the original contract exists briefly and is then replaced. Under open offer, used by several European CCPs, the CCP’s offer to be counterparty is open to the market and becomes the counterparty at the instant a matched trade is received, so the bilateral contract never exists between members. The legal timing of that transfer is the moment of “trade guarantee”, and it matters for insolvency law: it fixes from when the CCP, rather than the surviving member, owns the risk. Check the specific rulebook, because the wording varies by CCP and jurisdiction.

Worked netting arithmetic

Netting is where the economics come from. Consider one security, XYZ at 50 per share, and four members trading it during the day. The five trades are:

Trade	Buyer	Seller	Shares
1	A	B	1,000
2	B	C	600
3	C	A	700
4	D	A	400
5	A	D	200

Gross, that is 2,900 shares and $145,000 of notional changing hands in five separate deliveries. After novation, we ignore who traded with whom and add up each member’s position. A: +1,000 −700 −400 +200 = +100. B: −1,000 +600 = −400. C: −600 +700 = +100. D: +400 −200 = +200. The positions sum to zero, as they must.

Only B owes securities, 400 shares, and A, C and D receive 100, 100 and 200. Net settlement therefore moves 400 shares and $20,000 instead of 2,900 shares and $145,000. The securities netting efficiency is 1 − 400/2,900 = 86.2%. The cash net is also reduced: B receives $20,000, A pays $5,000, C pays $5,000, D pays $10,000. These figures are illustrative, but the arithmetic is exact and the Python demo later in this post reproduces it.

Real equity netting factors depend on the market. DTCC has described CNS as reducing a very large gross volume of trades to a much smaller net settlement obligation, but I did not retrieve a current percentage this run, so I do not quote one. What the arithmetic shows is the mechanism: netting savings come from members being on both sides of the same name during the day.

The CCP’s risk stack and default waterfall

Netting shrinks the movement of assets, but it concentrates risk in the CCP. Any CCP therefore runs a defence in layers. In broad order:

Initial margin posted by each clearing member, sized to cover potential losses on its positions over a margin period of risk.
Variation margin or marking to market, paid for movements since the last mark.
The defaulter’s default fund contribution, a pre-funded mutualised pool.
The CCP’s own capital committed ahead of other members’ contributions (“skin in the game”, in many CCP designs).
Surviving members’ default fund contributions and, depending on the rulebook, further assessments.

The PFMI require systemically important CCPs to hold resources to cover the default of the participant, and affiliates, with the largest exposure, in extreme but plausible market conditions. Many CCPs that are more complex or systemically significant are required to cover the two largest. The exact standard (cover one or cover two) depends on the CCP’s risk profile and regulation, which is why I recommend checking the specific CCP’s disclosure framework under the PFMI.

What happens when a clearing member defaults

On default, the CCP declares the member in default, freezes its accounts, hedges or auctions its portfolio, and applies the waterfall. For cash equities this is simpler than for derivatives because positions are short-lived, but obligations already guaranteed must still be settled. The CCP continues to deliver to receivers by buying in the missing securities, which is why the guarantee is the product: the buyer’s experience is that the trade settles even if the original seller does not.

This is also why the CCP’s data model matters. It needs exact, auditable position and margin state at every moment. Any mismatch between the clearing system’s view of a member and the member’s own is a reconciliation break, and the reconciliation engine architecture for payments and fintech post covers the matching techniques that apply here as well.

DvP Settlement and the Role of the CSD

Delivery versus payment (DvP) is a settlement mechanism that links the securities transfer and the cash transfer so that delivery occurs if and only if payment occurs. It removes principal risk, the risk that you deliver and are never paid, or pay and never receive. The BIS Committee on Payment and Settlement Systems classified DvP into three models in 1992, and the taxonomy is still the clearest way to compare systems.

Figure 3: The three DvP models differ in whether the securities leg and the cash leg are settled gross, trade by trade, or net, in batches.

The three models

Model 1: gross securities, gross cash. Each trade settles individually and simultaneously, with the securities and cash legs linked on a trade-by-trade basis. Principal risk is eliminated. The price is liquidity: a participant needs the full securities and full cash for each instruction at the moment it settles, with no netting offset.

Model 2: gross securities, net cash. Securities transfer trade by trade during the processing window, but the cash obligations are netted and settled at the end of the cycle. Final securities transfers are conditional on the net cash settlement, which means a failure to pay at the end can force reversal or create exposure. This model needs risk controls such as collateralised credit lines to stay safe.

Model 3: net securities, net cash. Both legs are netted and settled in batches. This minimises the securities and cash that must be on hand, which is the efficient end of the spectrum. It also leaves everyone dependent on the netting cycle succeeding, so a single failure to deliver can unwind the cycle.

Real systems are hybrids. US equity settlement combines a CCP netting layer (NSCC, with its CNS accounting system) with a depository (DTC) that settles the resulting delivery orders. DTC itself runs a Model 2-like arrangement, as I understand its design: securities move in real time through the day, and the cash is settled through a net funds mechanism backed by participant collateral and net debit caps. I did not verify the exact mapping from DTC’s rulebook to the BIS classification this run, so treat that sentence as a description, not an official categorisation.

What a central securities depository does

A central securities depository is the institution that holds securities in dematerialised form and records changes of ownership. It has three core functions: notary (initial recording of an issue), central maintenance (the master record of holdings), and settlement (operating a securities settlement system). In the EU, the CSD Regulation (CSDR, Regulation 909/2014) defines these as the core services and sets authorisation and conduct requirements.

CSDs sit on top of a tiered holding structure. The issuer’s securities are held in an issuer CSD or an investor CSD, banks hold them for clients, and clients hold them through custodians. Each layer keeps its own books, which is why the end investor’s entitlement is a claim within a chain of records, and why reconciliation across the chain is a first-class engineering concern.

DTC, Euroclear and Clearstream compared

Three names dominate cross-border conversations. They are not interchangeable.

Feature	DTC (US)	Euroclear Bank	Clearstream Banking
Home market	United States	International, Belgium-based	International, Luxembourg-based
Role	US CSD for equities, corporates, municipals	International CSD (ICSD) with banking services	ICSD with banking services
Clearing link	NSCC is the CCP, same DTCC group	Settles trades from many CCPs and markets	Settles trades from many CCPs and markets
Cash leg	Net settlement through DTC with participant collateral	Commercial and central bank money	Commercial and central bank money
Distinctive	Immobilised custody and the DTC settlement process	Eurobond market and collateral services	Eurobond market and collateral services

The ICSDs matter because they do two things national CSDs mostly do not: they settle internationally issued bonds that have no home market, and they provide banking-type services including cash accounts, credit and securities lending. That extra layer means their settlement design mixes CSD functions with bank-style liquidity management.

T2S: settling in central bank money across borders

T2S is the Eurosystem’s single platform for securities settlement, launched in 2015 with CSDs migrating in waves. Participating CSDs outsource their settlement engine to T2S while keeping their legal CSD roles. Securities settle against cash held on dedicated cash accounts with central banks, which gives the cash leg the strongest settlement finality. The important architectural lesson is separation: the CSD keeps the legal and custody relationship, while the settlement engine is a shared utility. I do not quote the current participant count because I did not verify it this run.

Settlement Fails, CSDR Penalties and Affirmation

A settlement fail occurs when a party does not deliver securities or cash on the intended settlement date. Fails are normal at a low rate and destructive at a high one, because one fail often cascades: the receiver who did not receive cannot deliver onward, and the chain of dependent trades lengthens.

Why fails happen

The causes cluster into a handful of types. Inventory fails: the seller does not have the securities, often because a loan has not been recalled or a purchase has not itself settled. Funding fails: the buyer lacks cash or credit, increasingly an FX problem for cross-border investors. Instruction fails: the two sides’ instructions do not match on details such as account, quantity or settlement date. Operational and cut-off fails: the instruction arrived too late for the batch.

T+1 shifts the mix. Recalling a securities loan and getting the securities back to a seller in time is harder in one day than two. That is a structural change in securities lending, and it links directly to why affirmation and early allocation dominate T+1 programmes: instruction fails are the most avoidable category.

The CSDR settlement discipline regime

The EU decided to price fails. CSDR’s settlement discipline regime introduced cash penalties for late settlement, and they went live on 1 February 2022. A participant that causes a fail pays a daily penalty, computed per instruction and distributed to the failing party’s counterparty, with the rate depending on the asset class and liquidity. Penalties are not revenue for the CSD, they are a transfer from the failer to the receiver, so the CSD must calculate and settle them every day.

The other half of the regime, mandatory buy-ins, has had a rockier path. Mandatory buy-ins were deferred repeatedly, and the CSDR Refit, politically agreed in 2023, kept them only as a last-resort measure that applies if cash penalties fail to sustain settlement efficiency and if they would not threaten financial stability. ICMA’s tracker also records that in 2024 ESMA floated higher and progressive penalties, which ICMA opposed as disproportionate and unsupported by data, and that ESMA’s report to the Commission on revising the technical standards was expected in 2025 with adoption expected in early 2026. I did not confirm whether that revision has been adopted, and I deliberately do not quote penalty rates, since they are set by technical standards that may have changed.

For engineers, the design implication is that penalty calculation is a ledger problem. Every instruction that settles late generates a per-day, per-instruction line item, which needs to be reproduced exactly by the CSD, the participant and the custodian. Disputes over penalties are reconciliation disputes.

Affirmation, STP and the straight-through rate

Straight-through processing (STP) means a trade flows from execution to settlement without manual intervention. Its measurable proxy is the affirmation rate by the cut-off. The US figures cited earlier are an object lesson: the self-affirming custodian segment rose from 51% to about 84% in four months, while the prime broker segment started high at 81% and finished at 98.6%. Segments with more automation reach the target earlier, and the long tail of manual processes is where fails are born.

A useful rule is to treat affirmation rate as a leading indicator and fail rate as a lagging one. If the first moves in the wrong direction on a given day, the second will follow a day later.

Margin, the Clearing Fund and Why T+1 Reduces Them

A CCP sizes margin to cover the loss it could suffer between a member’s last good margin payment and the point where positions can be closed out, known as the margin period of risk. A shorter settlement cycle shortens the time during which an unsettled position remains a CCP obligation. Less time means less price movement, which means less margin.

An illustrative calculation

Suppose daily price changes are independent with the same volatility. Then the standard deviation of price change over n days scales with the square root of n. Going from a two-day exposure to a one-day exposure reduces a value-at-risk style requirement by a factor of 1/√2, about 0.71, a cut of around 29%. This is a deliberately simplified model: real CCP margin models use filtered historical simulation, stress components, concentration add-ons and floors, and I do not claim that NSCC computes its requirement this way.

It is still instructive. DTCC reported the NSCC clearing fund falling 29% against the past-quarter average, a figure that happens to line up with the square-root illustration. I would treat that as a plausibility check rather than proof, because the clearing fund also moved with market volatility and composition over the period. The practical point stands: broker-dealers’ liquidity needs for clearing fund and margin fell, which was one of the main motivations for the change.

The trade-off that margin hides

Lower margin does not mean lower liquidity need overall. Funding and prefunding of purchases, securities lending recalls and FX for non-US investors become more urgent. Liquidity requirements shift in time from the clearing fund to intraday and overnight funding at the participant level. A firm that models only the clearing fund reduction is looking at the benefit and missing the cost.

Architecture of a Post-Trade Platform

Everything above converges on a software question: how do you build the systems a broker, bank or custodian runs between execution and the CSD? The answer is rarely a monolith. It is an event-driven pipeline in which each stage consumes facts from the previous one, adds its own, and exposes state for exceptions handling.

Figure 4: A post-trade platform built around an append-only event log, with netting, instruction generation, CSD connectivity, a fails monitor and reconciliation.

Core components

Event log as the source of truth. Every trade, amendment, affirmation, instruction and CSD status is an event with a business timestamp and a system timestamp. Bitemporal storage matters here because corrections arrive late: a counterparty amends a trade on T+1 that was agreed on T, and you must be able to answer both “what did we believe at 6 PM on T” and “what do we know now”. A CQRS split gives operations teams read models by account, security and status without loading the write path.

Matching and affirmation hub. This component compares your record of the trade with the counterparty’s or with the central matching service and tracks state transitions: unmatched, matched, affirmed. It is the place to put the T+1 deadline logic, with escalation timers that fire before the 9 PM cut-off rather than after it.

Netting and obligation engine. For participants that net internally (for example, a bank netting across its own books before sending to the CCP), this engine computes positions per account, security and value date, similar to the demo below. For CCPs it is the core and must be deterministic: the same input must always produce the same output, and every net obligation must be explainable back to the gross trades.

Settlement instruction engine. It converts obligations into messages to the CSD or custodian, applies standing settlement instructions, handles partial settlement, and tracks status. It also decides priority when inventory is short. Choose the allocation rule deliberately, because it has financial consequences, as the demo shows.

Fails and penalty monitor. It watches instructions approaching their intended settlement date, flags those at risk using inventory and funding signals, and triggers remediation: recall a loan, borrow stock, partially settle, or cancel and re-instruct. Under CSDR it also reproduces the daily penalty calculation to dispute errors.

Reconciliation. Three-way reconciliation compares your books, the custodian’s statement and the CSD’s. Each break needs an owner, an ageing bucket and an automatic match attempt before a human sees it.

Capacity and latency characteristics

Settlement load is spiky. Volume concentrates at the end of the trade date when allocation, affirmation and instruction creation converge, and again around CSD batch windows. A design that averages throughput over 24 hours is wrong. Size for the burst, with idempotent consumers and replayable queues, because a CSD link that drops mid-batch must resume without duplicate instructions.

Idempotency deserves emphasis. A settlement instruction carries a unique reference, and sending it twice must not produce two settlements. Design the instruction ID as a deterministic function of the obligation and the version of the standing instruction, not a random value generated at send time.

Data model essentials

At minimum, model these entities explicitly: Trade (immutable once executed, amendments as new events), Allocation (many per trade), Obligation (net or gross, by account, security, value date and currency), SettlementInstruction (with status history and a link to its obligation), Position (settled and pending views), Margin call and Penalty. Separate settled position from contractual position. They differ for one or two days, and the difference is exactly the pending settlement risk you are managing.

Runnable Demo: Novation, Netting and Partial DvP

The following Python script implements the mechanics from this post in about fifty lines: novation into CCP legs, multilateral netting of securities and cash, and a DvP step with partial settlement when a net deliverer is short. It uses the same five trades as the worked example. It is a teaching model, not a production system: no margin, no multiple currencies, no calendars.

from collections import defaultdict
from dataclasses import dataclass

@dataclass(frozen=True)
class Trade:
    tid: int
    buyer: str
    seller: str
    isin: str
    qty: int
    price: float

def novate(trades, ccp="CCP"):
    """Replace each bilateral trade with two legs facing the CCP."""
    legs = []
    for t in trades:
        legs.append((t.tid, t.buyer, ccp, t.isin, t.qty, t.price))   # buyer buys from CCP
        legs.append((t.tid, ccp, t.seller, t.isin, t.qty, t.price))  # CCP buys from seller
    return legs

def multilateral_net(trades):
    """Net securities and cash per member per ISIN."""
    sec = defaultdict(int)
    cash = defaultdict(float)
    for t in trades:
        sec[(t.buyer, t.isin)] += t.qty
        sec[(t.seller, t.isin)] -= t.qty
        cash[t.buyer] -= t.qty * t.price
        cash[t.seller] += t.qty * t.price
    return dict(sec), dict(cash)

def settle(sec, price, inventory, balances):
    """DvP per ISIN with partial settlement. Receivers are served pro rata from
    what the net deliverers can actually deliver; cash moves only against
    securities that moved. Returns the shortfall per ISIN."""
    shortfall = {}
    for isin in {i for _, i in sec}:
        givers = {m: -q for (m, i), q in sec.items() if i == isin and q < 0}
        takers = {m: q for (m, i), q in sec.items() if i == isin and q > 0}
        avail = sum(min(inventory.get((m, isin), 0), need) for m, need in givers.items())
        need_total = sum(takers.values())
        ratio = avail / need_total if need_total else 1.0
        for m, need in givers.items():
            d = min(inventory.get((m, isin), 0), need)
            inventory[(m, isin)] = inventory.get((m, isin), 0) - d
            balances[m] = balances.get(m, 0) + d * price
        for m, q in takers.items():
            got = int(q * ratio)
            inventory[(m, isin)] = inventory.get((m, isin), 0) + got
            balances[m] = balances.get(m, 0) - got * price
        shortfall[isin] = need_total - int(sum(int(q * ratio) for q in takers.values()))
    return shortfall

if __name__ == "__main__":
    trades = [
        Trade(1, "A", "B", "XYZ", 1000, 50.0),
        Trade(2, "B", "C", "XYZ", 600, 50.0),
        Trade(3, "C", "A", "XYZ", 700, 50.0),
        Trade(4, "D", "A", "XYZ", 400, 50.0),
        Trade(5, "A", "D", "XYZ", 200, 50.0),
    ]
    gross_qty = sum(t.qty for t in trades)
    gross_cash = sum(t.qty * t.price for t in trades)
    sec, cash = multilateral_net(trades)
    net_qty = sum(q for q in sec.values() if q > 0)
    net_cash = sum(c for c in cash.values() if c > 0)
    print("legs after novation:", len(novate(trades)))
    print("net securities:", {k[0]: v for k, v in sorted(sec.items())})
    print("net cash:", dict(sorted(cash.items())))
    print(f"gross {gross_qty} sh / ${gross_cash:,.0f}  ->  net {net_qty} sh / ${net_cash:,.0f}")
    print(f"netting efficiency: {1 - net_qty / gross_qty:.1%}")
    inv = {("B", "XYZ"): 300}          # B holds only 300 of the 400 it must deliver
    bal = {}
    short = settle(sec, 50.0, inv, bal)
    print("shortfall:", short)
    print("cash after settlement:", dict(sorted(bal.items())))
    print("inventory:", dict(sorted(inv.items())))

Running it prints ten novated legs, net securities of A +100, B −400, C +100 and D +200, and net cash of A −5,000, B +20,000, C −5,000 and D −10,000. Gross volume of 2,900 shares and $145,000 nets to 400 shares and $20,000, an efficiency of 86.2%.

Reading the partial-settlement result

B can deliver only 300 of 400 shares, so there is a shortfall of 100. The settlement step serves receivers pro rata: A and C each receive 75 shares and D receives 150. Cash follows the securities, so B receives $15,000 rather than $20,000, and A, C and D pay $3,750, $3,750 and $7,500. Nobody paid for shares they did not receive, which is the DvP guarantee in miniature.

The pro rata rule is a design choice, not a law. Real systems use priority ordering, cash-based rules, or algorithms that try to maximise settled value, and a CCP typically also begins buy-in or borrowing procedures for the shortfall. Try changing the rule to a first-come ordering and watch which members absorb the fail. That is the policy decision hiding inside “partial settlement”.

DLT, Tokenisation and the Limits of Faster Settlement

Distributed ledger technology (DLT) promises atomic settlement: securities and cash move in a single ledger transaction, which would make settlement effectively instantaneous, often called T+0. The idea is technically elegant. It is less obviously a replacement for what a CCP does, because novation, netting and default management are risk services rather than message-passing problems.

What is verified and what is not

The EU DLT Pilot Regime, Regulation (EU) 2022/858, has applied since 23 March 2023 and creates a sandbox for DLT market infrastructures to operate trading and settlement systems with some regulatory exemptions. Beyond that, central banks and market infrastructures have announced experiments with settling DLT securities against central bank money. I did not verify current statuses, volumes or go-live dates of specific projects this run, so I do not name any as production. Treat claims of “T+0 at scale” as unverified until you see the operator’s own volume statistics.

The structural argument

Two points sit behind the caution. First, instant settlement removes the netting benefit. A participant that previously needed $20,000 to settle a day’s trades may need far more if each trade settles gross and immediately. Real-time gross settlement moves the burden from the CCP’s risk models to the participants’ intraday liquidity. Second, shorter cycles amplify the importance of the pipeline described above. Even a perfect ledger cannot fix a mismatch in an upstream allocation.

My view, which is opinion, not fact: DLT will matter first in asset classes with poor existing infrastructure, such as private funds, collateral mobility and some bond markets, and least in liquid exchange-traded equities where netting, CCP risk management and mature STP already deliver most of the benefit.

Trade-offs, Gotchas, and What Goes Wrong

Settlement systems fail in boring, repeatable ways. Knowing them is worth more than any architecture diagram.

Concentration risk. A CCP or CSD is a single point of failure by design. Its resilience, recovery and wind-down plans are therefore regulatory subjects under the PFMI. As a participant you cannot diversify away from the utility, so your control is limiting your exposure to its operating hours and cut-offs.

Procyclical margin. Margin models that react to volatility raise requirements just when firms are least able to fund them. The market stress episodes of recent years showed how liquidity calls can compound. Shorter cycles help, but the model parameters and buffers still decide the outcome.

Cut-off cliffs. A single late message can push a trade into the next day, and under T+1 the next day is the settlement date. Build alerting around the cut-off, not around the batch completion.

Time zones and FX. For investors outside the US, funding the dollar leg within the T+1 window requires FX trades that settle same day in some currency pairs, with limited hours. The result is operational pressure in Asia-Pacific, and some investors responded by prefunding. This is a real cost that headline fail statistics do not show.

Securities lending recalls. A short seller or a lender who needs shares back has less time. Lending desks must automate recalls and match them to sales earlier.

Partial settlement and tolerance. If your system does not support partial settlement, one short position blocks the whole instruction. If it does, you must reconcile the remainder and manage penalties on it.

Corporate actions interplay. Record dates, ex-dates and entitlement calculations assume a settlement cycle. A change in cycle moves the ex-date relative to the record date, so entitlement logic and claims processing need testing.

Over-reading the numbers. Early T+1 statistics came from a period of benign markets. A calm fortnight tells you little about behaviour on a volatile day, and the DTCC figures should be read as a successful transition, not a proof of resilience.

Practical Recommendations

If you run post-trade technology, treat the cycle as a critical-path problem. List every step between execution and the CSD cut-off, assign each an owner and a latency budget, and attack the longest dependent chain first. Improvements elsewhere will not move settlement time.

Automate affirmation and push instructions as early as the data allows. The US experience shows that the segment with the lowest automation determines your fail rate. Allocate and confirm during the trading day rather than after the close wherever the workflow permits.

Model liquidity, not just margin. Reduction in clearing fund requirements is real but is partly offset by earlier funding and FX needs. Forecast intraday and overnight cash by currency and account, and test it against stress scenarios.

Make your data model explain itself. Event-sourced trades, bitemporal state and deterministic netting let you answer a regulator, a counterparty or an auditor without archaeology. Instruction IDs should be deterministic so a replay never duplicates a settlement.

For European and UK operations, anchor your plan to 11 October 2027 but confirm the final legal status with your regulator and CSD, and rehearse in industry testing windows when they are announced.

A short checklist:

Measure affirmation rate by segment and by hour, and alert before the cut-off.
Support partial settlement and pro rata or priority rules explicitly.
Automate securities lending recalls and buy-in or borrow triggers.
Reproduce CSD penalty calculations independently and reconcile them daily.
Prefund or pre-position FX for non-home-currency settlements where needed.
Keep a three-way reconciliation with ageing and ownership of each break.
Replay-test your instruction pipeline for idempotency.

Frequently Asked Questions

What is securities settlement?

Securities settlement is the final stage of a trade in which the seller delivers securities, the buyer pays cash, and the records of ownership are updated at a depository. It follows execution, confirmation and clearing. A well-designed system uses delivery versus payment so that neither side can be left delivering without being paid. Settlement is considered final when the transfer cannot be unwound under the system’s rules and applicable law.

What does T+1 settlement mean?

T+1 means a trade settles one business day after the trade date, instead of the earlier T+2. The US, Canada and Mexico moved on 28 May 2024. The UK Accelerated Settlement Taskforce recommended 11 October 2027 for the UK, and the EU and Switzerland were reported as aligned on the same date. The shorter cycle forces allocation, affirmation, funding and FX to finish much earlier.

What is CCP novation?

Novation is the legal replacement of a bilateral trade with two contracts, one between the CCP and the buyer and one between the CCP and the seller. The CCP then guarantees performance to both. This enables multilateral netting, since all obligations face one counterparty, and it moves credit risk from individual firms to the CCP’s margin and default fund. Some CCPs reach the same result through open offer instead.

What is the difference between a CCP and a CSD?

A CCP manages counterparty risk. It novates trades, nets obligations, collects margin and handles member defaults. A CSD manages ownership records and performs the actual transfer of securities. In the US, NSCC is the CCP and DTC is the CSD, both within DTCC. In Europe the roles are split across CCPs such as LCH and Eurex Clearing and CSDs such as Euroclear and Clearstream.

What are DvP models 1, 2 and 3?

They are the BIS classification of delivery versus payment systems. Model 1 settles securities and cash gross, trade by trade. Model 2 settles securities gross and cash net. Model 3 nets both legs. Model 1 needs the most liquidity but carries the least settlement-cycle dependency. Model 3 economises on liquidity but relies on the netting cycle completing. Most real systems blend features of several.

What are CSDR settlement penalties?

Under the EU CSD Regulation’s settlement discipline regime, participants who cause a settlement fail pay a daily cash penalty, calculated by the CSD and paid to the affected receiving party. The penalty regime went live on 1 February 2022. Mandatory buy-ins were postponed and, after the CSDR Refit, retained only as a last-resort tool. Check current rates and technical standards with ESMA, as they have been under review.