Skip to content
Corporate & Investment Banking My progress Français

Home / Capital Markets & Securities Services / Securities Services / Module 22

22Corporate Actions Processing

How a bank tracks, calculates and pays out every event on the securities it holds — a data-quality and deadline-driven discipline where getting a date or a ratio wrong costs millions

Dividends, splits and mergers, paid to the right holder on the right date.

Reading time
2 h 00
Sections
80
Recall questions
35
Quiz questions
16
Simulators
1

Two threads run through this module

The business thread follows the money and the choice. Every corporate action reduces to one sentence: a holder's position on the record date, multiplied by the issuer's terms, produces an entitlement that moves on the dates the market sets. Master that sentence and every event in the module — a fifty-cent dividend or a billion-euro merger — is a variation of it. The business question is always the same: who is entitled, to how much, when, and did they get to choose.

The technology thread follows the data and the deadline. Four of the five lifecycle stages are, at heart, data operations: ingesting announcements, validating them, calculating against positions, and moving values. The hard part is never the arithmetic — the formula is trivial. The hard part is doing it on the exact record-date position, with the right terms, correct rounding, correct currency, correct tax, and hitting an immovable deadline, every time, across millions of events. Corporate actions is a data-quality and workflow problem first, and an operations problem second.

The two threads are inseparable. A missed election is both a client who lost money and a workflow that failed to collect an instruction before a cut-off. This module keeps both in view at every step.

How to read this module

The module runs in five parts. Part 1 builds the concept from zero — what an event is, the classification that governs everything, and where the function sits in the bank. Part 2 walks the lifecycle end to end, announcement to payment, with the deep mechanics that cause real losses. Part 3 maps the actors and the internal teams. Part 4 opens up the systems, data and messaging. Part 5 covers the risks, three real failure cases, and the transformation agenda including generative AI. A cheat sheet and a glossary close it out.

Throughout, coloured boxes carry specific kinds of content. A Key insight box crystallises an idea worth remembering. A Going deeper box adds mechanism for the technically minded. An IT & data lens box names the systems and data structures behind a concept. A Watch out box flags a common error. A Worked example box follows our running case. An In a nutshell box recaps a part. A Check your understanding box lets you test yourself. Where a box is titled Europe & France in focus, it marks a regional specificity — this discipline is shaped as much by European market standards as by anything else.

A single fictional company, ACME Industries SA, runs through the module as a worked example. Its figures are illustrative only, chosen to make the mechanics concrete.

You will understand what a corporate action is, from the ground up

By the end you will be able to define any event as position times terms on a set of dates, classify it as mandatory, voluntary or optional, and explain why that classification decides everything that follows.

You will be able to walk the lifecycle end to end

You will trace an event through announce, capture, entitle, elect and pay, and you will know at each stage what is produced, who leads, and where the risk concentrates.

You will understand where the losses actually come from

You will be able to explain why the election deadline is the most dangerous moment in the process, why irreversibility makes corporate actions a fat-tail operational risk, and how controls are built to answer that.

You will be able to trace the data front to back

You will follow one event object from a raw announcement, through the golden record and the entitlement engine, to the payment, the confirmation and the regulatory report — and see why an error upstream reaches every holder downstream.

You will recognise the systems, standards and transformation landscape

You will know the messaging standards that carry events, the vendor stack that processes them, the European regulations that shape them, and where automation and AI genuinely help — and where a human must stay in the loop.

1What a Corporate Action Is, and Why It Matters

Before any processing, one idea has to land: a corporate action is an issuer's decision that reshapes a security, expressed as terms, dates and sometimes a choice. This part builds that idea from zero, places the function inside the bank, lays out every event type by complexity, and introduces the single classification that governs everything downstream.

A corporate action is an issuer decision that changes what investors hold

Start with the holder, not the event. An investor owns a security — say, one thousand shares of a company. That holding sits quietly until the company that issued the shares does something to them: pays a dividend, splits the stock, offers new shares, gets acquired. Each of those decisions changes the cash or the securities the investor ends up with. That decision, and everything the market must do to give effect to it, is a corporate action. The investor did nothing; the issuer acted, and the holding changed.

The event applies to everyone holding the security, identified by its ISIN. A corporate action is not a bilateral trade between two counterparties. It is a one-to-many event: the issuer announces terms, and those terms apply identically to every holder of that International Securities Identification Number, whether they hold one share or ten million. The processing problem is therefore a problem of scale and consistency — the same terms, applied correctly to every position, every time.

Every event, however exotic, reduces to one equation. Take the holder's position on the record date, apply the issuer's terms, and move the resulting cash or securities on the dates the market sets. Position times terms equals entitlement. A cash dividend is position times a per-share rate. A stock split is position times a share ratio. A rights issue is position times an entitlement ratio, then a choice about whether to subscribe. A merger is position times an exchange ratio into a new security. The arithmetic is trivial. The discipline lies in doing it on the exactly correct position, with the exactly correct terms, on the exactly correct dates.

The event begins as prose, and that is the root of the difficulty

Announcements arrive as legal text, not clean data. An issuer does not publish a tidy record with fields for ratio, currency and record date. It publishes a resolution, a press release, a prospectus, an offer document — dense legal language, often in the issuer's home tongue, sometimes ambiguous, occasionally with terms that are conditional on things that have not yet happened. Someone, or something, has to read that text and turn it into a structured event a system can process. That translation, done at scale and under time pressure, is where errors are born.

The terms are rarely as simple as they look. A one-for-five rights issue sounds trivial until a holder owns 12,343 shares and the division leaves a fraction. Fractions must be rounded on a rule that varies by event and market, and the fractional entitlement paid in cash-in-lieu. Currencies must be converted. Conditionality — a tender that only proceeds if a minimum is reached, a step-up in terms if a threshold is crossed — means the outcome is not known when the event is announced. The apparently simple ratio is where much of the real work hides.

The anatomy of any event, and what a real notification looks like

Any corporate action, however exotic, is built from the same four components. First, the issuer and security, identified by ISIN — the event applies to everyone holding that security. Second, the event type — dividend, split, rights, merger, tender — each with its own processing logic. Third, the terms, the ratio at the heart of the event: how much cash or stock per share held, the maths of the whole thing. Fourth, the dates — ex, record, deadline and pay — that decide who gets what, and when. Learn to read those four components off any announcement and you can process any event.

In practice a notification is exactly this: a set of coded fields. When the ACME rights issue reaches the bank as a structured message, it is not prose but a compact record. The event type and indicator tell the system how to process; the four dates set the timetable; the terms and options define the maths and the choice; and the default tells the system what to do with a silent holder.

Field Value Meaning
CAEV RHTS Event type: rights issue
CAMV VOLU Voluntary — election required
ISIN FR0000000001 The security the event applies to
EXDT 2026-03-04 Ex-date
RDDT 2026-03-05 Record date
MKDT 2026-03-18 Market election deadline
PAYD 2026-03-25 Pay date
TERMS 1 new per 5 held The entitlement ratio
PRIC EUR 42.00 Subscription price
OPTN 01 EXER / 02 LAPS / 03 SELL Exercise, lapse, or sell the rights
DFLT 02 LAPS Default if no instruction: rights lapse

Read that record and the whole event is legible. RHTS and VOLU tell the platform to fork into the voluntary-event workflow; the four dates schedule every downstream action; the ratio and price drive the entitlement calculation; and the default — here, that the rights lapse if the client is silent — is the single most important field for risk, because it decides what happens to the majority of holders who never respond.

Corporate actions is a servicing function, sitting on top of settled positions

It runs on holdings you already have, not on new trades. This is the cleanest way to place the function. Trading and settlement are about acquiring or disposing of a position. Corporate actions is about servicing a position for the whole life of the holding. Once a trade settles and the client holds the security, the bank — as custodian or as holder — becomes responsible for everything the issuer subsequently does to it, potentially for years. Corporate actions is the asset-servicing heartbeat of custody.

It sits inside post-trade, in the asset-servicing or securities-services division. In front-to-back terms, the chain runs: a trade executes, then settles; the holding sits in custody as the record of who owns what; every issuer event on that holding is captured and serviced by corporate actions; entitlements pay out and positions adjust; and everything reconciles back to the custodian and the street. Corporate actions is the servicing layer that runs continuously across the middle of that chain. It is not a front-office revenue engine; it is a client-servicing and operational-risk function, and it is judged on getting things right rather than on generating trades.

A wide set of internal stakeholders depend on it silently. Clients expect their dividends and their choices, correctly and on time. Trading desks need positions adjusted the instant an event takes effect, so that risk and profit-and-loss stay accurate. Finance and risk rely on holdings and cash being right after every event. And custody clients — the asset managers and institutions who pay the bank to hold and service their securities — judge the bank substantially on how cleanly events are handled. Servicing quality wins and loses custody mandates.

The systems that matter here are the securities master (reference data for every instrument the bank touches), the position-keeping or custody ledger (who holds what, where, and how much is lent out), and the corporate actions platform that sits on top of both. Corporate actions inherits the data quality of everything beneath it: if the position ledger is wrong, the entitlement is wrong, no matter how good the event data is. The function is only ever as reliable as the holdings data it stands on.

Where corporate actions sits relative to its neighbours

Corporate actions is one function in a family of post-trade services, and its distinctiveness is clearest against its neighbours. Settlement finishes trades — it turns an agreement to trade into a delivered position, and then it is done. Custody safekeeps the position and keeps the record of who owns what. Reconciliation continuously checks that the bank's records match the market's. Corporate actions is the one function among these that is driven by the issuer rather than by the client or the trade: it reacts to what the company does to its own securities, for the whole life of the holding, over and over. Settlement happens once per trade; a corporate action can happen many times on the same holding, each time reshaping it. That is why corporate actions is described as a servicing heartbeat — it is the recurring, issuer-driven pulse that runs on top of the static holding the other functions maintain.

How servicing quality turns into money, even though the function sells nothing

Corporate actions earns no fee of its own, yet it is central to a real revenue business, and that paradox is worth resolving early. The bank does not invoice a client for processing a dividend. Instead, corporate actions is one component of custody and asset servicing, a business the bank does charge for — typically a few basis points a year on assets under custody, sometimes with additional transaction, income or event fees. A client paying, say, one to three basis points on a large portfolio is paying, among other things, for every dividend, split and merger on that portfolio to be serviced correctly and on time. Corporate actions is bundled into that price. It is a cost centre that sits inside a revenue business, and its quality is part of what the client is buying.

The commercial logic is that of a hygiene factor with a fat downside. A useful analogy: corporate actions is the plumbing of custody. Nobody pays a premium for plumbing that works, and nobody notices it until it leaks — but a flood costs a fortune. Custody is won on price, reach and relationship, and it is lost on servicing failures. A bank cannot charge much more for excellent corporate-action processing, because clients expect it to be correct; but it can lose a mandate, and the fee stream attached to it, after a run of visible errors. The economics are therefore asymmetric in exactly the way the risk is: limited upside from doing it well, large downside from doing it badly.

That asymmetry is why the servicing contract matters as much as the servicing. Custody agreements specify service levels and, crucially, where liability sits when an event is mishandled — the bank typically indemnifies the client for losses caused by its own processing error. So a missed election is not merely an unhappy client; it is a contractual obligation to make the client whole from the bank's own funds, followed by a real risk of losing the relationship. The function that sells nothing can nonetheless write a cheque, and lose an account, on a single bad event.

The events, sorted by how much difficulty they carry

A handful of event types account for the vast majority of volume, and a different handful account for the vast majority of risk. Cash dividends are enormously frequent and almost entirely routine. Mergers and complex reorganisations are rare and consume most of the analysts' effort and cause most of the material losses. Understanding the landscape means seeing both axes at once — frequency and difficulty — because they pull in opposite directions and that tension defines where the function invests.

Event type Class Frequency Difficulty What makes it hard
Cash dividend Mandatory Very high Low Withholding tax; currency; scale
Stock split / bonus issue Mandatory Medium Low Ratio and fractions; price adjustment
Scrip dividend / DRIP Optional (CHOS) Medium Medium Cash-or-shares choice; defaults; pricing
Rights issue Voluntary Medium Medium–high Election, deadline, tradable rights, fractions
Tender offer / buyback Voluntary Low–medium High Election, pro-ration, conditionality
Merger / acquisition Mandatory or elective Low High New ISIN, exchange ratio, cash/stock mix
Spin-off / demerger Mandatory Low High New line created, cost-basis split
Scheme of arrangement Mandatory (court) Low Very high Conditional, shifting effective dates

Read the table as a map of where effort goes. The top rows are high-volume, low-difficulty events that should flow straight through with no human touch. The bottom rows are low-volume, high-difficulty events that resist automation because their terms are bespoke, their outcome is uncertain, and their data arrives as prose. The whole economic logic of the function is to push the routine through automatically and concentrate scarce human expertise on the complex tail. Part 5 returns to this as the eighty-twenty of corporate-action risk.

The events in words, from the routine to the fearsome

Each event type has its own logic, and knowing them by feel is half the discipline. A cash dividend pays cash per share — mandatory, and by far the most common event. A stock split gives more shares at a lower price, and a bonus issue gives free extra shares; both are mandatory and change the share count without changing total value. A scrip dividend or DRIP lets the holder take the dividend as cash or as new shares — mandatory with options, with a default for the silent. A rights issue grants the right to buy new shares at a discount, pro-rata, and is often voluntary with tradable rights. A tender offer or buyback is the issuer offering to buy shares back at a price — voluntary, and the holder must elect, often with pro-ration if oversubscribed. A merger or acquisition exchanges shares for cash or the acquirer's stock, usually mandatory, replacing the position with a new security. A spin-off hands holders shares in a newly separated entity, pro-rata, creating a new line in the account. And a scheme of arrangement is a court-sanctioned reorganisation whose effective date can move until the last moment. The first few are routine; the last few are where the effort and the losses concentrate.

The reference case: reorganisations are where the difficulty really lives

Dividends teach the mechanics; reorganisations teach the risk. For most of this module the running worked example is a rights issue — complex enough to show elections, deadlines, fractions and a real choice, simple enough to follow end to end. But the events that keep managers awake are the reorganisations: mergers that replace one security with cash and stock in another, spin-offs that create a brand-new line in every holder's account, schemes of arrangement whose effective date can move until the last moment. These are near one-offs. There is not enough repetition to automate them reliably, the cost of getting them wrong is high, and they arrive as dense legal documents. When this module refers to the hard end of the spectrum, it means these.

The numbers are large enough that small error rates produce large losses

The scale is the reason a routine function is a serious risk. A single global custodian processes on the order of 1.2 million corporate-action events a year, and the money that moves through dividends and reorganisations runs into the trillions. Even a tiny error rate, applied to that volume and those values, produces losses that matter. And the loss distribution is not smooth: most events are trivial, but a single mishandled voluntary event or a botched merger can dwarf a full year of ordinary processing cost. That asymmetry — routine most of the time, catastrophic occasionally — is the defining shape of the risk.

The industry numbers make the case that this is expensive, manual and under-automated. Industry estimates put the annual cost of corporate-action processing to the industry at roughly US$58 billion, growing by around ten percent a year, and suggest that close to half of all corporate-action data is still handled manually, with straight-through processing rates for the function below forty percent. The software market built to fix this — corporate-actions automation platforms — is itself estimated at around US$4 billion in 2024, projected to reach the low-teens of billions within a decade. For a single mid-tier custodian, the operational-risk exposure attached to the function has been estimated at around US$12 million a year. These figures come from vendor and market-research sources and vary widely, so treat them as orders of magnitude rather than precise measures — but the direction is unambiguous: a large, costly, still-manual function that the industry is spending heavily to automate.

Dimension Industry order of magnitude
Events per global custodian, per year ~1.2 million
Data handled manually ~46%
Straight-through processing rate below ~40%
Estimated annual industry cost ~US$58 billion, growing ~10%/yr
CA-automation software market (2024) ~US$4 billion
Op-risk exposure, mid-tier custodian ~US$12 million/yr
Realistic automation target ~80% supervised, ~20% manual

Figures are industry estimates from vendor and market-research sources; they vary by definition and should be read as orders of magnitude, not precise measures.

The whole subject in one sentence, made numeric

Position times terms equals entitlement is not a slogan; it is a calculation you can watch happen. Take the ACME client with 12,340 shares and the one-for-five rights issue. The position is 12,340. The terms are one new share per five held at €42.00. Apply them: 12,340 divided by 5 is 2,468 rights, so the entitlement is up to 2,468 new shares for €103,656.00, moved on the pay date if the client exercises before the deadline. Every element of the discipline is visible in that one line — the exact record-date position, the issuer's terms, the choice, the deadline, the money, the date. Master the sentence and every event on the pages that follow, however exotic, is a variation of it: a different ratio, a different option, a different date, but always position times terms on a timetable the market sets. All figures illustrative only.

One classification governs everything: mandatory, voluntary, or optional

The most important thing to know about any event is whether someone has to choose. Corporate actions divide into three classes, and that division decides whether the bank simply applies terms or must run a whole election process. It is encoded in the ISO messaging standards as an event indicator precisely because the processing must fork on it.

Mandatory events happen to everyone automatically

A mandatory event applies to all holders with no choice at all. A cash dividend, a stock split, a bonus issue, a merger imposed on shareholders — these simply happen. The bank's job is to apply the terms correctly and not make a mistake. There is no notification-and-collection cycle, no deadline for the holder, no default to manage. Mandatory events are an operational task: get the position right, get the terms right, move the money. The risk is real but bounded — it is the risk of a calculation or a date being wrong, not the risk of a missed instruction.

Voluntary events require the holder to elect

A voluntary event does nothing unless the holder actively participates. A tender offer, a buyback, a rights subscription — the holder must submit an instruction to take part, or do nothing and stay out. This adds a whole dimension of risk the mandatory events do not have: the bank must notify every entitled holder, collect their instructions, validate them, aggregate them up the custody chain, and hit an immovable market deadline. Miss the deadline and the entitlement or the choice is simply gone. Most catastrophic corporate-action losses are voluntary events where a client's instruction was missed, mis-keyed, or submitted after the cut-off.

Optional events happen, but the holder chooses how

A mandatory-with-options event will occur regardless, but the holder picks the form. The classic case is a scrip dividend: the event pays out, but the holder may take cash or new shares. A default applies to anyone who stays silent. The event is certain; only its form is chosen. The risk sits between the other two classes — the bank must still notify and collect, but a silent holder is not left with nothing, they are given the default.

Class ISO indicator Does the holder choose? Where the risk sits
Mandatory MAND No — applied automatically Calculation and date accuracy
Voluntary VOLU Yes — must elect to take part Notification, deadline, missed instruction
Mandatory with options CHOS Yes — chooses the form, default applies Correct default, consistent application

The business has its own rhythm, even though it is not cyclical like issuance

Corporate actions does not boom and bust with markets the way deal origination does, but it is far from steady. Equity issuance dries up in a downturn; corporate actions does not, because dividends, mergers and restructurings happen in every kind of market — arguably more restructurings happen in bad markets. What the function does have is a set of structural drivers that shape its workload and its investment priorities, and understanding them explains why transformation money flows when it does.

Dividend seasons concentrate the volume. Payments cluster around reporting calendars, so large fractions of the annual event count fall in a few weeks. The function must be built for the peak, not the average — capacity that sits idle much of the year but cannot fail in dividend season.

Regulatory waves reset the baseline. The move to shorter settlement cycles, the migration of messaging to a new standard, the arrival of shareholder-rights legislation — each of these is a step change imposed from outside that forces re-platforming and re-training across the whole industry at once. These waves, not market cycles, are what drive the big investment programmes.

The settlement cycle itself is tightening. As markets move from longer settlement to T+1, the window between the key dates compresses, and the time available to detect and fix an error before it becomes a loss shrinks with it. A shorter cycle raises the operational bar for the whole function.

Four reasons an unglamorous function sits near the top of the risk register

Corporate actions is easy to overlook and expensive to get wrong, and four features explain why. The function has none of the glamour of trading or origination — it produces no revenue and appears, from a distance, to be administrative. Look closer and it is one of the largest sources of operational and financial risk in the bank, for four connected reasons.

The numbers are huge. Across the market there are millions of events a year, moving very large sums in dividends and reorganisations, and a single missed election or mis-calculated entitlement can cost millions on one event. The sheer scale means that even a small error rate produces material absolute losses.

The errors are often irreversible. Once cash pays out to the wrong party or an election deadline passes, undoing it is hard or impossible. Unlike a mis-booked trade that can be cancelled and rebooked, a wrong corporate-action outcome frequently cannot be reversed, so the loss lands straight on the bank's own account.

It runs on bad data. Announcements arrive as unstructured text from many sources that do not always agree. Turning that raw material into one trusted version of the event — the golden record — is the core problem of the discipline, and it is a data-quality problem before it is anything else.

The deadlines are unforgiving. Record dates and election cut-offs are fixed by the market, not by the bank. Miss one and the entitlement or the choice is simply gone. There is no negotiating an extension, and no second attempt.

Put those four together and the risk shape is clear. Most events are tiny and routine, which tempts under-investment, but the loss distribution has a fat tail: a single mishandled voluntary event or merger can dwarf a full year of processing cost. That asymmetry is the whole argument for taking an administrative-looking function seriously — the average event is trivial, but the worst event is not, and it is the worst event that sets the requirement.

  1. State the one equation every corporate action reduces to. — Position × Terms = Entitlement, moved on the dates the market sets.

  2. Why is a corporate action described as a one-to-many event?

  3. Why is corporate actions called a servicing function rather than a trading function?

  4. What makes a voluntary event riskier than a mandatory one?

  5. What is the danger of mis-capturing an event's class as MAND when it is VOLU?

  6. Why does the function have to be built for the peak rather than the average?

  7. Name one European force that shapes corporate-action processing and say what it does.

1

  1. What equation does every corporate action reduce to?

  2. Why is a corporate action a one-to-many event?

  3. Why is corporate actions a servicing function?

2The Lifecycle, End to End

Every event, from a fifty-cent dividend to a billion-euro merger, flows through the same five stages: announce, capture, entitle, elect, pay. This part walks that spine in order, stops at each mechanic that causes real losses, and carries the ACME rights issue through from the issuer's declaration to the client's new shares.

The five stages are the same for every event

The lifecycle is a spine, not a menu. A cash dividend and a multi-way merger differ enormously in difficulty, but they travel the identical five stages. The issuer announces; the bank captures and validates the event into one trusted record; on the record date the bank snapshots positions and calculates entitlements; for voluntary events it notifies holders, collects instructions and manages the deadline; and finally it moves cash and securities, books them, reconciles and reports. Learning the spine once means every event becomes a variation on a known theme.

Stage Typical duration Output Who leads
1. Announce Event start Public terms and dates Issuer and its agent
2. Capture Days One validated golden record CA analyst / data management
3. Entitle On record date Per-holder entitlement CA operations / entitlement engine
4. Elect Until the deadline Collected, aggregated instructions Client services / CA operations
5. Pay On pay date Cash and stock moved, reconciled CA operations / settlement
Figure 1 The corporate-action lifecycle, with risk peaking at the immovable election deadline

The effort is not spread evenly — it spikes at the deadline

Most of the danger concentrates in the middle of the lifecycle. Announcement and capture are demanding but forgiving: an error caught here can be corrected before anything moves. Payment is high-value but largely mechanical if the earlier stages were right. The peak of risk sits at the election deadline, because it is immovable and because a missed or wrong instruction there converts directly into a loss with no way back. The intensity curve on the timeline diagram is not decoration — it is a map of where to put controls, senior attention and system investment. You harden the middle.

Stage 1 — The issuer announces, and the clock starts

An announcement is the starting gun, and it is rarely clean. The issuer, through its paying or issuer agent, declares the event and its terms. That information reaches the market through several channels at once — the agent, the central securities depository, data vendors, sub-custodians in each local market — and they do not always agree. From the moment of announcement the bank is on the clock: it must capture the event, validate it, and be ready to act on every date the issuer has set. Nothing downstream can be better than what is captured here.

The bank does not control the timetable. The issuer sets the ex-date, record date, election deadline and pay date, and the market conventions fix how they relate. The bank's entire job at this stage is to receive the announcement faithfully and lose no time — because every day spent capturing is a day removed from the already-tight window to notify holders and collect their choices.

The four sources, and why they disagree

Understanding capture means knowing exactly who announces an event and why their versions differ. Four kinds of source carry the same event, each with a characteristic strength and a characteristic weakness. The issuer or its agent is the primary, authoritative source — but it often arrives as prose, in the issuer's own language, on the issuer's own timetable. The central securities depository — Euroclear, Clearstream, DTC — is the official chain and highly structured, but it is one step removed from the issuer. Data vendors — Bloomberg, LSEG, SIX — provide clean, machine-readable feeds, but they are derived products that can introduce their own transcription and rounding errors. And the sub-custodian in each local market adds local knowledge but another hop and another chance to diverge.

Disagreement is structural, not accidental. A vendor may round a ratio the agent states exactly; a depository may carry a date the issuer later amends; a currency or decimal may be mis-transcribed on one feed and not another; an option present in the offer document may be missing from a structured feed. None of these sources is wrong often, but across millions of events even a low disagreement rate produces a steady stream of breaks — and acting on any single one without comparison is precisely how the ten-times-rate error reaches a payment. This is the concrete reason dual-source validation is the first control in the framework: the sources exist to be checked against one another, because each is fallible in a different way.

Stage 2 — Capture builds the golden record

The first real work is turning many imperfect messages into one truth. The same event arrives from the issuer's agent, the central securities depository, one or more data vendors, and the local sub-custodian. Each is a source; none is guaranteed correct; and they frequently disagree on a date, a ratio, a currency or an option. Acting on any single unverified source is how losses start. The output of capture is a single validated, versioned, time-stamped event record — the golden record — that every downstream calculation will trust and nothing else.

Figure 2 Multiple imperfect sources are reconciled into one validated golden record

Why one source is never enough

Discrepancies between sources are common, not exceptional. A vendor feed may carry a rounded ratio where the agent carries the exact one. One source may show a dividend rate ten times another because of a decimal or currency error. An option may be missing from one feed. The primary source — the issuer or its agent — is authoritative but often arrives as prose, while the structured feeds are convenient but derived and therefore capable of introducing errors of their own. The discipline is to compare, never to trust blindly.

The scrubbing process, step by step

Scrubbing is the industrial term for manufacturing one truth from many inputs. It runs in four steps. First, ingest and normalise: take every source and map it into one common event structure so the fields line up. Second, compare and flag: match the sources field by field automatically, and raise a flag wherever they disagree on a date, a term or an option. Third, investigate and resolve: an analyst reviews only the flagged breaks, goes back to the primary source, and decides the correct value. Fourth, publish the golden record: one validated, complete, versioned event that becomes the single version of truth. Events where every source already agrees flow straight through untouched — the scarce human effort goes only to the genuine breaks.

This is an exception-based operating model, and it is the single most important design idea in the whole function. Automated matching handles the clean majority; an exception queue surfaces only the disagreements to an analyst. The golden record is versioned and time-stamped, so every later calculation can point to exactly which version of the event it used, and any correction leaves an auditable trail. Build the platform around this pattern and volume scales without headcount; build it any other way and every event competes for the same analyst's attention.

The data model beneath the golden record

A good event model lets one piece of code handle a scrip dividend and a multi-way merger. Analysts and developers structure every event around the same entities. An EVENT header carries the type, the mandatory/voluntary/optional indicator, the ISIN, the status and the golden-record version. A set of DATES carries ex, record, election deadline and pay, plus the bank's own internal cut-offs, each of which drives a scheduled action. One or more OPTIONs represent the choices on a voluntary or optional event — one row per choice, each with its own terms and a default flag. TERMS hold the maths of each option: the cash rate, the share ratio, the price, the currency, the rounding rule. An ENTITLEMENT is a holder's record-date position combined with the chosen option's terms. And an INSTRUCTION is the client's election against an option, with quantity, timestamp, validation status and audit trail.

The relationships are what matter. An event has many options; each option has terms; a holder's position plus a chosen option yields an entitlement; an instruction selects an option. Get that one-to-many structure right once and the system absorbs new event types as data rather than as code. Get it wrong and every complex event becomes a bespoke special case that has to be hand-built and hand-checked.

Events change after they are announced, and the record must keep up

An announcement is a starting position, not a final one, and handling amendments cleanly is part of capture. Issuers revise events: a date shifts, a ratio is corrected, an option is added, a conditional term resolves. Each revision arrives as a new message — an MT564 replacement or an updated seev.031 — and the golden record must be re-versioned to reflect it without losing the history of what was believed before. This is why the golden record is versioned and time-stamped rather than simply overwritten: downstream calculations must be able to point to the exact version they used, and a late amendment must ripple through to entitlements and notifications that may already have gone out. An amendment caught and propagated is routine; an amendment missed, so that the bank keeps processing a superseded term, is a break waiting to surface at pay date.

Stage 3 — Entitlement is position times terms, on the record-date snapshot

On the record date the system freezes each account's position and applies the terms. This is the calculation at the centre of every event. The formula is trivial — multiply the record-date position by the terms — but the difficulty is in the word "position." The settled position is not the whole story, and getting the snapshot wrong entitles the wrong holders entirely.

What "who really holds this" actually means

Three complications turn a simple snapshot into one of the hardest questions in post-trade. First, pending trades: a buy that has not yet settled by the record date means the buyer is not yet the holder of record, even though they will be — this triggers a market claim, covered below. Second, securities lending: stock that has been lent out sits with the borrower on the record date, so the lender needs the income "manufactured" back to them. Third, short positions: a short seller owes the entitlement to the party they sold to. If the system snapshots only the naive settled position and ignores these, it pays the wrong parties and misses the claims.

The arithmetic, worked

Every event is this calculation, scaled to millions of positions. Consider a one-for-five rights issue at €42.00 and a record-date position of 12,340 shares. Dividing 12,340 by five gives 2,468 rights exactly, so the holder is entitled to 2,468 new shares with no fraction. If they exercise, they pay 2,468 times €42.00, which is €103,656.00, and receive 2,468 new shares. If the rights lapse — the default on this event — they receive nothing and forgo the value of the discount, a real loss. Now change the position to 12,343 shares: dividing by five gives 2,468.6, so the holder is entitled to 2,468 whole new shares and a fraction of 0.6. The fraction is typically rounded down and paid in cash-in-lieu on a rule that varies by event and market. Doing that rounding consistently and correctly across thousands of holders is a genuine systems requirement, not an afterthought.

The key dates are the timetable that decides everything

Get one date wrong and the right holder is paid the wrong amount, or on the wrong day. The dates are not administrative trivia; they are the machinery that decides who is entitled. There are five that matter.

Figure 3 The five key dates — and the ex/record relationship that creates market claims

The announcement date starts the clock — the moment the issuer declares the event and the bank must begin capturing and acting. The ex-date is the point from which the security trades without the entitlement: buy on or after ex-date and you do not get the event. The record date is the snapshot — whoever is the holder of record on this date is entitled. The election or market deadline is, for voluntary events, the last moment to submit an instruction; it is immovable, and after it only the default applies. The pay date is when cash or securities actually move to the entitled holders, and the event completes.

The ex/record relationship is the root of market claims

Ex-date and record date are linked by the settlement cycle, and that link is where claims come from. The ex-date is set so that anyone buying from it settles after the record date and is therefore not the holder of record. But a trade agreed before the ex-date may not settle until after the record date — which means the seller, still the holder of record, receives an entitlement that economically belongs to the buyer. The mechanism that moves that entitlement from the party who received it to the party who deserves it is a market claim, covered in detail below. Getting the ex/record alignment wrong is one of the most expensive mistakes in the discipline.

Stage 4 — Election is the most dangerous moment in the whole process

For voluntary and optional events the bank must inform every holder, collect a choice, and act on it before a deadline it cannot move. This is where catastrophic losses cluster. The election chain has five links, and each is a place to fail.

First, notify every entitled holder that the event exists, what the options are, and when the deadline falls. Second, collect each client's instruction — cash or shares, exercise or lapse, and how many. Third, validate that each instruction is consistent with the holding and the available options. Fourth, aggregate and submit the instructions up the custody chain to the agent or depository before the market deadline. Fifth, apply the default to any holder who stays silent — and record that they were properly informed.

The deadlines are nested, and the bank's own is the earliest

The bank must set its own cut-off well before the market's, because it loses time at every hop. The client response deadline comes first, to leave time to process. Then the bank's internal cut-off, to validate, aggregate and resolve exceptions. Then the custodian or agent's deadline further up the chain. And finally the immovable market deadline. Each hop consumes time, so the bank's client-facing cut-off is often days before the market's — a fact clients rarely appreciate as they push their instructions right up to the last moment.

Figure 4 Nested deadlines — the bank's client cut-off sits days before the immovable market deadline

Why elections cause losses

Four failure modes recur. Late client instructions arrive after the bank's cut-off but the client still expects them honoured. Manual rekeying of "cash or shares, how many" under time pressure introduces errors exactly when the stakes are highest. The wrong default applied to silent holders becomes a systematic error across many accounts at once. And aggregation errors, when many individual elections are rolled into one message up the chain, corrupt the whole submission. Silence, crucially, is itself a decision: most clients do not respond, so the default must be correct, applied consistently, and provably communicated.

Banks control election risk with technology aimed squarely at the two worst failure modes — rekeying and missed deadlines. Digital election capture through a client portal or an API removes manual rekeying by taking the instruction in structured form and validating it against the holding automatically. Automated notifications with reminders and escalation and chasing of non-responders shrink the pool of silent holders before the cut-off. And a full audit trail proves who was told what, and when — the evidence the bank needs if a client later claims they were not informed or were mishandled. Every one of these is a direct answer to a specific, historically expensive mistake.

Stage 5 — Payment moves the money, books it, and proves it ties out

The final stage moves cash and stock, records it correctly, and proves it all reconciles. It runs in six steps. The system recomputes each entitlement from the golden record and the confirmed positions, applying rounding and cash-in-lieu rules consistently. It books the movements, crediting and debiting cash and securities to each account and adjusting the underlying position. It withholds tax at the correct rate for each holder's jurisdiction and status before paying net proceeds. It reconciles to the custodian, matching what the bank booked against what the depository or agent actually paid. It confirms to clients with statements and confirmation messages. And it reports and closes, feeding finance, risk and regulatory reporting and archiving the full audit trail.

The reconciliation gate is the last line of defence

The bank must never pay clients more than it receives from the market. Reconciling the entitlement it booked against the cash and stock the custodian actually delivered is the final control before money leaves. A break caught here, before pay date, prevents a loss. The same break caught after pay date is the loss. This is why the reconciliation gate is non-negotiable: no client payment is released on an unreconciled break.

The difficulty lives in the complex events

Dividends are routine; reorganisations consume the effort and cause the big losses. A small fraction of events — mergers, spin-offs, schemes, conditional tenders — absorb most of the analysts' time and account for almost all the material losses, because their terms are bespoke, their outcome is uncertain, and their data arrives as prose. This is the eighty-twenty of corporate-action risk.

Mergers and acquisitions exchange shares for cash, stock, or a mix, sometimes with an election and fractional handling; a new security with a new ISIN replaces the old position. Spin-offs and demergers give holders shares in a newly separated entity, pro-rata, creating a brand-new line in the account and requiring the cost basis to be split. Schemes of arrangement — court-sanctioned reorganisations common in the UK — carry conditions and effective dates that can shift, leaving high uncertainty until they become effective. Reverse splits and consolidations turn many shares into few, with fractions and odd-lots that must be handled and often cashed out, and small errors that scale across every holder. Multi-stage and conditional events — tenders with minimum-acceptance conditions, step-up terms, or pro-ration if oversubscribed — have outcomes that are not known in advance. And distributions in specie pay out in assets rather than cash, needing valuation, custody set-up and their own tax treatment.

Every income event is also a tax event

Withholding tax makes dividends far more than a simple cash movement. The source country taxes dividends and income before payment, so the bank must apply the right rate for each holder. Double-tax treaties reduce that rate for eligible holders — but only if the right documentation is in place. The bank can either apply the lower rate up front (relief at source) or pay the full rate and reclaim later, a long and manual process. And the tax character of the event itself — income, return of capital, or a taxable gain — changes the client's outcome. Getting withholding wrong cuts both ways: over-withhold and the client is short-paid; under-withhold and the bank may be liable to the tax authority.

Market claims redistribute entitlements to the party who deserves them

When a trade straddles the record date, the entitlement lands on the wrong party and must be moved. The entitlement is paid to whoever is the holder of record, but a trade agreed before ex-date may not settle until after record date, so the seller receives a dividend that economically belongs to the buyer. A market claim moves that entitlement from the party who received it to the party who deserves it. A reverse claim runs the other way, where the buyer was wrongly paid and owes the seller. And a transformation re-shapes a pending trade for some events — for example into the merged security or adjusted terms — rather than simply claiming the cash. Detecting, valuing and settling claims across many pending trades is complex, and a missed or mis-valued claim is a direct loss.

Cross-border processing multiplies every difficulty

The same event type behaves differently in every market, and most large holders are cross-border. Different markets use different ex/record conventions, deadlines, default options and rounding rules. Custody chains lengthen — a global custodian relies on a sub-custodian relying on a local depository, each adding a hop, a deadline and a reconciliation. Tax regimes differ in every source market. And local-language announcements and non-standard data raise the risk at capture. Harmonisation helps — CAJWG and SMPG standards and ISO 20022 make European events more consistent, and T2S centralises much of the settlement — but company law, tax and meeting rules stay national, so full uniformity is not achievable. Non-European markets differ more still.

Lending and shorts turn one entitlement into several payments

When stock is lent or sold short over a record date, the entitlement has to be recreated for the party who is economically owed it — and that is manufactured income. Suppose our client owns 12,340 ACME shares but has lent 2,000 of them to a borrower over the record date. On the record date the borrower is the holder of record for those 2,000 shares and receives the dividend; the lender, our client, is contractually owed the same economic value. The bank must therefore manufacture the payment — recreate the entitlement for the lender out of what the borrower received — so that the client is made whole despite not being the holder of record. The same logic runs through short positions: a short seller who does not own the stock still owes the entitlement to the party they sold to, and the chain must reproduce it.

Manufactured payments are where lending, tax and corporate actions collide. A manufactured dividend is not the same instrument as a real dividend for tax purposes, so its withholding treatment can differ, and getting that wrong is exactly the kind of ambiguity the cum-cum schemes exploited. The processing difficulty is that the bank must track, for every holding, how much is lent or borrowed over each record date, reproduce the income correctly, and apply the right tax treatment to the manufactured leg. It is a clean illustration of why the record-date position is the hardest question in the discipline: the naive holding is not who is economically owed.

The same event, three countries, three behaviours

Cross-border processing is where every difficulty in this part compounds at once, and a single worked case shows why. Imagine ACME pays a dividend and three of the bank's clients hold it through different chains: one directly in France, one through a global custodian into Germany, and one through a sub-custodian in an Asian market. The event is one issuer decision, but the servicing forks three ways. The French holding follows domestic ex/record conventions and French withholding, relieved at source where documentation allows. The German holding travels a longer chain — global custodian to sub-custodian to local depository — each hop adding a deadline and a reconciliation, and the withholding follows German rules with treaty relief depending on the client's residency evidence. The Asian holding may follow a different ex/record convention entirely, arrive as a local-language announcement, and carry a withholding regime with a slow, manual reclaim rather than relief at source. One event; three timetables, three tax outcomes, three reconciliation chains. The bank's platform must absorb all of it as data, because the alternative — a separate manual process per market — does not scale.

The internal gates a big event must pass before it pays

Routine events flow; large or complex ones run a deliberate internal gauntlet before any money moves. The controls in the next section are the standing framework; on top of them, high-value and high-complexity events pass through named decision gates, so that no single person can push a large or unusual event to payment alone. The first gate is event review: for a complex reorganisation, a senior analyst or an event-review forum signs off that the golden record correctly captures bespoke terms and conditions before the event goes live. The second is the entitlement sign-off: for large-value events, the independently recalculated entitlement must be approved before notifications go out. The third is the payment release gate: a large pay run is not released until the reconciliation has tied out and a designated approver has authorised it. And running alongside is an escalation path, so that a break that cannot be resolved, or an approaching deadline that is at risk, is raised to management rather than left on one analyst's desk.

The gates exist because irreversibility removes the usual safety net of correcting mistakes afterward. In a reversible process, a single approver is tolerable because errors can be unwound; here they cannot, so the size and novelty of an event determine how many independent eyes must see it before it moves. A fifty-cent dividend on a widely held stock passes almost invisibly; a billion-euro merger with an election passes through every gate, with senior sign-off at each. Calibrating the gauntlet to the stakes — light for the routine, heavy for the exceptional — is what lets the function process millions of events cheaply while still surrounding the dangerous few with control.

Because errors are irreversible, the process is wrapped in layered controls

Corporate actions is controlled the way it is precisely because its mistakes cannot be undone. Six controls form the standard framework, and none is optional. Dual-source validation means never acting on a single source — at least two independent feeds must agree, or the break is investigated. Four-eyes on the golden record means a second qualified person independently verifies event terms and dates before they go live. Independent entitlement checks recalculate entitlements separately from the primary system and reconcile the totals before any payment. Deadline and default controls raise automated alerts on approaching cut-offs and require explicit sign-off on the default applied to silent holders. The reconciliation gate blocks any client payment on an unreconciled break. And a full audit trail with management information records who did what and when, and tracks errors, breaks and near-misses to drive fixes.

Two reorganisations, worked in detail

The routine events teach the spine; the reorganisations teach where it strains. Two worked cases show how the same five stages behave when the terms are bespoke and the outcome uncertain.

A spin-off, and the cost-basis split

A spin-off creates something from nothing in every holder's account, and the difficulty is accounting, not arithmetic. Suppose ACME separates its industrial division into a new listed company, ACME Systems SA, distributing one ACME Systems share for every four ACME shares held. Our client with 12,340 ACME shares receives 3,085 ACME Systems shares — a brand-new line that did not exist in the account the day before, with a new ISIN, its own reference data, and its own future events to service. The arithmetic is a simple ratio. The real work is that the client's original cost basis must be split between the two lines for tax purposes, on a rule set by the transaction, and that the new instrument must be set up in the securities master before the distribution can be booked. A spin-off mishandled at set-up produces a holding the bank cannot correctly service or report. All figures illustrative only.

A scheme of arrangement, and the moving effective date

A scheme is a court-sanctioned reorganisation whose terms are certain but whose timing is not, and that uncertainty is the risk. Common in the UK, a scheme of arrangement takes effect only when a court sanctions it and conditions are met, so the effective date can move until the last moment. The servicer must hold the event open, keep positions and pending trades aligned to a date that has not yet fixed, and be ready to process the exchange — often into cash, or the acquirer's stock, or a mix — the instant the scheme becomes effective. The classic failure is treating a provisional effective date as final: booking the exchange early, or snapshotting positions on the wrong day, when the court timetable then shifts. Schemes resist automation precisely because their timing is conditional and their terms arrive as legal prose.

  1. Name the five lifecycle stages in order.

  2. What is the golden record, and how is it built?

  3. Why is the record-date position harder than it looks?

  4. A holder of 12,343 shares in a one-for-five rights issue is entitled to what?

  5. Why is the bank's client cut-off earlier than the market deadline?

  6. Why do market claims exist?

  7. What does the reconciliation gate guarantee?

  8. Why must a silent holder's default be provably communicated?

2

  1. Which order is right for the lifecycle?

  2. How is the golden record built?

  3. Why is the record-date position harder than it looks?

3Actors, the Chain, and the Teams Inside the Bank

A corporate action passes down a long chain from the issuer to the end investor, and elections travel back up the same chain. This part maps the four camps, follows the flows in both directions, details the investors who receive and elect, and names the six teams inside the bank that make the function work.

Four camps, and the bank sits in the middle of every flow

The ecosystem has four kinds of party, and the bank touches all of them. At one end is the issuer and its agent — the company that initiates the event and the paying or issuer agent that publishes the terms and pays out. At the other end are the end investors — the asset managers, funds, corporates and retail clients who receive entitlements and make elections. Between them sits market infrastructure — the central securities depositories and the global and sub-custodians that form the official chain along which securities, cash and messages move. And running the servicing itself is the bank, capturing, scrubbing, entitling, collecting elections and paying. The defining feature of the map is that the bank is a party to almost every interaction; it is the servicing hub that connects the issuer's decision to the investor's account.

Figure 5 Event information flows down the custody chain; elections and votes flow back up

Information flows down, instructions flow up

The single most useful mental model is a two-way chain. Event information flows down — issuer to agent to depository to custodian to bank to investor — as each layer receives the announcement and passes it on. Elections and votes flow back up — investor to bank to custodian to depository to agent — as each layer collects instructions and aggregates them. Every hop in either direction adds a deadline and a reconciliation. This is exactly why harmonised standards and the migration to modern messaging matter so much: each additional hop is a place for time to be lost and for data to be corrupted, and standardising the message shrinks both risks.

From To Stage What flows
Issuer / agent (I)CSD, vendors Announce Event terms and dates
(I)CSD / sub-custodian Bank Capture Structured notification (MT564 / seev.031)
Data vendors Bank Capture Independent event feeds for scrubbing
Bank End investor Elect Notification of options and deadline
End investor Bank Elect Election instruction
Bank (I)CSD / agent Elect Aggregated instruction (MT565 / seev.033)
(I)CSD / agent Bank Pay Cash and securities, confirmations
Bank End investor Pay Entitlement, confirmation, statement

The buy-side: who receives, elects, and lends

The investors at the bottom of the chain are not a uniform block — their behaviour shapes the bank's workload. In an issuance module the buy-side matters because of who buys; in corporate actions it matters because of who elects, who lends, and who claims. The same four broad types recur, and each interacts with events differently.

Long-only asset managers

Large, index-aware managers are the bulk of the elections and the bulk of the lending. They hold broadly and for the medium term, so they touch almost every event. On voluntary events they usually have a clear policy and respond, though the largest of them can be slow because instructions route through their own operations. Critically, they are heavy securities lenders, which means their record-date positions are frequently lent out and their income must be manufactured back — a recurring source of entitlement complexity.

Pension and sovereign wealth funds

Patient, long-horizon holders whose scale makes any error large. They tend to hold and elect conservatively, often taking the default or the income option, and they rarely trade around events. Their significance for servicing is size: because their positions are enormous, a mis-calculated entitlement or a mis-applied withholding rate on a single event is a large absolute number, and their governance means they scrutinise servicing quality closely.

Hedge funds

Active, event-driven, and the source of most of the complexity around record dates. Hedge funds trade around corporate actions deliberately — capturing dividends, arbitraging rights, tendering into buybacks — and they are heavy users of short positions and borrowed stock. That activity generates the pending trades, the shorts and the borrows that make the record-date snapshot hard, and therefore the market claims and manufactured payments that follow. They also respond late and precisely, pushing elections to the cut-off.

Retail investors

Many small holders, high in number, politically visible, and the hardest to reach for elections. Individually tiny, collectively they can be a large part of a holder base, and reaching every one of them with a notification and collecting a valid instruction before the deadline is genuinely difficult. Retail non-response is the norm, which makes the default and the proof of communication especially important. Retail also carries a conduct and political dimension — a poorly handled event affecting thousands of small holders becomes a reputational matter, not just an operational one.

Investor type Holding horizon Election behaviour Why the bank must watch them
Long-only manager Medium Policy-driven, sometimes slow Heavy lenders; positions often lent over record date
Pension / sovereign fund Long Conservative, default-leaning Huge positions magnify any entitlement or tax error
Hedge fund Short, event-driven Late, precise, active Shorts and borrows drive claims and manufactured income
Retail Mixed Low response rate Hard to reach; default and proof of notice are critical

Retail carries a dimension the others do not

Retail holders bring a conduct and political weight out of all proportion to their individual size. A processing failure that affects one institutional client is a commercial matter resolved between professionals. The same failure spread across thousands of small holders — a missed dividend, a lapsed right nobody was told about clearly — becomes a conduct issue, a press story, and sometimes a regulatory one. This is why SRD II's insistence that information reach the end investor matters most for retail, and why the proof of notification and the correctness of the default are scrutinised hardest where the holders are many and small. The bank's obligation to reach every holder is not just operational diligence; for retail it is a supervised duty with reputational teeth.

Where corporate actions touches the money-making desks

A function with no revenue of its own still moves numbers the front office trades on, and the connection is closer than it looks. Corporate actions is not sealed off from the trading floor; it feeds and is fed by several revenue businesses, and errors in it show up as someone else's profit-and-loss.

Dividends are a traded asset, not just a servicing task. Desks trade the dividend itself — through dividend futures and swaps, through index arbitrage around the ex-date, and through strategies that harvest or avoid the dividend depending on tax treatment. Those desks depend on the same event data the servicing function scrubs: the rate, the ex-date, the withholding treatment. A wrong dividend forecast is a wrong hedge, so the quality of corporate-action data is a direct input to trading risk, not only to client servicing.

Securities lending is a revenue business that collides with corporate actions on every record date. The desk that lends the bank's or its clients' stock earns a fee for doing so, but lending over a record date is exactly what creates the manufactured-income and market-claim complexity of Part 2. There is a genuine tension: the lending desk wants the position out on loan earning a fee; the servicing function has to reproduce the entitlement and get the tax right on the manufactured leg. Recalling stock before a sensitive event, or lending through it, is a commercial decision with a servicing cost attached.

Event-driven trading turns corporate actions into an investment strategy. Merger arbitrage, rights-issue trading and tender participation are whole strategies built on corporate events, and the desks running them are among the most demanding consumers of accurate, timely event data and clean elections. When the servicing function processes an event late or wrong, an event-driven position can be mispriced or mis-hedged. The back-office function and the front-office strategy are reading the same event; one services it, the other bets on it.

Confidentiality and information barriers still apply

Corporate actions is post-trade servicing, but it is not free of inside information. A pending merger, a scheme of arrangement, or an as-yet-unannounced tender can be material non-public information, and the teams handling it must be walled off from anyone who could trade on it. The bank maintains restricted lists of securities where knowledge is sensitive, controls access to event data so that only those who need it can see it, and keeps an audit trail of who saw what and when. The principle is the familiar one: an institution that knows something confidential must not let one part of itself exploit what another part knows.

Enforcing confidentiality is a technology problem, not just a policy one. Access controls on the corporate-actions platform limit event visibility to the servicing team; restricted-list systems flag securities under embargo; surveillance monitors for dealing that lines up suspiciously with event knowledge; and immutable logs record every access for later review. As new communication channels proliferate — chat, collaboration tools, mobile — the surface that must be monitored keeps expanding, which makes the barrier a moving target rather than a fixed wall.

Six teams inside the bank make the function work

Corporate actions blends scarce human expertise, a large operations base, and a specialised technology platform, and it takes six distinct teams. Understanding who does what — and where the information barrier runs through them — is essential to knowing how work actually flows.

CA analyst / event manager

The domain experts who read announcements and own the golden record. They interpret the legal text, build and validate the event, and manage the complex reorganisations that resist automation. They are the reputation of the desk — the scarce, deeply experienced people whose judgement decides the hard cases. Their primary technology is the capture and scrubbing platform.

CA operations / processing

The high-volume engine room that runs entitlements, elections, payments, claims and reconciliation to deadline. Where the analysts handle the exceptional, operations handles the throughput — thousands of routine events moved correctly and on time. Their primary technology is the entitlement engine and the payment and reconciliation systems.

Client services

The relationship layer around events. They notify clients, chase elections before the cut-off, answer queries, and manage the client experience of every event. When a retail holder does not respond or an institution disputes a default, client services is the interface. Their primary technology is the notification and client-portal systems.

Data and reference management

The foundation the whole function stands on. They own the securities and event reference data, the vendor feeds, and the mappings between identifiers and standards. If reference data is wrong, everything above it is wrong, so this team is quietly one of the most consequential. Their primary technology is the securities master and reference-data platform.

Risk, control and audit

The owners of the four-eyes framework and the loss record. They run the control checks, track errors, breaks and near-misses, and provide the operational-risk oversight that keeps the function honest. They are independent of the processing teams by design, so that the people checking are not the people doing. Their primary technology is the control, exception and management-information systems.

Business analysis, IT and change

The team that owns the platform and where transformation lands. They run the corporate-actions system, the entitlement engine, the messaging layer, the feeds and the automation, and they deliver the change agenda — automating capture, migrating to modern messaging, tightening controls. This is where consulting and technology work concentrates, because the transformation of the function is fundamentally a platform-and-data programme.

Voting rides the same rails as elections

General-meeting voting is, mechanically, a voluntary corporate action, and the bank services it on the same chain. A record date fixes who may vote; holders must be notified of the meeting and the resolutions; instructions — the votes — are collected and passed up the chain; a deadline applies; and confirmation flows back. That is the election cycle in all but name, which is why the same intermediary chain, the same messaging, and the same deadline discipline used for corporate-action elections are reused for proxy voting, often on the same platforms and by the same teams.

Specialist proxy and voting agents sit alongside the bank in this flow. Where corporate-action elections route through custodians and depositories, votes often route through specialist proxy platforms that handle shareholder identification, meeting notification and vote collection. The bank connects to them the way it connects to the rest of the chain, and reconciles that the votes it submitted were recorded.

SRD II turned best-effort servicing into a legal obligation

The second Shareholder Rights Directive made the flow of information and rights along the chain a matter of law, not courtesy. Under SRD II, issuers can require intermediaries to identify who their shareholders are, and the chain must respond. Meeting and event information must reach the end investor quickly, in standard electronic format. Votes and elections must be facilitated back up the chain, with confirmation that the vote was recorded. And all of this runs to prescribed timelines and standardised formats. What was once a servicer doing its reasonable best became a set of enforceable obligations with deadlines — which is exactly why proxy and shareholder-communication processing has been pulled into the same standardised, electronic, deadline-driven world as corporate actions.

Two adjacent servicing functions that ride the same chain

Corporate actions rarely travels alone; two neighbouring services run on the same custody rails and are often owned by the same teams. Recognising them completes the picture of what asset servicing actually delivers.

Class-action and securities-litigation servicing recovers money for holders after the fact. When an issuer settles a securities lawsuit, the money is distributed to investors who held the security during a defined period — mechanically, a distribution against a historical position, which is corporate-actions logic run backwards in time. A servicer that offers this monitors eligible settlements, reconstructs which clients held the affected security in the relevant window, files the claims, and distributes the proceeds. Done well it recovers real money for clients and is a differentiator; done poorly it leaves client money unclaimed on the table. It uses the same position history, the same reconciliation discipline, and often the same teams as corporate actions.

Proxy voting is where servicing meets stewardship and ESG. Part 2 and the voting section showed that a general meeting is mechanically a voluntary event; commercially, it is also where asset managers discharge their stewardship obligations — voting on directors, executive pay, and increasingly environmental and social resolutions. For many end investors, being able to vote their holdings easily and reliably is now a regulatory and reputational requirement, not a nicety, and SRD II made facilitating it a legal duty. So the same chain that pays a dividend also carries a manager's ESG vote, and the quality of the voting service has become part of how asset managers are themselves judged by their clients. Servicing quality, in other words, now extends into the stewardship agenda.

Global and sub-custodians: the same holding, seen at different depths

Two kinds of custodian sit in the chain, and confusing them is confusing where a deadline actually falls. A global custodian offers a client a single point of access to many markets, holding their securities worldwide through a network of local agents. A sub-custodian is one of those local agents, holding securities in a single market and connecting directly to that market's depository. For a corporate action, the event is captured at the local level by the sub-custodian, passed up to the global custodian, and passed on to the client — each layer adding its own cut-off. The practical consequence is that a client's usable deadline is set by the deepest, earliest cut-off in their particular chain, not by the market deadline they might read in the announcement. Knowing which custodians sit in a client's chain is knowing when their instruction is really due.

The operating model: who does what, at which stage, on which system

The six teams are not a flat list; each owns a stage and a system, and the information barrier runs through them. Mapping the teams onto the lifecycle shows how work actually hands off, and where a break in one team becomes a problem for the next.

Team Primary stage Primary system Division
CA analyst / event manager Capture Capture & scrubbing platform Asset servicing
CA operations Entitle, elect, pay Entitlement engine, payments Operations
Client services Elect Notification & client portal Client-facing ops
Data & reference management Cross-cutting Securities master, feeds Data management
Risk, control & audit Cross-cutting Control, exception, MI tools Independent control
Business analysis, IT & change Cross-cutting The platform itself Technology / change

Read the table as a relay, not a hierarchy. An event passes from the analyst who builds the golden record to operations who entitle and pay it, with client services running the election in parallel, data management holding the ground beneath them, and risk and control checking across the whole. The change and technology team sits underneath all of it, owning the platform that the others use. A weakness anywhere in the relay — a stale reference record, a slow exception clear, a missed notification — surfaces downstream as someone else's break, which is why the teams are designed to check each other rather than work in isolation.

  1. Describe the two-way chain in one sentence.

  2. Why does every hop in the chain matter?

  3. How does hedge-fund behaviour increase servicing complexity?

  4. Why is investor type a predictor of processing risk?

  5. Why does corporate actions still need information barriers?

  6. Which team owns the golden record, and which owns the platform that will be transformed?

3

  1. Which way do elections and votes flow?

  2. Why does every hop in the chain matter?

  3. How do hedge funds increase servicing complexity?

4Systems, Data, and the Standards That Bind Them

Corporate actions is a data-quality problem wearing an operations costume, so the systems and the standards are the heart of the discipline. This part opens the platform from feeds to payment, maps the vendor stack, follows one event object front to back, and details the identifiers, integration patterns, messaging standards and regulations that hold it all together.

Start with the functions, before naming any vendor

Ask first what any corporate-action capability must do, regardless of who builds it. Every servicer, however it is assembled, needs the same functional blocks in the same order. Feeds and capture ingest vendor, depository and agent messages plus unstructured announcements. Scrub and golden record match sources, flag breaks, and produce one trusted event. The entitlement engine applies terms to record-date positions and computes cash and stock per holder. Election and workflow notify holders, capture instructions, apply defaults and manage deadlines. And payment, tax and reconciliation book the movements, withhold tax, reconcile to the custodian, and confirm and report. Running underneath all of them are two cross-cutting layers: reference and position data feeding the whole chain, and controls, audit and management information running across every stage.

Figure 6 The functional architecture — five blocks over two cross-cutting layers
Functional block What it does Core data it owns
Feeds & capture Ingest and normalise every source Raw source messages
Scrub & golden record Reconcile sources into one truth Validated, versioned event
Entitlement engine Position × terms per holder Per-holder entitlements
Election & workflow Notify, collect, default, deadline Instructions and defaults
Payment, tax & recon Move, withhold, reconcile, confirm Bookings and reconciliations
Reference & position (cross-cutting) Instrument and holdings data Securities master, ledger
Controls, audit & MI (cross-cutting) Four-eyes, exceptions, audit trail Control and near-miss records

The vendor landscape: mostly buy, and the value is in the integration

Corporate actions is a vendor-heavy domain, because the standards run so deep that reinventing them makes no sense. The stack has several layers, and no single vendor does everything. Event data providers — Bloomberg, LSEG (Refinitiv), SIX, and specialist corporate-action data houses — supply the structured feeds banks scrub against, and several are used at once for cross-checking. Corporate-action processing platforms — specialist vendor systems — provide capture, scrubbing, entitlement, elections and payments, and most banks buy and integrate one rather than build. Custodian and depository platforms — the global custodians and the central securities depositories — run their own event engines that the bank connects to and reconciles against. SWIFT and messaging carries the standardised event messages between all parties. Proxy and shareholder-rights platforms handle meeting notification and vote collection. And reconciliation and workflow tools sit around the core platform to control exceptions.

Most banks buy the platform and build the integration. The domain is too deep and too standards-driven to rebuild, so the platform is bought. But the platform alone does nothing — the real work, and the real competitive difference, is wiring it into the bank's post-trade estate: connecting feeds, positions, payments, messaging and reconciliation so the stack talks cleanly. The one component banks tend to keep proprietary is the entitlement engine at the core, because it must be exactly right, fast, and fully auditable, and because it encodes the bank's own rounding, tax and booking logic.

The entitlement engine is the heart of the platform. It takes the golden-record terms and every holder's record-date position and computes the exact cash and securities movement — with correct rounding, fractions, currency and tax. It must be right, fast and fully auditable, because a wrong engine is a wrong payment at scale. This is why, even in a bought stack, the engine and its calculation logic are the component banks scrutinise most and often keep closest.

Build, buy, or hand the whole thing to someone else

The build-versus-buy choice has a third option that is really a business decision: outsource the function entirely. Beyond building a platform or buying and integrating one, a bank can outsource corporate-action processing to a third party — a large asset servicer or a business-process provider that runs the capture, scrubbing and even the operations on the bank's behalf. Smaller banks and asset managers, for whom a full in-house capability is uneconomic, frequently take this route, effectively renting another institution's scale. The trade is the familiar one: outsourcing converts a fixed cost and a scarce-expertise problem into a variable fee, but it hands operational control — and some of the client relationship — to a provider, while the accountability to the end client and the regulator stays with the bank. You can outsource the processing; you cannot outsource the liability.

The logical endpoint of that thinking is the industry utility, and it is a recurring ambition. Because every bank scrubs the same public announcements into the same golden records, the work is substantially duplicated across the industry, which invites a mutualised utility that captures and validates events once for everyone. Several such efforts have been attempted, and the pull is real: shared cost, shared standards, one clean source. The resistance is equally real — banks are wary of depending on a shared critical service, of losing a differentiator, and of concentrating risk in one place. The result is a persistent tension between the obvious efficiency of doing the scrubbing once and the reluctance to let a single utility sit at the centre of everyone's servicing. Where the function lands on the build-buy-outsource-mutualise spectrum is one of the defining strategic questions of the domain.

The data model, laid out as a table

The event model introduced in Part 2 is worth seeing as a structure, because it is the schema every corporate-action system is built around. Six entities, in a one-to-many hierarchy, absorb every event type as data rather than as code.

Entity What it holds Relationship
EVENT Type, MAND/VOLU/CHOS indicator, ISIN, status, golden-record version One per event
DATES Ex, record, deadline, pay, plus internal cut-offs One set per event
OPTION Each choice (exercise, lapse, cash, shares) with a default flag Many per event
TERMS Cash rate, share ratio, price, currency, rounding rule One set per option
ENTITLEMENT Record-date position combined with the chosen option's terms One per holder per option
INSTRUCTION The client's election, with quantity, timestamp, validation, audit One per holder

The power of the model is that complexity becomes data, not new code. A scrip dividend is one event with two options; a multi-way merger is one event with several options and richer terms; a mandatory dividend is one event with a single implied option. The same six entities describe all of them. Get this structure right once and the platform handles the next exotic event by configuration; get it wrong and every unusual event becomes a bespoke build, which is exactly how backlogs and errors accumulate.

The vendor stack, layer by layer

No single vendor spans the whole function, so a bank's capability is a stack that must be assembled and integrated. The layers are distinct, and each has its own providers.

Layer What it provides Typical providers
Event data Structured feeds to scrub against Bloomberg, LSEG (Refinitiv), SIX, specialist data houses
Processing platform Capture, scrub, entitle, elect, pay Broadridge, FIS, SmartStream, SS&C
Custodian / depository Their own CA engines to reconcile against Global custodians, Euroclear, Clearstream, DTCC
Messaging Connectivity for MT and ISO 20022 SWIFT
Proxy / SRD II Shareholder ID, meeting, vote collection Specialist proxy platforms
Reconciliation / workflow Break management and case workflow Matching and workflow tools

The processing platforms have names worth knowing. In practice a bank choosing not to build will look at a short list of specialist systems: Broadridge (a large global asset-servicing and corporate-actions platform, offered both as software and as a managed service), FIS (whose SWIFT-accredited Corporate Actions Suite spans the event lifecycle and layers on robotic process automation), SmartStream (whose Smart Actions covers all ISO event types with SRD II proxy support), and SS&C, among others. On the data side the scrubbing feeds come from Bloomberg, LSEG (Refinitiv), SIX and S&P Global (IHS Markit); on the reconciliation and workflow side, tools from SmartStream (TLM), Gresham and others sit around the core. Naming them matters because the real decision a bank faces is not "platform or no platform" but which combination of these to integrate, and each carries different strengths in data, automation and proxy coverage.

The integration reality is the strategic point. Because no product does everything, the bank's real asset is the wiring: feeds into the platform, positions from the ledger, payments out, messaging across, reconciliation around. Most transformation value comes from making that stack talk cleanly rather than from swapping any single component, which is why corporate-action programmes are integration programmes first and product selections second.

Follow one event object from announcement to regulatory report

The clearest way to understand the platform is to trace a single object through it. An event begins as raw source messages at capture, becomes a validated golden record after scrubbing, is combined with positions to produce entitlements in the engine, generates instructions during election, produces bookings and confirmations at payment, and finally flows into the regulatory and audit reporting that closes it. At every step, the object created depends on the object before it.

Figure 7 Data lineage — each object depends on the one upstream, so an error cascades

The cascade principle is the reason data quality is everything. Because each downstream object depends on an upstream one, an error at the top reaches every system below it. A wrong ratio in the golden record produces wrong entitlements for every holder, wrong instructions, wrong payments, and a wrong regulatory report — the cascade has no natural stopping point except a control that catches it. This is why the investment goes into getting the golden record right, and into the checks that sit between the golden record and the payment: they are the only places the cascade can be interrupted before money moves.

Identifiers are the glue that lets systems agree what they are talking about

Before any two systems can process the same event, they must agree on the same security, the same entity, and the same event. Three identifier families do most of that work, and a fourth is specific to corporate actions.

Identifier What it names Where it bites
ISIN The security (the instrument) The event applies to an ISIN; wrong ISIN, wrong holders
LEI The legal entity (issuer or holder) Ties the event and the holder to a legal identity for reporting and tax
FIGI Vendor-neutral instrument mapping Reconciles the same instrument across data sources
COAF / official event reference The specific corporate-action event Lets every party refer to one event unambiguously up and down the chain

The event needs its own identifier, not just the security's. A single ISIN can have many events over its life — several dividends, a split, a rights issue — so processing needs an unambiguous reference for the specific event, not just the instrument. The official corporate-action event reference (in Europe, the COAF, the Common Official Corporate Action Reference) gives every party one code to refer to the same event, which is what makes automated matching and claims possible across the chain. Without it, two systems can agree on the security and still be processing different events.

Integration patterns: from nightly batch to real-time streaming

How systems pass data to one another is not a plumbing detail — it decides how fast errors surface. Three patterns coexist. Batch is the legacy default: files exchanged on a schedule, typically overnight, simple but latent — a break introduced in the morning may not surface until the next cycle. APIs allow on-demand queries and updates, the modernising step that lets systems ask for current data rather than wait for a file. Event streaming — a publish-and-subscribe model in the style of Kafka — pushes changes in real time as they happen, so downstream systems react immediately.

The migration from batch to streaming is more than a speed upgrade — it changes the control model. Under batch, systems drift out of step between cycles and must be reconciled back together, which is why so much of the function is reconciliation. Under streaming, a change is pushed to every subscriber as it happens, so systems stay in step and the need for after-the-fact reconciliation shrinks. Moving to real-time messaging therefore attacks the reconciliation burden at its source rather than just processing it faster.

What a modern corporate-actions platform actually looks like

The abstract move from batch to streaming has a concrete architecture, and it is worth naming the parts. A contemporary platform is decomposed into microservices rather than one monolith: a capture service ingesting feeds, a scrubbing service producing the golden record, an entitlement service doing the maths, an election service running the workflow, a payment service booking movements. They communicate over an event backbone — commonly a Kafka-style log — where each state change (a golden record published, an instruction received, a payment booked) is an event other services subscribe to. This is the technical realisation of the exception-based model: services react to events as they happen instead of waiting for the next batch file.

APIs replace file drops as the way systems ask each other for data. Where the legacy pattern exchanged fixed-format files overnight, the modern pattern exposes REST or messaging APIs returning structured JSON or ISO 20022 XML, so a client portal can query an event's status in real time and an election can be submitted and validated on the spot. Increasingly the platform runs in the cloud, which matters less for cost than for elasticity: dividend season concentrates volume into a few weeks, and cloud capacity can scale to the peak and back down rather than sizing owned hardware for a load that sits idle most of the year. An event-sourced design — storing the full sequence of state-changing events rather than only the latest state — also gives, almost for free, the append-only audit trail the function is obliged to keep.

Cloud, microservices and event streaming are not modernisation for its own sake; each answers a specific corporate-actions problem. Microservices let the risky, bespoke parts (complex-event handling) be changed without redeploying the whole platform. The event backbone removes the reconciliation that batch drift creates. Cloud elasticity meets the dividend-season peak without permanent over-provisioning. And event sourcing turns the regulatory audit obligation into a property of the architecture rather than a bolt-on. A platform that adopts these without tying each to a problem it solves has bought fashion; one that maps them to these four needs has bought leverage.

Testing an entitlement engine is its own engineering discipline

Because a wrong engine is a wrong payment at scale, how the calculation core is tested is as important as how it is built. An entitlement engine cannot be signed off on a handful of happy-path cases; it must be exercised against the full zoo of event types, rounding rules, currencies and edge cases. Four techniques do the heavy lifting. Regression suites run a library of known events with known correct outputs after every change, so a fix to one event type cannot silently break another. Parallel running processes real events through the new engine alongside the incumbent and compares the results before the new one is trusted, catching discrepancies on live data without risking a client. Golden-copy validation checks each computed entitlement against an independently derived figure, the automated form of the four-eyes principle. And reconciliation tooling matches what the engine booked against what the custodian actually delivered, closing the loop from calculation to settlement.

The hard part of testing is generating the edge cases, not running them. The events that break engines are the fractional positions, the multi-option mergers, the conditional tenders and the cross-border tax cases — exactly the bespoke long tail that arrives rarely in production. Good practice is to build a synthetic event library that deliberately includes these, so the engine meets them in a test harness before it meets them for real. An operation that only tests against the events it has recently seen is an operation that will meet its next complex event live, which is precisely where the losses are.

Compliance-critical records must be append-only, and the engine stays proprietary

Because corporate-action records may be needed to prove what the bank did, they must be immutable. The golden record, the entitlement, the client's instruction and the payment all form part of the evidence trail, so the systems that hold them are append-only: records are added and superseded, never overwritten or deleted, so the full history of every event survives. Governance platforms enforce this and make the trail queryable. And the decision logic at the centre — the entitlement engine and the rules it applies — is almost always proprietary, because it encodes the bank's own judgement about rounding, tax and booking and must be defensible line by line.

Immutable audit and the proprietary decision engine

Two design choices define a defensible corporate-action platform: records that cannot be quietly changed, and a calculation core the bank fully controls. Because the golden record, the entitlement, the instruction and the payment may all be needed to prove what the bank did and why, the systems that hold them are append-only — a record is superseded by a new version, never overwritten, so the full history survives every correction and every dispute. And because the entitlement engine encodes the bank's own rules for rounding, fractions, tax and booking, it is almost always proprietary and closely held, even when the surrounding platform is bought. The bank can outsource the plumbing; it cannot outsource the judgement that decides what a client is paid.

The audit trail is not a passive log — it is an active control. When a client disputes a default, when a regulator asks why an event was processed a certain way, or when an internal review reconstructs a near-miss, the append-only history is the evidence. It records who set each golden-record value, who verified it under four-eyes, when each notification went out, what instruction was received and when, and exactly what moved on pay date. A platform that cannot reconstruct that chain for any event is not merely inconvenient; it is unable to prove it did the right thing, which in a supervised, irreversible business is itself a failure.

Messaging protocols: two standards, mid-migration

Corporate-action information travels the custody chain as structured messages, and the industry is between two standards. The incumbent is ISO 15022, carried over SWIFT as MT messages: MT564 notifies of an event, MT565 carries an instruction, MT566 confirms, and MT567 reports status. The successor is ISO 20022, the richer, more structured XML standard, whose securities-events (seev) messages — seev.031 to notify, seev.033 to instruct, seev.036 to confirm — carry more detail with less ambiguity. Alongside both, unstructured sources — issuer PDFs, press releases, prospectuses — still need human or machine interpretation. The market is mid-migration from MT to ISO 20022, so systems must handle both and map between them.

Figure 8 The messaging sequence — notify down, instruct up, confirm back
Layer Standard Carries Timescale
Notification MT564 / seev.031 Event terms, dates, options On announcement, then updates
Instruction MT565 / seev.033 The holder's election Before the deadline
Status MT567 Acknowledgement and status Around instruction
Confirmation MT566 / seev.036 What actually moved On pay date

The two messaging worlds, made concrete

Corporate-action messaging splits into a notification-and-instruction world and a settlement-and-confirmation world, and the split mirrors the front-to-back structure of post-trade. The notification messages (MT564, seev.031) and instruction messages (MT565, seev.033) carry the event and the choice — the servicing conversation between the bank, the depository and the client. The confirmation and status messages (MT566, MT567, seev.036) carry what actually moved — the settlement record. The two worlds run on the same rails but answer different questions: one asks "what is the event and what do you choose", the other reports "here is what was delivered". A servicer has to speak both fluently and reconcile one against the other, because a notification that was never confirmed, or a confirmation that does not match the instruction, is a break.

Inside a real message: the MT564 and its ISO 20022 successor

A technologist building or assessing a platform needs to see the actual message, not a stylised one, because the structure is the specification. An ISO 15022 MT564 corporate-action notification is a sequence of coded fields grouped into blocks, and each field has a qualifier that tells the parser what it is. The event reference sits in a `:20C::CORP//` field; the official market reference — the pan-European COAF — in `:20C::COAF//`. The event type is carried in `:22F::CAEV//` with a four-letter code (DVCA for a cash dividend, RHTS for a rights issue, TEND for a tender, MRGR for a merger); the mandatory-versus-voluntary indicator in `:22F::CAMV//` (MAND, VOLU or CHOS). The security is in `:35B:` as the ISIN; the dates arrive as `:98A::` fields with their own qualifiers — RDTE for record date, XDTE for ex-date, MKDT for the market deadline, PAYD for pay date. Options and their terms sit in repeating blocks, one per choice.

ISO 20022 carries the same meaning in a richer, self-describing XML structure. The successor message is `seev.031` (CorporateActionNotification), with `seev.033` for the instruction and `seev.036` for the movement confirmation. Instead of positional qualifiers it uses named XML elements — an event type element, a set of date elements, a repeating options structure — which are easier to validate against a schema and less prone to the silent mis-parse that a mis-placed qualifier can cause. The trade-off is verbosity: an ISO 20022 message is far larger than its MT equivalent, which is one reason the migration is a genuine engineering programme and not a find-and-replace.

Element ISO 15022 (MT564) ISO 20022 (seev.031)
Event reference `:20C::CORP//` Named reference element
Official / COAF reference `:20C::COAF//` Official corporate-action reference element
Event type `:22F::CAEV//DVCA` Event-type code element
Mand/volu indicator `:22F::CAMV//VOLU` Mandatory-voluntary element
Security `:35B:` ISIN Financial-instrument element
Record / ex / pay date `:98A::RDTE//`, `XDTE`, `PAYD` Named date elements
Options and terms Repeating `:22H:`/`:92x:` blocks Repeating options structure

The practical consequence for a platform is that it must run both standards at once for years, and map cleanly between them, because the market is mid-migration and counterparties send whichever they support. That mapping is not trivial: a rich ISO 20022 message can carry detail an MT message cannot, so down-converting loses information, while up-converting from MT has to synthesise structure that was implicit. A corporate-actions platform's messaging layer is therefore one of its most tested components, and the SWIFT Standards Release cycle — which updates message definitions annually — means it is never finished.

Integration in practice: where the value and the fragility both live

A bought platform is only as good as the connections around it, and those connections are where projects succeed or fail. The entitlement engine must read positions from the custody ledger, terms from the golden record, tax rules from a tax engine, and reference data from the securities master; it must write bookings to the payment system and confirmations to the messaging layer; and every one of those interfaces is a place where data can arrive late, in the wrong shape, or not at all. This is why most of the transformation value in corporate actions comes not from any single product but from making the stack talk cleanly — reducing the number of hand-offs, standardising the data that crosses them, and moving from nightly batch toward real-time streaming so that the systems stay in step rather than drifting apart between cycles.

Operational resilience: some moments have no second chance

Corporate actions has irreversible moments, and resilience is about surviving exactly those. A pay date, an election deadline, a large merger's effective date — these are moments where a system failure at the wrong instant is not an inconvenience but a loss, because the deadline does not wait for the system to come back. Resilience is built from precise components. Redundancy means no single point of failure. Automatic failover means a standby takes over without manual intervention. Disaster recovery means the ability to restore service at another site. And business continuity means the people and processes can keep working when systems or premises are unavailable.

Concept What it guarantees Why it matters on the critical day
Redundancy No single point of failure A component failure does not stop processing
Automatic failover Standby takes over seamlessly The deadline is met even mid-incident
Disaster recovery Service restored elsewhere A site loss does not become a missed pay date
Business continuity People and process keep running Elections still collected if systems degrade

Resilience is specified in numbers, and the numbers are unusually demanding. Two targets define it. The recovery time objective (RTO) is how quickly service must be restored after a failure; the recovery point objective (RPO) is how much data, measured in time, the bank can afford to lose. For most systems these are measured in hours; for the deal-critical moments of corporate actions — the election deadline, the pay run, a large merger's effective date — the practical RTO is near zero, because a system that comes back an hour after an immovable deadline has come back too late, and the RPO is effectively zero, because a lost instruction is a lost client choice. Throughput matters too: the platform must clear the dividend-season peak, which can be several multiples of the average daily event count, without the exception queue backing up past the cut-offs. These are the non-functional requirements that separate a platform that works on an ordinary Tuesday from one that works on the one morning that cannot be repeated.

Five regimes shape every system in the chain

Corporate actions is governed as much by market standards as by law, and each regime reads as a system specification. Five forces shape the platform, and every one of them pushes the same way — toward standardisation, electronic messaging, shorter deadlines and greater transparency.

SRD II, the second Shareholder Rights Directive, mandates that event and voting information reach the end investor in standard electronic format, that issuers can identify their shareholders, and that rights are facilitated up the chain — turning best-effort servicing into a legal obligation with prescribed deadlines. The ISO 20022 migration is the data-standardisation programme that makes messages richer and more automatable. CAJWG and SMPG standards harmonise how events are processed across Europe so the same event behaves consistently across borders. CSDR and T2S underpin harmonised settlement, automated market claims and shorter cycles across the euro area. And tax rules, including the EU FASTER directive, aim to standardise and digitalise withholding relief and reclaim. Around all of these, general conduct, client-asset and operational-resilience requirements make correct, timely servicing a supervised obligation rather than a courtesy.

Read as a table, the regimes are a specification, not a legal summary. Each one imposes concrete requirements on the platform — formats to support, deadlines to meet, records to keep, reports to produce.

Regime What it mandates System implication
SRD II Shareholder ID, information down, rights up Standard electronic messaging, prescribed deadlines, vote confirmation
ISO 20022 migration Richer structured messaging Dual MT/ISO support and mapping during transition
CAJWG / SMPG Harmonised market practice Configurable market rules, consistent event handling
CSDR / T2S Harmonised settlement, automated claims Connection to T2S, claim reconciliation
Tax (FASTER, treaties) Correct withholding, relief and reclaim Tax engine, residency documentation, reclaim tracking
Conduct / resilience Correct, timely, resilient servicing Audit trail, proof of notice, failover to the deadline

The workflow and break-management layer around the core

The core platform computes; the workflow layer decides what a human sees and when. Around the entitlement engine sits a set of case-management and matching tools whose only job is to control exceptions — to make sure that the scarce human attention goes to the right break at the right time and that nothing falls through. Matching engines compare sources and positions and raise breaks; case-workflow tools route each break to the right analyst with the context to resolve it; and deadline-driven alerting escalates anything approaching a cut-off. This layer is what makes the exception-based operating model real: without it, an analyst would have to hunt for the breaks; with it, the breaks come to the analyst, prioritised by urgency and value.

Break management is where the function's day is actually spent. On any given day, the routine events flow through untouched and the team works a queue: a source disagreement here, an unreconciled item there, a non-responding client to chase, a tax document to confirm. The quality of a corporate-action operation is largely the quality of its break management — how fast breaks are surfaced, how well they are prioritised, and how reliably they clear before the deadline that would turn them into losses.

Reference data is the quiet foundation under everything

Every calculation in the function assumes the reference data beneath it is correct, and that assumption is where silent failures begin. The securities master holds the identity, terms and characteristics of every instrument; the position ledger holds who holds what, where, and how much is lent; and the mappings between identifiers and standards let the systems agree what they are processing. None of this is glamorous, and all of it is load-bearing. A stale instrument record, a mis-mapped identifier, or a position that does not reconcile to the custodian will produce a wrong entitlement no matter how perfect the event data is. This is why data and reference management, though it never touches an individual event, is one of the most consequential teams in the function: it maintains the ground the whole platform stands on.

  1. Why start with functions before naming vendors?

  2. Where do banks tend to build rather than buy, and why?

  3. State the cascade principle.

  4. Why does a corporate action need its own identifier beyond the ISIN?

  5. What does moving from batch to event streaming change beyond speed?

  6. Name the notification, instruction and confirmation messages in both standards.

  7. Why is redundancy alone insufficient for resilience here?

  8. In what sense does ISO 20022 underpin the rest of the transformation agenda?

4

  1. What do banks often keep proprietary?

  2. An error in the golden record reaches…

  3. Why name functions before vendors?

5Risk, Real Failures, and the Transformation Agenda

Corporate actions is a low-frequency, high-severity operational risk with a fat tail, and the losses are real, direct and often irreversible. This part lays out the risk taxonomy, shows how to tell a well-run function from a lucky one, examines three real failures covering three distinct modes, and assesses the transformation agenda — including where generative AI genuinely helps and where a human must stay in the loop.

Irreversibility is the organising principle of the whole discipline

One property explains why corporate actions is controlled the way it is: its mistakes usually cannot be undone. In most parts of a bank an error can be corrected — a mis-booked trade is cancelled and rebooked, a wrong price is amended before settlement. Corporate actions does not offer that mercy. Once cash pays to the wrong party, once an election deadline passes with no instruction, once a position is reorganised on a wrong term, the outcome is largely fixed and the loss lands on the bank. This single fact reorganises everything upstream. It is why the controls sit before payment rather than after it, why a second person verifies the golden record before it goes live, why the reconciliation gate blocks any payment on an open break, and why resilience is measured against the deadline rather than against average uptime. Every design choice in the function is an answer to the same question: how do we catch the error while it can still be caught, because afterwards there is no afterwards.

Read the whole module through that lens and it becomes coherent. The golden record exists because a wrong term is irreversible once paid. The nested deadlines exist because a missed election is irreversible once the market closes. The independent entitlement check exists because a wrong calculation is irreversible once it moves. The three real failure cases are three ways irreversibility bit. And the caution about generative AI is caution about handing an irreversible decision to a system that can be confidently wrong. Irreversibility is not one risk among many; it is the reason the discipline has the shape it does.

Six ways a corporate action goes wrong

The risks are best understood as a small, complete set rather than a long list. Six categories cover the ways an event turns into a loss, and each has a distinct owner and a distinct defence. Seeing them as a map, rather than a catalogue, shows that they are not equally controllable — some are arithmetic and preventable, others are judgement calls under time pressure.

Figure 9 The six risk types — most catastrophic losses cluster on election and deadline
Risk type What it is Who owns it Main defence
Data / capture Wrong terms or date captured CA analysts / data mgmt Dual-source scrubbing, four-eyes
Entitlement Miscalculated position × terms CA operations Independent recalculation
Election / deadline Missed or wrong instruction Client services / operations Digital capture, deadline alerts
Settlement / recon Pay more than received CA operations / settlement Reconciliation gate
Tax / withholding Wrong rate, failed reclaim Tax operations Documentation, four-eyes on rates
Conduct / regulatory SRD II breach, client harm Risk and compliance Audit trail, proof of notice

The loss distribution has a fat tail, and that shapes everything

Most events are tiny and routine, so it is tempting to under-invest — and that is the trap. The overwhelming majority of corporate actions are small dividends that process without incident, which makes the function look cheap to run and easy to neglect. But the loss distribution is not smooth: a single mishandled voluntary event or a botched merger can dwarf a full year of ordinary processing cost. The rare, complex, high-value event — a large merger, a mandatory tender, a scheme — is exactly where the biggest losses cluster and exactly the hardest to process. That asymmetry, routine almost always and catastrophic occasionally, is why controls, data quality and straight-through processing earn their keep even though most days nothing goes wrong.

Telling a good operation from a lucky one

"No losses this year" is not evidence of a good process — it may be evidence of a quiet year. Because the tail is fat, an operation can be badly run and still lucky for a while. The way to tell them apart is to measure the process, not just the outcome. Five metrics do most of the work.

Metric What it measures Rule of thumb
Straight-through rate Events processed with no human touch Higher is better; the routine should not need hands
Exception clear time How fast breaks are resolved before deadline Must clear well inside the cut-off
Error / loss rate Events with a client-impacting error The headline risk metric; trend matters most
Election response rate Share of holders instructing before cut-off Higher means fewer defaults applied blind
Reconciliation break rate Unmatched items at pay date Low and falling; breaks are near-losses

Reading the metrics together, not one at a time

No single metric tells the truth about a corporate-action operation; the diagnosis is in how they move together. A high straight-through rate with a rising reconciliation break rate suggests automation is running ahead of data quality — events are flowing untouched that should have been stopped. A low error rate with a falling election response rate suggests the operation is quietly applying more defaults to silent holders, storing up disputes rather than avoiding them. A fast exception clear time is only reassuring if the exception queue is not also growing. The craft is to read the panel as a system: the healthy signature is a rising straight-through rate, a stable or shrinking exception queue cleared well inside deadlines, a falling reconciliation break rate, and an error rate trending down — all at once. Any one of them improving while another deteriorates is a signal that risk has moved rather than fallen.

Three real failures, three distinct modes

The instructive failures are not random — they are the three ways the process breaks. One is a failure of terms and tax discipline at scale, one a failure to capture what the event actually was, and one a failure that became permanent the moment cash moved. Together they are a map of how corporate-action and asset-servicing risk actually materialises.

Case one — cum-ex and cum-cum: wrong tax treatment, industrialised

The largest tax scandal in modern European finance grew directly out of weak controls around dividend withholding. In the cum-ex and cum-cum schemes, run roughly between 2001 and 2012, shares were traded and lent around the dividend record date so that tax authorities were induced to refund withholding tax that had either been paid only once or never paid at all. Investigations led by the Correctiv newsroom estimated the total cost to European treasuries at more than €55 billion, with Germany the hardest hit and material losses in France, Italy, Denmark and Belgium. The mechanics sat squarely in corporate-action territory — record dates, holder-of-record entitlement, securities lending, manufactured dividends and withholding-tax reclaims — and the schemes exploited exactly the ambiguity about "who really holds this on the record date" that Part 2 identified as the hardest question in the discipline. The lesson is that the tax character of an income event and the identity of the true holder are compliance-critical controls, not administrative details, and that weakness there scales into losses measured in tens of billions.

Case two — GameStop's 2022 split that was really a dividend

A single ambiguity about what kind of event this was propagated a wrong outcome across the whole custody chain. In July 2022 GameStop implemented a four-for-one stock split structured as a stock dividend — additional shares distributed to holders — rather than a conventional split. Several intermediaries, including brokers in Europe, processed it as an ordinary split, and reporting indicated the German regulator BaFin observed that the event had been treated as a stock split rather than a stock dividend by some entities. The consequence was that some European holders did not promptly receive the additional shares they were due. No fraud was involved; the failure was one of event classification at capture, cascading down a cross-border chain exactly as the lineage principle predicts. It is a textbook illustration of why the golden record must capture not just the ratio but the precise nature of the event, and why cross-border processing multiplies the risk when the classification is subtle.

Case three — Citi and Revlon: the irreversibility that governs everything

The clearest demonstration of why asset servicing is built around irreversibility came from a payment that could not be recalled. In August 2020 Citibank, acting as loan agent, intended to send about $7.8 million in interest to Revlon's lenders but instead wired roughly $900 million — the full outstanding principal — from its own funds. When it asked for the money back, some lenders refused, and in 2021 a US federal judge ruled the transfers "final and complete transactions, not subject to revocation," leaving Citi unable to recover a large part of the money until later settlements. The event was a loan-agency payment rather than a corporate action, but the principle it crystallises is the one that governs the whole of asset servicing: once cash moves to the wrong party, undoing it is hard or impossible. Unlike a mis-trade, which can often be cancelled and rebooked, a wrong payment on a corporate action lands as a direct loss. This is the asymmetry that justifies the reconciliation gate and the four-eyes framework — the controls exist because the mistake they prevent cannot be reversed.

Case Failure mode Where it maps in the lifecycle Lesson
Cum-ex / cum-cum Wrong tax treatment at scale Entitle / pay (tax, true holder) Tax character and holder identity are compliance-critical
GameStop 2022 Event mis-classified at capture Capture (golden record) Capture the exact nature of the event, not just the ratio
Citi / Revlon Irreversible wrong payment Pay (settlement) Once cash moves it is gone; controls must sit before payment

The honest state of play, before the ambition

Any credible transformation story has to start by admitting where the function actually is. Corporate actions remains, after decades of investment, one of the most manual and least straight-through functions in the bank — not because the teams are behind, but because the problem is genuinely hard: unstructured input, a bespoke long tail, fragmented data, and an irreversible cost of error. The transformation agenda is not a promise to make that difficulty disappear. It is a disciplined attack on its causes — standardise the data, source it once, automate the routine, and apply new tools only where they change economics or risk. The sections that follow describe that attack, and they are deliberately sober about what it can and cannot achieve.

Generative AI: real help at the edges, governed tightly

Generative AI attacks the exact place corporate actions hurts most — turning prose into data — but its failure mode is the loss the function exists to prevent. There are genuine, emerging uses. NLP on announcements can extract structured event data — type, dates, ratio — from PDFs and prose, attacking the root data problem. Auto-scrubbing and anomaly detection can flag likely errors, such as a ten-times rate or an implausible date, and prioritise the exception queue. Source reconciliation can match and resolve disagreements between feeds more intelligently. And election and query assistance can summarise complex offers and help staff and clients act faster.

Under the hood, the technical approach is more specific than "use AI". Extracting an event from a prose announcement is a document-processing pipeline: optical character recognition where the source is a scanned PDF, then named-entity extraction to pull out the event type, the dates, the ratio and the currency, then mapping those into the structured event model. The extraction can use classical natural-language techniques — rules and trained models tuned to the narrow, repetitive language of announcements — or a large language model prompted to return structured fields; in practice the two are combined, with the narrow models handling the common patterns and an LLM handling the messier prose. Crucially, every extracted field carries a confidence score, and the pipeline is tuned so that anything below a threshold is routed to a human rather than accepted. This is the technical shape of "human in the loop": not a person checking everything, but a system that knows what it is unsure of and asks. The same confidence-scoring logic drives anomaly detection — a ten-times rate or an implausible date is flagged not because a rule caught it but because it sits far from what the model expected.

But the caveats are not soft, because the cost of a wrong output is precisely the loss being avoided. An NLP mis-read of a ratio or a date causes exactly the error the discipline exists to prevent, so every output must be verified. The human stays in the loop by design: AI can propose the golden record, but four-eyes still approves it, and accountability does not move to the model. The bespoke, conditional events that cause the biggest losses are the hardest to automate reliably, so AI helps most on the routine, not the tail. And AI is a second step, not a first — without clean feeds, ISO 20022 and a golden source, AI simply automates a mess faster.

Use of GenAI The gain The governing risk
NLP extraction from announcements Attacks the prose-to-data root problem A mis-read ratio or date is a direct loss
Anomaly detection on feeds Prioritises the true exceptions False confidence if it misses a subtle break
Source reconciliation Resolves disagreements faster Opaque logic must remain explainable
Election / query assistance Faster, clearer client servicing Confidentiality and MNPI must be protected

The enduring pain points, and where they all trace back to

Despite decades of automation, corporate actions remains one of the least straight-through functions in the bank, and its pain points share a single root. Six recur. Data quality and unstructured input: announcements arrive as prose, feeds disagree, and reference-data gaps remain the root cause of most errors. Manual, deadline-driven work: complex events, elections and exceptions are worked by hand under time pressure — the classic error window. Fragmented systems and data: positions, feeds, elections and payments live in different systems that must be reconciled event by event. Cross-border variation: every market's rules, tax and formats differ, and long custody chains multiply deadlines and reconciliations. Shrinking timelines: T+1 and shorter deadlines compress the window to detect and fix errors. And key-person and resilience risk: deep event expertise concentrates in a few experienced analysts, making scaling and continuity genuine concerns.

Every one of these traces back to data — its quality, its structure, its fragmentation, and the manual work needed to make it usable. That is why the transformation agenda is fundamentally a data-and-standards agenda rather than an operations one: fix the data at source and standardise the messages, and most of the pain points shrink together, because they were all symptoms of the same underlying problem.

Why corporate actions resists full automation

Understanding the limits matters as much as the ambition, because corporate actions is a genuinely hard automation problem — and pretending otherwise is how programmes fail. Six forces keep a human at the critical gates. The input is unstructured and inconsistent: events originate as legal prose in many formats, languages and conventions, and perfect machine-reading of every announcement is not yet reality. The long tail is bespoke: schemes, conditional tenders and reorganisations are almost one-off, so there is not enough repetition to train a reliable hands-off model. The cost of error is asymmetric: because mistakes are irreversible and expensive, the bar for removing the human is very high — you automate the easy ninety percent, not the risky ten. The data lives across silos: positions, lending, tax and payments sit in different systems, and end-to-end automation requires integrating all of them cleanly. Standards are still landing: the ISO 20022 migration is mid-flight and markets differ, so until the data is uniform, automation stays partial. And accountability cannot be automated: someone must remain answerable for a payment to a client, so governance keeps a human decision at the critical gates by design.

The realistic target is not zero-touch, but maximum straight-through on the routine and minimum, well-controlled human effort on the complex. The goal is to shrink the manual surface and concentrate expertise where it genuinely adds value — a sober, achievable ambition rather than a promise of a lights-out operation. Any programme that promises the latter has misunderstood the shape of the risk.

Governing the model, not just using it

Deploying AI in a function where a wrong output is a direct loss means governing the model as rigorously as any other critical control. Model-risk governance asks a specific set of questions before an AI system is allowed near a golden record. Can its output be explained, so that an analyst approving a proposed event term knows why the model read the announcement the way it did? Is it validated against known-correct events, so its error rate is measured rather than assumed? Is there a monitored fallback for the cases it flags as uncertain or handles poorly — the bespoke reorganisations where it is least reliable? And is confidentiality preserved, so that material non-public information in an announcement is not leaked through the model? These are not obstacles to adoption; they are the conditions that make adoption defensible. An AI that proposes and a human who disposes, with the model measured and monitored, is a genuine efficiency gain. An AI trusted to act unsupervised in a function with irreversible losses is the cum-ex lesson waiting to be relearned with a new tool.

Five myths that cause real losses

Some of the most expensive errors in corporate actions come from beliefs that sound reasonable and are wrong. Naming them directly is worth more than any amount of process documentation, because each myth, left unchallenged, points straight at a loss.

Myth one: the settled position is who holds the security. It is not. Pending trades, lent stock and short positions mean the party who appears to hold on the record date is often not the party economically entitled. Believe the naive position and you pay the wrong holder and miss the claim.

Myth two: the same event can be processed the same way in every market. It cannot. Ex and record conventions, defaults, rounding and tax all vary by market, and a global holder touches several at once. Hard-code one market's rules and you silently mishandle every other.

Myth three: a split and a stock dividend are the same thing. They are close enough to look identical and different enough to be processed differently — the GameStop case turned on exactly this. The golden record must capture the precise nature of the event, not just the ratio.

Myth four: no losses this year means the process is sound. With a fat-tailed loss distribution, a quiet year can hide a process one complex event away from a large loss. Judge the machine by its process metrics, not by an empty loss column.

Myth five: automation removes the need for expertise. It removes the need for expertise on the routine ninety percent and concentrates it on the risky ten. The complex tail becomes more dependent on scarce human judgement as the routine is automated away, not less.

5

  1. What is the organising principle of corporate actions control?

  2. What did GameStop's 2022 “split” teach?

  3. Why is “no losses this year” weak evidence?

  4. Where must controls sit, per the Citi/Revlon case?

The module in six ideas

Six ideas carry the whole subject, and everything else is detail hung on them.

Idea one — the event type defines the platform, not the volume.

A function built for the millions of routine dividends will still fail on the rare merger; the hard tail, not the easy bulk, sets the requirements. Design for the reorganisation and the dividend takes care of itself.

Idea two — data cascades, so errors are never contained.

The golden record is the atom, and every entitlement, instruction, payment and report is a transformation of it. An error at the top reaches every holder, which is why the correctness of the golden record is the precondition for everything else and why the controls cluster between it and the payment.

Idea three — the election deadline is where losses are actually made.

Capture and payment are demanding but forgiving; the immovable deadline is not. A missed or wrong instruction there is irreversible, and the behaviour of the holders — hedge funds pushing to the cut-off, retail not responding at all — determines how much of that risk the bank carries.

Idea four — controls are core architecture, not an afterthought.

Because mistakes cannot be undone, dual-source capture, four-eyes on the golden record, independent entitlement checks and the reconciliation gate are structural, not optional. The messaging standards and the resilience that meets the deadline are part of the same architecture of correctness.

Idea five — the fat tail quietly distorts investment.

Because most events are trivial, the function looks cheap and gets under-invested, right up until a single complex event produces a loss that dwarfs a year of running cost. The distribution, not the average, is what should drive the budget.

Idea six — real transformation moves cost or risk.

Standardising to ISO 20022, sourcing the event once, automating capture and claims, and using AI on the data problem all change economics or risk — and each must be governed, generative AI most of all, because its failure mode is the loss the whole function exists to prevent.

These six ideas are also a checklist for assessing any corporate-action operation.

Ask whether the platform was built for the hard tail or only the easy bulk; whether the golden record is genuinely one trusted source or several disagreeing ones; whether the election process reliably reaches holders and hits deadlines; whether the controls sit before payment or only report after it; whether the budget reflects the fat tail or the trivial average; and whether the transformation spend is changing economics and risk or only appearances. An operation that answers all six well is sound; one that answers any of them poorly has found where its next loss will come from.

  1. Why is "no losses this year" a weak signal of a good operation?

  2. Which risk type carries most of the catastrophic losses, and why?

  3. What distinct failure mode does each of the three cases illustrate?

  4. Give the filter that separates real transformation from decoration.

  5. Where does generative AI genuinely help in corporate actions today?

  6. Why must a human stay in the loop for AI-proposed golden records?

Process Cheat Sheet — The Whole Event on One Page

A single reference for the entire corporate-action flow, from the issuer's announcement to the client's confirmation.

The lifecycle at a glance

Stage Timing Core activity Key output Lead and systems
1. Announce Event start Issuer declares terms and dates Public terms Issuer / agent
2. Capture Days Scrub many sources to one truth Golden record Analysts; capture platform
3. Entitle Record date Snapshot position × terms Per-holder entitlement Operations; entitlement engine
4. Elect To the deadline Notify, collect, default, aggregate Instructions submitted Client services; workflow
5. Pay Pay date Move, withhold tax, reconcile, confirm Cash and stock moved Operations; settlement

Numbers and rules of thumb

Metric Rule of thumb
The core equation Position × Terms = Entitlement, on the dates the market sets
Event classes MAND (no choice), VOLU (must elect), CHOS (choose the form)
Settlement cycle (Europe) T+2 today; T+1 from 2027, compressing every window
Ex vs record Buy on or after ex-date and you do not get the event
Fractions Rounded down, paid as cash-in-lieu on a market-specific rule
Bank cut-off Set days before the market deadline to allow aggregation
Notification message MT564 / seev.031; instruction MT565 / seev.033; confirm MT566 / seev.036
Golden record Two independent sources must agree, or the break is investigated
Reconciliation gate Never pay clients more than received; no payment on an open break
Where losses cluster Voluntary events, missed deadlines, complex reorganisations
Cum-ex / cum-cum scale Estimated to have cost European treasuries more than €55 billion

The whole event in one sentence

Take the holder's position on the record date, apply the issuer's terms to get an entitlement, collect any choice before an immovable deadline, then move the cash or stock, withhold the right tax, and reconcile before you pay — because once it pays, it cannot be undone.

Capstone Exercise — Process ACME End to End

The surest test of the module is whether you can now run an event yourself. Work the case below before reading the answer. It uses the ACME rights issue, with one new twist, and touches every stage of the lifecycle. All figures are illustrative only.

The brief

ACME Industries SA (ISIN FR0000000001) announces a one-for-five rights issue at €42.00 per new share, voluntary, with the rights lapsing as the default. The key dates are: ex-date 4 March, record date 5 March, market election deadline 18 March, pay date 25 March. Your client holds 12,343 shares on the record date — note the odd number. The client has lent 500 of those shares to a borrower over the record date. A vendor feed shows the subscription price as €4.20, while the issuer's own offer document and the depository feed both show €42.00. The client, when notified, elects to exercise in full.

Work through, in order: what you do with the price discrepancy; the record-date position you use; the entitlement, including any fraction; the cash the client owes if they exercise; what the lent shares require; and what has to be true by 18 March and by 25 March for this to end cleanly.

The worked answer

The price discrepancy is a capture break, and it stops the event. Two independent sources agree on €42.00 and one vendor shows €4.20 — a classic decimal error. Dual-source validation flags it, the event does not flow straight through, and an analyst confirms €42.00 against the primary offer document before the golden record is set and verified under four-eyes. Had you trusted the vendor, every entitlement downstream would have been wrong by a factor of ten.

The record-date position is 12,343 shares, and the lent stock does not reduce it — but it does add a step. The client still owns 12,343 shares economically; 500 are simply out on loan. The lender remains entitled, so the 500 shares' entitlement must be manufactured back from the borrower, and the tax treatment of that manufactured leg checked. The naive "settled" view that the client holds only 11,843 would under-entitle them.

The entitlement carries a fraction. 12,343 divided by 5 is 2,468.6, so the client is entitled to 2,468 whole new shares and a fraction of 0.6, which is rounded down and paid as cash-in-lieu on the event's rounding rule. If they exercise the 2,468 rights, they owe 2,468 × €42.00 = €103,656.00 and receive 2,468 new shares.

By 18 March the instruction must be in, aggregated and submitted; by 25 March everything must reconcile. The client's exercise must be captured, validated against the holding, and rolled up the chain ahead of the market deadline — remembering that the bank's own client cut-off falls days earlier. On pay date the 2,468 shares are credited and the €103,656.00 debited, the manufactured entitlement on the 500 lent shares is settled, the fraction is paid in cash, and the whole event is reconciled against what the agent delivered before any client payment is released. Miss the deadline and the rights lapse for value; miss the reconciliation and a break becomes a loss.