FOUR HUNDRED THOUSAND DOORS

FOUR HUNDRED THOUSAND DOORS

What your money can actually do, and how many ways it can do it

Article 1 of 8  ·  Series III of III  ·  Published 13 January 2027  ·  Analysis → Forecast → Recommendations

Series III opens. The first series asked why the old arrangement is coming apart; the second measured where that shows up in physical infrastructure. This one asks what replaces it at the level of the actor — a person, a business, an investor, a territory — and how to act inside it.

Eight articles, eight domains, one instrument. Section two sets out the instrument; the rest of the article runs it on the first domain.

1. Four Hundred Thousand Doors

In one recent financial year, British banks closed 453,230 customer accounts. A decade earlier the figure was 45,091. Roughly a tenfold rise in ten years, and most of it is not scandal — it is ordinary risk management, operating at scale and without much explanation.

The most-discussed case involved a politician, a private bank and a leaked internal document, and it produced a chief executive's resignation and a regulatory review. The case was unusual in exactly one respect: the customer found out why.

From 28 April 2026 the rules changed. Banks must give ninety days' notice before closing a current account, up from sixty, and must set out their reasons in writing. Customers can take an unfair closure to the ombudsman. Immediate closure remains available where financial crime is suspected, which is reasonable and which also covers a great deal of ground.

One clause in that reform is worth reading twice. The new rules apply only to accounts opened on or after 28 April 2026.

So the protection exists, it is real, and it reaches nobody who already had an account on the day it arrived. Several hundred thousand people a year are having a door closed under the old terms, and the fix that was built for them applies to somebody else's account.

2. What Series III Measures

Series II measured seven physical layers and found that the honest questions were not only about the world but about the relationship between an actor and the thing it depends on. Series III starts from that finding and turns it around.

The premise

For most of the last forty years, optionality was supplied rather than held. You did not need alternative payment routes, because the banking system provided them invisibly. You did not need to think about which jurisdiction you could operate in, because most of them were open. You did not need a second supplier on another continent, because the first one was reliable and cheap.

That supply is being withdrawn, unevenly and mostly without announcement. Not replaced by chaos — the second series found these systems absorb a great deal — but moved from the system's balance sheet to yours. The thing that replaces the old arrangement is not a new arrangement. It is a requirement that actors hold what the arrangement used to hold for them.

Series II measured how fast the world's decision space is shrinking. Series III measures how much of yours is left, and what it costs to keep.

The instrument

Four questions, asked of any option an actor believes it has.

The second question is the one that does unexpected work. An option with no carrying cost is not an option anybody is managing; it is a convenience that happens to exist, and conveniences are withdrawn without notice. Where this series finds a genuinely robust position, somebody is paying for it on purpose.

Four audiences, not three

Series II addressed individuals, business and capital. This series adds a fourth, because the domains it covers are ones where a municipality, a region or a small state faces the same question with different instruments: territory. Every article carries all four.

3. Running the Instrument on Payment Access

The first domain is the one that converts everything else into action. Not wealth — access. The speed and reliability with which what you hold becomes what you can do.

Read that column down and the position is unusual among the domains this series will cover. The cost of holding an alternative is genuinely low. The control is almost entirely external. And the count is the quantity most people are confident about and most often wrong on.

Why this is the first article

Because every other kind of optionality resolves through this one. A second supplier you cannot pay is not a second supplier. A jurisdiction you can enter but cannot bank in is not an option. A reserve of capital that takes three weeks to move is not a reserve on the timescale most decisions run on.

Payment access is the conversion layer between holding something and doing something, and the instrument above is harshest on exactly the two questions — count and control — where confidence is highest.

4. Subtheme One — The Count Collapses Under Inspection

Ask a mid-sized business how many ways it can move money internationally and the answer is usually a number between two and five. Trace those routes and the number falls.

The network behind the network

Cross-border payments do not travel directly. A bank without a presence in the destination country uses a correspondent — another bank that holds accounts for it and settles on its behalf. That network is the actual infrastructure, and it has been shrinking continuously for fifteen years.

Relationships fell about 22 percent between 2011 and 2019, and roughly 30 percent between 2011 and 2022, including a 4 percent drop in 2022 alone. The striking feature of the series is not the rate. It is that there is no year in it where the network grew.

The decline is also concentrated rather than spread. Advanced economies lost around 23 percent on a country average. The Caribbean lost about 52 percent, Polynesia 54 percent, Melanesia 63 percent. A regional body put the Pacific's loss at roughly 60 percent since 2011, about double the world figure.

The actors who had fewest routes to begin with lost the largest share of them, which is the general shape of this series and will recur in every article.

What this does to a count

Two banks that both clear dollars through the same correspondent are one route wearing two names. A business with accounts at three institutions in one country, all of which de-risk the same destination because they read the same guidance and buy the same screening models, has one position and three invoices.

This is the convergence pattern Series II found in cables, magnets and satellites, arriving in a domain where it is even harder to see — because correspondent relationships are not published to customers and most banks will not tell you who clears your transfers.

The honest counterweight

Something genuinely improved in the same period. Payment technology diversified: instant domestic rails, card networks, specialist cross-border providers, stablecoin settlement at the margins. For a small payment to a common destination, there are more usable routes in 2027 than in 2015.

The caution is that most of those alternatives terminate in the same banking system, and the ones that do not are small, concentrated and themselves dependent on a handful of banking relationships. A payment provider that loses its own bank loses every customer at once, which is a correlation ordinary diversification does not address. And a global body reviewing progress in October 2025 concluded that the cross-border payment targets set for the end of 2027 are unlikely to be met.

5. Subtheme Two — The Option That Stays Legal and Stops Working

We expected the cash data to show a straightforward decline, and it does not. The actual finding is more interesting and it describes a failure mode that will recur throughout this series.

What the numbers say

In a euro-area survey conducted in early 2026 across all twenty-one member countries, 92 percent of companies with physical locations accepted cash, up from 90 percent in 2024. The report states plainly that the earlier decline has paused. Card acceptance was 88 percent; mobile payment acceptance had jumped from 36 percent to 68 percent in two years.

Of companies accepting cash, 92 percent expect to continue over the next five years. Six percent say they will not.

On those figures the option is intact, and anybody forecasting the end of cash on a five-year view is arguing against the data.

What is underneath the numbers

Two figures in the same survey point the other way, and they describe infrastructure rather than intention.

Among companies that do not accept cash, the share citing inconvenience in depositing or withdrawing it rose from 22 percent in 2024 to 35 percent in 2026 — the sharpest movement in the reasons given, and a statement about bank branches and cash-in-transit rather than about customers. And among companies operating self-checkout terminals, 48 percent accept cash at none of them.

The right to use cash is intact and the machinery for using it is thinning. Those are different things, and only one of them appears in the headline acceptance figure.

This is the failure mode worth naming at the start of a series about optionality. An option can remain fully legal, be formally offered, appear in every statistic as available, and still become unusable — because the supporting infrastructure that made it practical was withdrawn by parties with no obligation to maintain it and no reason to announce it.

Measured by the instrument in section two, this is a failure of the second question rather than the first. The count did not fall. The cost of holding the option rose, quietly, and was transferred from the institutions that used to carry it to the individual trying to use it.

6. What Most Analysis Gets Wrong

Four errors recur, and each changes a decision.

—  Counting balances instead of routes. A sum held in a form or a place from which it cannot be deployed at the speed a decision requires is not available in any sense that matters. Wealth and access are different quantities and only one of them is usually measured.

—  Counting institutions instead of infrastructure. Three banks sharing a correspondent, a screening vendor and a risk appetite constitute one route. The relevant number is almost never the number on the statements.

—  Reading availability as usability. Cash acceptance is up and cash handling is harder; both are true, and only the second determines whether the option works on a given afternoon in a given town.

—  Assuming notice. The default assumption is that access ends with warning. The actual mechanism in several hundred thousand cases a year is a letter, a period measured in weeks, and in some circumstances nothing at all.

7. Base, Stress and Extreme

Base — narrowing access, improving rails

De-risking continues at roughly the current rate, with closures concentrated in categories banks find expensive to assess rather than in categories that are actually risky. Payment technology keeps improving for ordinary transactions. The two trends run past each other: more ways to pay, fewer institutions willing to hold the relationship underneath. Most people never notice, and the minority who are closed out have a worse experience each year as the alternatives thin.

Stress — the category widens

A compliance category expands to cover a population that did not previously attract attention — an industry, a nationality, a pattern of transactions, a jurisdiction of residence. Closures arrive in a cluster rather than individually, and the affected parties discover simultaneously that their alternatives shared a correspondent. The consequence is not loss of savings; it is weeks or months during which a business cannot be run and a household cannot be paid.

Extreme — the conversion layer fails for a region

A small state or territory loses its remaining correspondent relationships, as several have come close to doing. Money continues to exist and stops being able to move: remittances reroute through informal channels at higher cost, trade finance becomes unavailable, and an economy that was functioning is cut off from settlement without any sanction being imposed. This has happened at the margins and remains the clearest case of the decision space closing for an entire population at once.

8. Forecast — One Year, to end-2027

Direction. At least one further G7 jurisdiction announces or enacts notice-and-reasons requirements for bank account closure comparable to the British rules of April 2026. Probability 0.55. Confidence: Medium.

The template now exists, it is modest, and the political appeal is unusual in that it crosses the normal divisions — the British version was drafted by one government and enacted by another. Debanking has also become a live political question in the United States, with federal agencies issuing warning letters to the payment industry during 2026. What holds the probability near even is that American activity so far has run through enforcement and executive action rather than through a notice requirement, which is a different instrument.

Second-order effect. Notice periods change behaviour more than they change outcomes. A bank required to give ninety days and a written reason has an incentive to decline relationships at onboarding rather than terminate them later, which moves the exclusion earlier and makes it harder to see and harder to appeal.

What weakens it. Financial regulators are occupied with other things, and account closure is a consumer-protection matter in jurisdictions where that sits with a separate body. The reform could equally arrive as supervisory guidance, which would not resolve this forecast.

9. Forecast — Three Years, to 2029

Direction. The number of active correspondent banking relationships recorded in international settlement statistics is lower at the end of 2029 than at the end of 2026. Probability 0.75. Confidence: Medium-High.

Fifteen years of data with no year of growth is a strong base rate, and the drivers have not changed: compliance cost per relationship keeps rising, the revenue from thin corridors does not, and the penalty for getting one wrong is asymmetric. Reversing this requires either a change in how compliance liability is allocated or a decision by large banks to carry unprofitable relationships for systemic reasons, and neither is visible.

Second-order effect. Each withdrawal pushes volume toward the remaining providers, which raises their concentration and their own compliance exposure, which raises the cost of the next marginal relationship. The mechanism is self-reinforcing, which is why the series has no year of growth rather than a gentle oscillation.

What weakens it. Reporting lags and definitional changes are the real risk to this forecast rather than the underlying trend. If the published series changes basis, or if new payment arrangements are counted as correspondent relationships, the measure could rise while the thing it measures does not.

10. Forecast — Five Years, to 2031

Direction. At least one G7 country enacts a statutory right of access to a basic payment account that cannot be withdrawn on commercial grounds alone. Probability 0.45. Confidence: Medium.

Below even, and the reasoning is about where the burden falls. Basic account access rules already exist in parts of Europe, and strengthening them into a right that survives a bank's commercial judgement is a different step: it obliges an institution to hold a customer it has decided it does not want, for reasons it is not required to accept. That is a significant intrusion into private contracting and it is resisted accordingly. What makes it plausible at all is that the alternative — a growing population with no banking access in an economy that is functionally cashless for many purposes — produces a political problem that eventually demands an answer.

Second-order effect. A right of access that banks must honour would most likely come with a designated provider of last resort, which is how the equivalent problem was solved in insurance and in postal services. That changes the architecture rather than the rules, and it is the shape to watch for rather than the right itself.

What weakens it. Financial inclusion is usually addressed through obligations to offer a basic product rather than through a right that overrides commercial discretion, and the softer version is both cheaper and easier to legislate. A basic-account duty with wide exemptions would not resolve this.

11. Forecast — Ten Years, to 2036

Direction. In at least one euro-area country, the share of companies with physical locations accepting cash falls below 75 percent. Probability 0.40. Confidence: Low-Medium.

Set deliberately against our own expectation going into this article. The 2026 data shows acceptance rising to 92 percent across the euro area and the decline paused, with national figures ranging from 99 percent in Greece and Italy down to 76 percent in Cyprus. Cyprus is therefore already close to the threshold, and 51 percent of its small firms say they may not accept cash in future, which is the single most striking national figure in the survey.

The probability sits below even because the infrastructure thinning we describe in section five operates slowly, because several governments are legislating in the opposite direction, and because the euro area has an institutional attachment to cash acceptance that is unlikely to weaken quietly.

Second-order effect. If one country crosses it, the argument changes everywhere from whether cash will remain usable to who is obliged to make it so, and the answer determines whether the carrying cost of that option sits with banks, retailers or the state.

What weakens it. Mandatory acceptance legislation is spreading rather than retreating, and a legal duty to accept cash would hold the measure above the threshold regardless of whether handling it remains practical — which would make this forecast resolve no while the underlying usability continues to fall.

12. Signals to Watch

Five observable items, each published, each moving before the outcomes above.

—  Correspondent banking relationship counts and active corridors in international settlement statistics. The single best measure of whether the conversion layer is narrowing, published with a lag and almost never read outside the industry.

—  Account closure volumes and the reasons recorded, where regulators collect them. The British figures exist because a parliamentary committee asked; most jurisdictions do not publish, and the appearance of such a series elsewhere is itself a signal.

—  Cash deposit and withdrawal infrastructure — branch and ATM counts, cash-in-transit coverage, and what retailers say about handling rather than about accepting.

—  Whether payment providers disclose their own banking dependencies. A fintech that names its settlement bank is offering information its competitors are not, and the absence of that disclosure across the sector is the correlation nobody is pricing.

—  Onboarding refusal rates, where any regulator begins to collect them. If notice requirements push exclusion from termination to refusal, this is where it will show first, and currently nobody measures it.

Read together: the first three measure whether routes are narrowing, the last two measure whether the narrowing is being moved somewhere it cannot be seen. The second is the more likely development and the harder one to detect.

13. Recommendations — Individuals

Immediate — 30 days

Count your actual routes rather than your accounts. Two accounts at institutions in the same country, under the same regulator, running the same screening, are one route. The question is how many genuinely independent ways you can receive income and make a payment if one stops working tomorrow.

Build — 12 months

Hold a second route that works and keep it alive. An account at a different institution, ideally under a different regulator, used occasionally rather than left dormant — dormant accounts are closed, and an account that has not moved in two years is not an option. This is the carrying cost from section two made concrete, and it is small.

Keep enough outside the banking system to cover a few weeks. Not as a hedge against collapse, which the evidence does not support, but against the ordinary and reasonably common event of an account being frozen or closed while a decision is reviewed.

Position — 3 years

If your income or your residence crosses a border, treat payment access as a real factor in those arrangements rather than an administrative detail. The populations losing access fastest are the ones whose transactions are expensive for an institution to assess, and that is a function of pattern rather than of conduct.

Avoid. Treating this as a reason to move money somewhere exotic. The failure mode here is loss of access to an ordinary route, and the answer is a second ordinary route, not an unusual one that attracts exactly the attention that causes closures.

Why this works. The carrying cost of a second banking relationship is among the lowest of any option in this series, and the control question says the decision is not yours — which together make this the clearest case in the series for paying a small amount to hold something open.

14. Recommendations — Business

Immediate — 60 days

Map your payment routes to the settlement layer rather than to the relationship. For each corridor you depend on, establish which institution actually clears it. Most banks will not volunteer this and many will answer if asked directly, and the answer frequently collapses three relationships into one.

Then establish what notice you would get. Read the termination clause in each banking agreement, and note which of your accounts predate any notice reform in that jurisdiction. In Britain, accounts opened before 28 April 2026 sit under the old terms.

Build — 12 months

Hold a second settlement route in a different jurisdiction, active rather than standby, with enough flow through it that it would survive a dormancy review and a relationship manager who knows the business. The cost is measured in fees and a few hours a month, against a failure that stops invoicing entirely.

And ask your payment providers who banks them. A provider is a route only as long as its own banking relationship holds, and a sector where nobody discloses this is a sector where customers cannot see the correlation they are exposed to.

Position — 3 years

Treat payment access as a condition of market entry rather than an operational matter. Expanding into a corridor where correspondent coverage has been falling means accepting that the cost and reliability of getting paid there may deteriorate over the life of the investment, and that is a commercial judgement rather than a compliance one.

Avoid. Concentrating payment operations for efficiency without pricing what that concentration costs. Consolidating to one banking group is a real saving and it reduces the count to one, which is exactly the trade the second question in the instrument exists to make visible.

Why this works. Tracing to the settlement layer and asking providers who banks them are the two questions that convert a comfortable count into an accurate one, and both cost a letter.

15. Recommendations — Capital

Immediate — this quarter

For holdings whose business depends on cross-border flow, establish correspondent exposure rather than banking relationships. Concentration here is invisible in disclosure and material to operations, and it is knowable by asking the company.

Build — 12 months

Watch de-risking as a cost driver in corridors rather than as a compliance story. Where correspondent coverage thins, transaction costs rise, settlement slows and working capital requirements increase — effects that appear in margins and are attributed to almost anything else.

And distinguish payment businesses by what sits underneath them. A provider whose entire settlement runs through one bank has a single point of failure that its customer count conceals, and that structure is standard rather than exceptional in the sector.

Position — 3 years

The structural observation is that compliance liability is currently allocated in a way that makes thin corridors unprofitable to serve, and that every proposed fix — safe harbours, shared utilities, public settlement options — is a reallocation of that liability rather than a technical improvement. Where that reallocation lands determines which corridors have banking at all in a decade. This is an observation about a policy variable, not a recommendation about any instrument.

Avoid. Reading payment technology improvement as access improvement. More rails on top of a narrowing settlement layer increases convenience and does not increase the number of independent routes.

Why this works. Correspondent statistics, regulatory consultations and account closure data are public, slow and specialist, and in this domain they describe an operating cost that reaches company accounts two or three years later under another name.

16. Recommendations — Territory

New in this series, and this domain is where it matters most — a jurisdiction can lose its settlement access entirely, which no individual or firm within it can remedy.

Immediate — this quarter

Know your correspondent position. How many institutions outside your jurisdiction clear your currency and your trade, who they are, and how that number has moved over ten years. For small states this is an existential statistic and it is frequently not held anywhere in government.

Build — 12 months

Treat the relationship with remaining correspondents as a diplomatic and supervisory matter rather than a commercial one between private parties. The withdrawals are driven by compliance cost and perceived risk, and both respond to what a jurisdiction can demonstrate about its supervision — which is a thing a government can actually change, unlike most items in this series.

Position — 3 years

Build the evidence that makes you cheap to assess. Published supervisory statistics, a functioning registry, verifiable beneficial ownership data and a responsive regulator reduce the cost of holding a relationship with you, and cost is the mechanism driving withdrawal. Regional pooling arrangements, where neighbours share compliance infrastructure, are the other available route and the one most often discussed and least often completed.

Avoid. Responding to withdrawal by seeking alternative arrangements outside the mainstream settlement system. It is available, it is expensive, and it raises the assessed risk that caused the withdrawal — which makes the original problem permanent.

Why this works. Correspondent withdrawal is a cost decision rather than a judgement about a place, which means it responds to anything that lowers the cost of assessment. That is unusual among the exposures in this series, most of which a territory cannot influence at all.

17. What Would Change Our Mind

Each forecast carries its own weakening condition. Three developments would undermine this article as a whole.

—  Correspondent decline reverses. Fifteen years without a year of growth is the backbone of this article. Two consecutive years of expansion would mean the compliance economics have changed and the narrowing is a phase rather than a structure.

—  Alternative rails prove genuinely independent. We treat them as terminating in the same banking system. A provider operating at scale with settlement that does not depend on a correspondent relationship would be a real second route rather than a second interface.

—  Access turns out to be recoverable. Our case rests on closure being hard to reverse. If appeal and reinstatement rates under the new notice rules turn out to be high, the exposure is a delay rather than a loss and the carrying cost of a second route is harder to justify.

Jurisdiction and sourcing, recorded on the first row. This article is British and euro-area in its evidence, with American regulatory activity noted rather than measured. The account-closure series exists because one parliament asked for it; most jurisdictions publish nothing comparable, so the tenfold rise is documented in one country and assumed nowhere else. Correspondent banking data is the strongest material here — international, consistent and long-running — but the published series runs to end-2022, and we have treated the direction as continuing without newer figures. The cash material is a large and recent survey and is the best-sourced element. Figures carry their class in the table below.

The Series III Table — one row

One row established. The next is jurisdiction — where you are permitted to be, and what that permission is actually worth.

18. Bottom Line

Four hundred and fifty-three thousand accounts closed in one country in one year, against forty-five thousand a decade earlier. The reform that followed requires ninety days of notice and a written reason, and it applies only to accounts opened after April 2026 — so it reaches future customers and not the people it was written about.

That gap is the subject of this series. Not what you own, but how many independent routes exist between what you own and what you can do, what each costs to keep open, when it closes, and who decides.

On payment access the answers are uncomfortable in a specific way. The cost of holding an alternative is genuinely low — a second relationship, a minimum balance, occasional use. The control is almost entirely external. And the count, the number people are most confident about, collapses under inspection: correspondent relationships are down about 30 percent since 2011 with no year of growth in the whole series, and the actors who had fewest routes lost the most — around 60 percent across the Pacific against 23 percent in advanced economies.

The cash finding went the other way and is the more useful one. We expected decline and found acceptance rising to 92 percent in 2026, with the earlier fall paused. The option is intact. Underneath it, the share of non-accepting firms citing difficulty depositing and withdrawing jumped from 22 to 35 percent in two years, and nearly half of self-checkout operators take no cash at all. The right persists and the machinery thins, and only one of those appears in the headline figure.

Which is the pattern to carry into the rest of this series. Options rarely disappear by announcement. They are kept formally available while the infrastructure that made them practical is withdrawn by parties with no obligation to maintain it — and the carrying cost of using them quietly moves from the institution to you.

For anyone with a decision in front of them: count routes rather than accounts, trace them to the layer where they actually settle, and keep one alternative alive rather than dormant. It is the cheapest option in this entire series and it is the one most people discover they needed on the day they no longer have it.

Forecast record

Four forecasts, one per horizon, each with a threshold, a named source that will settle it, and a date fixed before the answer is known.

Correlation. One policy pair, one structural standalone, and one loose inverse link. The first and third share a parent cause in political attention to financial exclusion and move together, though the third is a far higher bar than the first. The second is driven by compliance economics rather than by consumer policy and is close to independent of both. The fourth bears weakly and inversely on the policy pair: a jurisdiction legislating access rights is also the kind of jurisdiction likely to mandate cash acceptance, which would hold the measure above the threshold. So the set is not four independent observations, and the weakest member is the fourth, which carries the longest horizon and the softest mechanism.

Directional statements elsewhere in this article carry no threshold and are excluded from the record.

Sources

THRIVE IN CHAOS

Signal Over Noise 

AI intelligence system with human editorial oversight. Analytical material, not investment, legal or tax advice.