Patrick McKenzie
Banks are often hard to use because we want them to be

Glossary

A strange thing: knowing what a SAR is can get you a SAR

Title 12 of the U.S. Code of Federal Regulations, section 21.11, contains a sentence that boils down to this: no bank, and no employee of a bank, may disclose a Suspicious Activity Report, or any information that would reveal that a SAR exists. Not even under subpoena.

McKenzie unpacks what that means. If the U.S. government itself brings its subpoena power to bear on a teller about a SAR, the teller is supposed to say nothing. "It is designed to be enforced against the interests of the United States of America!" Customers get less than the government does. So does the teller: to keep staff from breaking the rule every day, functioning banks have long since walled SARs off in a subsystem seen only by the people who draft them.

That subsystem then has to interact with every other system in the bank, which is "a site for extremely painful hilarity to happen." If a SAR is misfiled because the subsystem has the wrong idea of who owns an account, investigating the error would mean telling the customer they were investigated, which is exactly what you cannot do. Then comes the line that stops you short:

And because this is insufficiently Kafkaesque, at some financial institutions, you can get a SAR filed for knowing what a SAR is, because “advanced knowledge of anti-moneylaundering procedure” is a characteristic only of financial professionals and terrorists.Seeing like a Bank, 2023-11-07

The next paragraph: "You might think I am joking. I am utterly not joking."

That scene is Bits about Money in miniature. Something that sounds like a joke turns out to rest on a regulation in black and white. Behind the regulation is something society wants. To get it, the system has built an arrangement that makes no sense from outside and in which every step made sense from inside. For five years McKenzie has been taking such arrangements apart in public. Occasionally, when he reaches the bottom, he concedes: most infelicities have a logical explanation, "And sometimes the world is absolutely batshit insane."

Who he is

In high school he interviewed for an order-entry job at an office-supplies retailer. The interviewer used the word "dropship," realized it was jargon, and began to explain. He cut in: that's when you buy the item after the customer orders it and have your supplier ship it directly; your free-shipping offer doesn't apply, and the delivery floor needs an elevator or it's an extra $75. How did a high-schooler know that? "It's in your catalog, in the fine print after the order form." Why did he have their catalog? "I made it my business to know your business." He calls it the smartest thing he will ever do in a job interview.

The rest of his career reads like an extension of that answer. Straight after college he moved to Japan and spent six years working for Japanese companies, building Bingo Card Creator, a tool for teachers to print bingo cards, on nights and weekends; it let him quit his day job in 2010. He then built Appointment Reminder, a SaaS business, and eventually sold both. Then came six years full-time at Stripe, where he is now an advisor; along the way he helped build what he calls "America's shadow vaccine info infrastructure." He paid for college partly by working as a tier-one customer service rep, wrote "a few hundred" complaint letters to banks on behalf of consumers as a hobby, and has been debanked twice himself. In early 2023 he stepped back from full-time work at Stripe and made the newsletter reader-supported while keeping every essay free; over the following two years he moved his family from Japan back to the U.S.

Almost every essay opens with the same disclaimer: he used to work at Stripe, is now an advisor, and Stripe does not necessarily endorse what he says in his personal spaces. It appears so often that it has become a kind of signature.

What he's wrestling with

His first issue, in October 2021, was about American community banks, which sounds like a sleepy subject. In it he made a claim he has not let go of since.

Technologists keep asking why ACH, the U.S. bank-transfer network, stayed slow for decades, and conclude that banks are technically incompetent. "Close but no cigar," he writes. The big money-center banks have buildings full of programmers shaving microseconds off trade execution; running batch jobs twice a day doesn't intimidate them. The drag comes from thousands of small banks with no programmers of their own, dependent on a few core processors' expensive software, with "disconcertingly few testing environments" and several dozen incompatible regulatory regimes to answer to. When same-day ACH was rolled out, the community bank lobby argued loudly for delay so as not to fall behind competitively.

Why does the government protect these small banks? Because they work for it:

A bank branch is a retail point-of-presence for the SSA getting funds to a pensioner, for the IRS collecting payroll taxes, and for the DEA interdicting fentanyl, staffed and funded by the private sector.Community Banking and Fintech, 2021-10-22

The section heading is "The financial system is a policy arm". That is his root question: how much of the financial system's apparently stupid, cruel or inefficient behavior is really a side effect of jobs society has assigned to it? There are two standard stories. Silicon Valley's says banks are hidebound and incompetent, waiting to be disrupted. The progressive one says banks are greedy and need regulating. McKenzie buys neither. His starting point is a programmer's instinct: if thousands of fiercely competing institutions, all knowing their customers hate something, fail in almost exactly the same way, it isn't a bug in any one of them. It's in the requirements document.

He ran into the question in concrete ways. At Stripe he watched engineers spend a day embedded with Operations, then rush back to their laptops apologizing ("I can't believe we did that to you!") to fix bugs that had existed for months or years. Ops replied that long experience in finance had taught them Ops never gets software that isn't broken, and that complaining about it is like complaining about gravity. In Japan he once checked a U.S. retirement account from a Japanese IP address, as he had many times before; this time it got him a phone call and the account was restricted and closed. In the U.S., a bank asked why he was averaging two incoming ACH transfers a day. He explained that he sold software online. The bank agreed the business sounded legitimate, then gave him 30 days to move its money elsewhere.

At the time those episodes were annoying and baffling. He has since spent a career working out why they happen. Over five years the claim hasn't softened; it has become blunter. By 2026 it is the first sentence of an investigative piece: "The financial industry understands itself to be an arm of the government."

His ideas, told as he found them

1. Fraud isn't an enemy; it's a slider

The story begins with that cut line about the optimal amount of fraud. McKenzie knows it sounds glib, so he first shrinks the problem: just one kind of fraud, in which a bad actor steals a card number and buys things online, costing the world something like $10 to $20 billion a year. Then he says something harsher: this fraud is possible because the best minds in government, finance, payments and business got together and decided they wanted it to be possible.

His lever is a question: who pays? Follow the money and you get a waterfall:

That the waterfall ends at the merchant decides everything. If fraud losses are a business expense like rent or salaries, then the Director of Fraud runs a budget. What should the target be?

Zero?! Do you think the Director of Marketing desires to spend zero on marketing!? That would be an objectively silly goal. They would clearly be fired and replaced with someone who understands marginal returns.The optimal amount of fraud is non-zero, 2022-09-02

The reason is that every anti-fraud measure lands on fraudsters and honest customers alike. His example is absurd and exact: requiring a firm handshake before doing business is a genuinely effective anti-fraud measure, since international fraud gangs are too far away to shake your hand. For the same reason, so are most of your customers. Online versions are milder. Asking for a billing address (for address verification) costs conversions. Sites push you to create an account partly so they can concentrate fraud attention on risky first-time buyers. Allowing guest checkout is a decision to accept more fraud in exchange for marginal sales.

Then he swaps in a word with a different emotional charge: trust. All fraud is an abuse of trust; you could get fraud to zero by never trusting anyone. But trust is an enormously useful social technology. He adds a line only he would write: civilization has a fundamental vulnerability, in that any human can be trivially killed while sleeping, and huge social effort goes into making sure it almost never gets exploited. "God has, reportedly, closed all bug reports claiming that it is a feature and won’t be patched any time soon."

The core problem of payments is letting a business decide, within seconds, to trust a stranger on nothing more than a promise to pay later. Hence "A promise! Mere words!" Next comes margin. High-margin businesses (games, streaming, software, often 90%+) loosen up and eat fraud to win conversions; sellers of resalable, low-margin goods like Apple hardware or game consoles hold first orders for review. Between them lies a whole spectrum, and the spectrum itself is a good thing. He also points out something often missed: one reason fraud has grown over the decades is that legitimate commerce has exploded.

Then he pushes the argument somewhere uncomfortable:

We should, as a society, accept non-zero amounts of benefits fraud. We should accept non-zero amounts of cheating on taxes.The optimal amount of fraud is non-zero, 2022-09-02

He admits these tradeoffs are almost impossible to make in the open: who wants to be the politician in favor of benefits fraud, or the bank CEO who thinks they aren't laundering enough money?

The idea stands on two earlier pieces. One is The fraud supply chain (late 2021): fraud is not mostly impulsive amateurs but a professionalized industry, with HR departments, forums, certifications and headhunters. Carders who steal credentials are separate from cashers who turn them into money for the same reason legitimate finance slices risk and sells it to whoever wants it. He narrates this with a refrain that plays like a comedy routine: email lists assembled by "specialist market research firms (but evil)" using the work of "data scientists (but evil)"; marketplaces with "customer service teams (but evil)" running "reputation systems (but evil)" so that unsuspecting customers (but evil) don't get defrauded by fraudsters while defrauding. Along the way:

Amateurs talk strategy or tactics. Professionals talk logistics.The fraud supply chain, 2021-12-03

His wife once saw an ad offering about $1,000 to "recycle" a company you were closing down, and suggested he take it. "Honey, that’s a crime and the person buying the business is a criminal," he told her. When the police come looking for the stolen money, the name on the documents will be his.

The other is Finality does not exist in payments (early 2022). Finality, he argues, is a probability distribution, not a property. Even Bitcoin: about $70 million stolen in 2016 was partly clawed back in 2022. His summary: "If you and the United States federal government disagree whether a transaction is final, you are wrong." Wires get reversed too, after a breezy phone call between operations people at two banks and a "hold harmless" letter. Credit cards go further and sell reversibility as a feature:

They increase social trust between peers on the network by decreasing technical trust. This is akin to alchemy.Finality does not exist in payments, 2022-02-15

Put together, that is his picture of fraud: fraud is an industry, fighting it is a budget line, reversibility is a product feature, and only the three together made people willing to pay strangers on the internet. In the late 1990s it was a live question whether anyone could transact online at all ("Aren’t hackers just going to take all money exposed to the Internet?"). They didn't, but answering the question required money to be exposed to the internet in the first place, and card reversibility made that happen.

2. Banks are an unacknowledged police force

KYC and AML: beyond the acronyms (January 2023) is the most linked-to essay on the site, cited 17 times. It doesn't open with banks. It opens with traffic deaths.

You are a large nation with too many people dying on the roads, and speeding is the culprit. How do you regulate? A strict nationwide speed limit? Rural roads mostly carry freight, so capping speed there hurts commerce and costs more to enforce than it saves in lives; big cities have their own police you can't directly command. You'll probably land on stochastic management: rather than controlling every road and every foot on every accelerator, you create new "laws of physics" for road authorities so that speed falls statistically. He gives the imaginary statute the acronym Roads Obviously Aren't Done Speedily, a dig at financial regulation's love of acronyms. The real goal was never speed. It was fewer deaths.

Then he turns the card over:

the goal is not to achieve banks having good knowledge of their customers or to prevent money laundering. It is to stochastically manage crime and terrorism at the margins by requiring an oft-unrecognized policy arm, the financial industry, to implement their own stochastic management of their books of business.KYC and AML: beyond the acronyms, 2023-01-23

And the stranger corollary: KYC and AML don't have to be effective on their own terms to serve their purpose. He titles the section "Effects achieved without effectiveness."

How? He takes apart several mechanisms. First, the dumb-looking questionnaire actually catches some people. Compliance departments will tell you, aggravated, that they read plenty of "Purpose of Transaction: Arms Smuggling and Prostitution", and you'd also be surprised how dumb some criminals are. Second, and more important, it acts at a different point in the crime's lifecycle: the trap described above, where a crime that touches money eventually needs a bank, needs a lie, and turns into a provable felony. Federal indictments, he notes, follow the same template again and again: two pages on the crime itself, then ten pages reconstructing how the money moved and which lies were told about it. He has reservations, and says so: "Candidly, I have some concerns about how gamesmanship sometimes enters that practice."

The third mechanism is the sharpest: the law almost never requires specific acts; it requires a written program and adherence to it. Many people believe the law requires a bank to see your ID in person. It doesn't. It requires certain identifying information; how to collect it is written in the bank's own Customer Identification Program, which regulators read and don't object to. That is why you could apply for a credit card on a plane over the Pacific decades ago and have it processed by mail, while deposit accounts "had to" be opened in a branch. The regulations didn't require it; banks didn't think it worth the work and blamed the regulator. When the pandemic hit, many institutions suddenly found the combination of words that allowed online account opening, something that had been legal every day for decades. In a multistakeholder system, he writes, everyone thinks they are uniquely responsible for the good and a passive observer of the bad.

The logic of "society assigns goals to banks" produces one especially cold example. In surveys of the unbanked, a certain group of men keeps saying banks will take their money. They're basically right, because collecting child support takes priority over those men being bankable:

Almost nobody is comfortable saying that they intend this in as many words. But as a society, yes, we unquestionably intend this outcome.KYC and AML: beyond the acronyms, 2023-01-23

The last layer is culture. Compliance is a performance, and its tone is "ingratiating", like a courtier before a monarch who is not unreasonable but does have a reputation for executing insufficiently diligent courtiers. You may privately think the questionnaire is bureaucratic nonsense, but inside a regulated institution you must never say so. "I’ve never met anyone who takes genuine pride and interest in KYC qua KYC. But you definitely cannot say that out loud."

It all adds up to one picture: society wants many institutions to lay overlapping, porous filters over different slices of society, so that most interesting activity hits at least one working surveillance apparatus, without wanting every apparatus to run at full strength. The optimal amount of fraud is not zero, and the optimal amount of KYC compliance is not 100%. The logic of the regulation reproduces itself inside the regulated.

3. Banks lack object permanence

In November 2023 the New York Times reported a spate of banks suddenly closing customers' accounts. McKenzie's response, Seeing like a Bank, borrows its title from James C. Scott's Seeing like a State, the classic account of bureaucracies that can only see what their forms can capture.

The oddity he starts from: these bad experiences happen in nearly identical ways across thousands of banks that compete viciously and know their customers hate this. Closed accounts, Zelle fraud, debit card reissues, foreclosures: every time you call, the person on the line is like a bit player in Groundhog Day, remembering nothing you said or they promised. His diagnosis:

Banks are extremely good at tracking one kind of truth, ledgers. They are extremely bad at tracking certain other forms of truth, for structural reasons.Seeing like a Bank, 2023-11-07

Then he debugs the bank the way you would debug a distributed system. First stop: system boundaries. Bank systems accrete "like sedimentary layers", and after every merger two systems run in parallel for years. "Every time responsibility moves between subsystems... some percentage of cases will simply break." Security holes live at the same boundaries. He is his own example: he banked at First Republic, which Chase then bought. The Chase app can now tell that the same person has an account on the other side, which would have amazed bank CTOs ten years ago. "Chase paid tens of billions of dollars over the years to get to the point where one engineer could bang this in-app banner out in 30 minutes." Yet Chase still eagerly sends timelines for transitioning a home loan First Republic never wrote, while knowing nothing about the line of credit it actually extended.

Second stop: support tiers. He worked tier one himself to pay for college. Tier one exists because the median call isn't really a problem; the caller is unsophisticated about the account, about finance, and often about much else in life. Here he writes the most impolitic passage on the site: some customers don't understand why a $45 charge and a $32 charge would overdraw an account holding $70, and the bank will not succeed where twelve years and $100,000 of public schooling failed. That customer calls much more often than you do. So you have to jump through hoops to reach a decision-maker because the system believes, on voluminous evidence, that you are probably that guy until proven otherwise. Tier one is largely a read-only interface to money and passes along a tweet's worth of context; tier two can authorize small "accommodations"; tier three are professionals with careers, no longer managed on tickets closed per hour.

Third stop: the side doors are deliberate. Nearly every bank has an "escalations" path that skips the tiers and reaches people with authority. A reporter calling about a widow on the verge of an improper foreclosure gets routed there within minutes; a paper letter to the VP of Retail Banking or Investor Relations often works too (he wrote a few hundred as a hobby). Banks don't hate it when you find the side door; it was designed, and the password is often performing membership in the professional-managerial class. No bank can say it offers different service levels by education, class and capacity to bring power to bear. Every bank does.

Yet he doesn't condemn the tiers. Quite the opposite:

But if you want phone calls to the bank to be both answered at 2 AM and absolutely free, you want the tiering system. Society wants the tiering system. It is why a high school student with a paper route can open a checking account, get a debit card, and start buying things on Amazon.Seeing like a Bank, 2023-11-07

Only at the fourth stop, the SAR confidentiality rule, does he write that one-line paragraph about banks sucking because we want them to. The order matters: first he rules out incompetence, then malice, and what remains is a blind spot society chose. "Every future with 12 CFR § 21.11(k)(1) in it will include many Americans whose bank accounts are closed for no reason that can be disclosed to them. Many of them will have done nothing wrong."

A year later the analysis faced its biggest test. In late 2024 the venture capitalist Marc Andreessen, on his own podcast and on Joe Rogan's, argued that debanking was being wielded as a weapon by specific political actors, against individuals and whole industries, without due process, crypto first among them. McKenzie answered with roughly 24,000 words titled Debanking (and Debunking?), the title itself a statement of method: verify, and also debunk. He opens by declaring himself "somewhat notoriously" a crypto skeptic, then: "Arguments aren’t soldiers; the truth is the truth. The truth sometimes favors crypto advocates in this discussion, and where it does, I will cite sources extensively."

First he separates two things the word blurs: having an account closed and being refused one. "It’s the difference between getting divorced and being turned down for a date." Most people whose accounts are closed lose them over credit risk or operational cost, such as repeated overdrafts. Then he turns to the archetypal victim, the corner bodega. Sell money orders or do remittances on the side and you become a money services business, on every bank's high-risk list. He goes through FinCEN's own public trophy case of enforcement actions and finds what look like bodegas, fined $10,000 for lacking a written AML policy. Banks read the signal, and most simply won't serve such customers. The closure letter will say "business decision" and "final"; he calls it a "ritual object", which professionals know not to take literally.

Then the sharpest moral question: shouldn't there be a presumption of innocence?

We do not have this assumption embedded in our banking system.Debanking (and Debunking?), 2024-12-09

Lending has always worked this way: build probabilistic models from observable factors, minus those society forbids (race, and zip codes because they proxy for race). "Finance is not high school; 92% is not an A- anymore."

But he doesn't stop at defending banks. He sends the bill to government, without regard to party. During the Obama-era Operation Choke Point, the Justice Department pressed banks using what it internally called a "self-affecting" theory of liability, and the FDIC circulated a 30-item high-risk list running from scams to pornography to "a grab bag of things we dislike" (he names racism and... fireworks?); regulators also decided that banking gun sellers carried "reputational risk." He borrows a Japanese word, sontaku: the diligent subordinate anticipating the boss's wishes without being told. "You are a very, very bad Compliance professional if you aren’t constantly sontaku-ing your regulator." As for government-ordered freezing of individuals' accounts for political reasons, he says it has happened, in Canada: during the 2022 trucker protests, the government invoked the Emergencies Act and froze more than 200 accounts, reaching people who were not protest organizers.

And so when people say that targeted debanking can be used, even in a democracy, for arbitrary and capricious punishment of disfavored individuals on the political speech, laundered through the banking system, with no substantial procedural recourse, I agree with those people that that is a risk. We just watched it happen.Debanking (and Debunking?), 2024-12-09

As for the larger claim, that debanking conservatives for their politics is ordinary U.S. practice, he says: it is not. His argument is disarmingly plain. If Republicans really were at constant risk of losing their bank accounts, they would buy their sneakers only in cash, fundraising dinners would have burlap sacks next to the tablecloths, and you would have had an awkward conversation with a conservative friend about splitting a check on Venmo. That isn't the world you live in, and your own eyes are enough to rule it out. By his own account in the retrospective, the essay drew public praise from both a crypto VC and a former federal bank regulator.

4. Everything is credit, carried on a train full of checks

The last idea is lighter, but it welds the other three together. The Long Shadow of Checks (December 2023) starts from an innocent-sounding question: why is the standard American bank account called a checking account?

His answer is a history of path dependence. Checks made clearing a problem: two banks need two messengers a day, four banks need twelve, thousands of banks need... oh dear. Hence clearinghouses. Every later way of moving money between U.S. banks was designed around the capability to clear paper checks nationwide; ACH wears that history in its name. Today's settlement timelines descend directly from an era when data moved between New York and San Francisco by rail, as handwritten paper, processed by floors of clerks with mechanical adding machines.

Deeper still: a checking account is a credit product. When you pay for groceries by check, the store must decide whether to release your chicken on badly incomplete information. Why accept a promise contingent on unobservable future events? "Because systematically taking relatively small amounts of risk created immense value. Groceries prefer selling more groceries to selling less... Consumers prefer to eat chicken versus going hungry. Chicken do not get a vote." To make it work, the state backed checks with a threat: a dispute with your grocer over a check can land you in jail. The crime is called "uttering." In practice, an overdraft costs some people a $30 fee and others a criminal record:

If a grocery store loss prevention employee, a police officer, and a prosecutor each refuse to stop the process, NSF means you’re Now Somebody's Felon.The Long Shadow of Checks, 2023-12-13

From which follows a conclusion few people reach: card payments make the economy more efficient, and one way to measure it is that card users rarely go to jail. Part of what a card fee pays for is rewards, part is computers, and part is "not jailing poor people." And because the standard account is a checking account, opening one is a credit decision, which is why there is an industry database that exists mainly to blacklist people from checking accounts. "Creditworthy," he writes, sounds like a value judgment but is "just a prediction about the weather." Advocates who want to bank the unbanked are often simultaneously furious at banks for extending credit improvidently, without seeing that they are asking for two contradictory things.

The thread runs all the way to The gift card accountability sink in December 2025. The AARP runs ads saying that being asked to pay by gift card is always a scam. Of course it isn't, he says: gift cards are a payments rail, and whole businesses use them to serve people without bank accounts. So why does debit card fraud usually get reimbursed while gift card fraud doesn't? Because gift cards were deliberately carved out of Regulation E and much of AML "in the interest of facilitating commerce", with retail industry groups making exactly that case in comment letters. Add the fact that the cards are actually issued by program managers you've never heard of, and the big retailer that sold you the card will tell you, truthfully and unhelpfully, that it doesn't issue them.

That decision has a cost. Grandma sometimes pays it.The gift card accountability sink, 2025-12-19

The four ideas fuse into one: every bit of friction is a tradeoff, and every tradeoff has someone paying for it. McKenzie's work is finding that someone and naming them.

How he thinks

Ask who pays before anything else. Nearly every essay starts by walking the money: the card-loss waterfall, the split of interchange, the fact that "free" brokerage trading is subsidized by customers' cash parked at small banks (he notes that the popular belief that brokerages live on payment for order flow is "straightforwardly wrong for every brokerage except Robinhood"). Find who finally absorbs the cost and the shape of the system appears.

"Banksplain that offboarding?" In Debanking he tells the story of each of his own two account closures and follows each with that line, then explains the seemingly arbitrary decision with a single structural cause. It's his signature move: first grant that the experience was awful and that he didn't understand it either, then reconstruct it from inside as a chain of individually reasonable decisions. He believes most infelicities have a structural cause, without ruling out that the world is occasionally just insane.

Read the primary documents and quote them. He pastes whole regulatory paragraphs, cites page numbers in DOJ Office of Professional Responsibility reports, quotes Inspector General findings, and counts bodegas in FinCEN's enforcement list. His whole case on Choke Point comes down to: "I can read emails. They say what they say."

See institutions with a software engineer's eyes. Bank problems become systems problems: subsystem boundaries, state lost at handoff, read-only interfaces, the invariant a ticketing system enforces, two systems running in parallel after a merger. No system can answer every interesting question about a user, "that is formally undecidable in computer science." Ticketing systems look trivially simple to technologists; to him they are one of the major technological advances of recent decades.

State your confidence. On Choke Point he writes that gun sellers don't make particularly intense use of ACH pulls ("confidence: 99%, on general industry knowledge") and don't have particularly high dispute rates ("confidence: 95%, same"). He puts numbers on his judgments and separates what he read from what he heard and what he infers.

Declare your position, then speak. Stripe advisor, crypto skeptic, "definitely not a secret Warrenite," former unpaid advocate against debt collectors. He puts it all on the table, then writes the other side's position in a form they'd recognize: the closing section of Debanking is a long list of what crypto wants, written in crypto's own voice.

Use jokes as part of the argument. The "(but evil)" refrain, "Now Somebody's Felon," "Chicken do not get a vote," God closing the bug report. None of them is decoration; each compresses a mechanism into a shape you can't forget.

His advice always points to where the rules are written. Want to help vulnerable people? Read the comment requests of obscure regulators. Want a bank to fix your problem? Work out whether the fix is within tier two's powers or only available to a professional earning six figures, and route yourself to the right person.

Where he might be wrong

Explanation can slide into absolution. His signature move is reconstructing an absurd-looking thing as a chain of reasonable tradeoffs. It has enormous explanatory power, and a systematic tilt: almost every status quo ends up looking like something "society chose." The question is who "society" is. Gift cards were carved out of consumer protection after retail industry lobbying, which he calls "the ordinary and appropriate work of a democracy", with a hint of irony, but his conclusion is still that it was done "for good reasons." The people who make the choice and the people who pay for it (the bodega owner, the scammed grandmother, the immigrant without a driver's license) are often not the same people. The "we" in "we want them to suck" sometimes deserves to be broken out and named.

"Optimal fraud" takes the curve as given. The essay drew over 300 points and more than 200 comments on Hacker News. One commenter called it "a naïve cost-benefit analysis" that ignores "the fact that a non-trivial amount of fraud affects individuals"; another said this kind of thinking in the payments industry was "infuriating", since cards already contain tamper-resistant chips and stolen credentials could be engineered away; another asked what happens if different technical measures let you jump to a point with less fraud and more revenue. European readers cited chip-and-PIN and phone confirmation. The critique has weight: he argues about where to sit on a given technical frontier, while the U.S. industry's long reluctance to adopt strong authentication is itself a "choice" open to criticism. In fairness, in Finality does not exist in payments he concedes the design space is "underexplored," and his 2025 gift card essay reads almost as a direct answer to the "fraud lands on individuals" objection.

The insider's view. He spent years at Stripe, remains an advisor, and works constantly alongside compliance teams, bank operations and regulators. That gives him knowledge outsiders lack, and a natural sympathy for institutions' constraints. He discloses it every time, and readers should keep it in mind: his explanations of why banks do things are almost always more detailed than his judgments of whether they should.

America-centric. Despite his many years in Japan and essays on Japanese payments, most of his analysis concerns peculiarly American arrangements: five thousand banks, interchange cap exemptions, checking accounts. Many conclusions, such as where the fraud slider should sit, may need recalculating under Europe's strong-authentication regime.

Crypto. He is an openly declared crypto skeptic. Debanking tried to be fair to both sides and, by his own account, was praised by both. This page has not read a systematic rebuttal from crypto advocates, so it cannot judge whether his account of the Silvergate and Signature failures has been seriously challenged.

The harshest thing he says about his own method sits in the SAR passage itself:

Most of the times infelicities in the world have a logical explanation to them, a structural cause where each individual link in the chain sounded good at the time and the result just happens to be suboptimal. And sometimes the world is absolutely batshit insane.Seeing like a Bank, 2023-11-07

That is the boundary of his method, drawn by himself: structural explanation is powerful, but it isn't everything.

Sources

• Community Banking and Fintech (2021-10-22)The first issue. Community banks, why ACH was slow, how the Durbin exemption funded a generation of fintechs, and "the financial system is a policy arm."
• The optimal amount of fraud is non-zero (2022-09-02)The loss waterfall, the Director of Fraud's budget, the handshake as anti-fraud measure, and the extension to benefits fraud and tax cheating.
• KYC and AML: beyond the acronyms (2023-01-23)The speed-limit parable, stochastic management, effects without effectiveness, lying to banks as a prosecutor's tool, ingratiating compliance. The site's most linked-to essay.
• Seeing like a Bank (2023-11-07)Banks without object permanence: system boundaries, support tiers, deliberate side doors, SAR secrecy, and "we want them to suck."
• The Long Shadow of Checks (2023-12-13)Clearinghouses, the checking account as a credit product, the crime of uttering, and why card payments keep people out of jail.
• Debanking (and Debunking?) (2024-12-09)About 24,000 words. Bodegas and money orders, the closure letter as ritual object, Operation Choke Point, Canada's freezing orders, and what isn't true. The author's own pick for his best piece of 2024.

Updates