Fifty-six hours, a hundred and four, or two hundred, decided by a thing the reader last thought about in 2021

1

The rail on the screen and the rail the money moves on

Tapping the wallet mark at a cashier does not send money from the wallet. It sends a substitute reference, which the cashier’s payment chain exchanges for an instruction against something else entirely: a current account, a debit card, or a prepaid card loaded in advance. The substitution is the whole product and it works beautifully in the direction of travel it was built for.

Coming back, the substitution falls away. Money is returned to the instrument the substitute stood in for, by whatever means that instrument is ordinarily paid, so the payout clock a reader should be reading is not the wallet’s. It is the clock belonging to a thing most people set up once and have not looked at since.

Thirty-six hours is what this note posits for the part that is the same whatever sits underneath: twenty hours for a desk to consider the request and sixteen for an identity examination. Everything after those thirty-six hours is a function of the answer to one question, and the answer is not printed anywhere a reader is likely to look.

36 posited hours are fixed, and the rest is not

None of the lengths in this note was measured, clocked or taken from a published schedule. They are declared so that one question can be priced: how much does it matter which instrument the substitute resolves to. The answer, worked through parts three and four, is 144 hours of spread, which is six days.

2

Four things to settle before a wait can be predicted

Prediction needs resolution, and resolution is four questions long. Every one of them has an answer available in a minute or two, and between them they turn a shrug into a figure. Until every one is settled, the only honest statement about the wait is that it lies somewhere inside a six day band.

Question two is the one nobody expects to matter. A reader who knows perfectly well that their wallet is pointed at a card may still not know whether that card draws on a current account or sits on a stored balance, and the two are paid back along different roads at different speeds. Sixteen working-day hours separate the first from the second on the posited legs below, and eighty separate the second from the third.

Question four is the one with the largest tail. Where the terms say money returns to whatever funded the deposit, the reader has no say in the destination, and part five costs what happens when that destination is wrong. Where the terms allow a nominated account instead, the reader chooses the leg and can choose the quick one.

Four questions, two of them answered inside the wallet, one at the cashier’s saved details and one in its terms. Four minutes of looking, on a generous estimate, against a band 144 hours wide. No other move in a payout queue pays back at that rate, and this one has to be made before a penny is deposited rather than after.

3

Fifty-six, a hundred and four, two hundred

One request, made at nine on a Monday morning, resolved three different ways. The thirty-six posited hours of holding run continuously and finish at nine on Tuesday evening, so the first working day available to any leg is the Wednesday. From there the three legs diverge, and the divergence is the page.

One hundred and forty-four hours separates the top of the first three rows from the bottom, which is six days, and nothing about the request changed between them. The same sum, the same morning, the same desk, the same thirty-six hours of holding. What moved was a setting in a wallet that the reader very likely chose by accepting a default.

Two hundred pounds held across those three totals costs 3.8 pence, 7.1 pence and 13.7 pence at an assumed three per cent a year. Those are the money figures and they are correctly tiny. The useful figures are the other ones: two and a third days, four and a third days, and eight and a third days, each of them a different answer to the question of when this is over.

4

A mean of 120 hours that describes nobody at all

Fifty-six, a hundred and four and two hundred add to 360, so the mean of the three is exactly 120 hours, which is five days. It is a clean number and it is useless as a prediction, because not one of the three outcomes is anywhere near it. A reader quoting 120 hours is quoting a figure nobody will experience.

Sixty-four hours below, sixteen below, and eighty above: those are the three distances from the mean. Averaged without signs they come to 53.33 hours, so an unresolved reader guessing the mean is wrong by more than two days on average. Resolution takes that error to nought, which is why part two is four questions rather than a shrug.

Averages behave this way whenever the thing being averaged has a small number of widely separated outcomes, and a payout leg is exactly that. There is no continuum of clearing speeds between one working day and five. There are three roads, and a reader is on one of them.

Thirty-six hours is the only part of the chain where an average means anything, because it is the only part that does not depend on the resolution. Thirty per cent of the quickest total is holding, 34.6 per cent of the middle one and eighteen per cent of the slowest. The fixed part shrinks as a share precisely when the total grows, which is the arithmetic signature of a variable dominating a sum.

53.33 h of average error from guessing the mean

5

Back to source, where source has since been replaced

Money goes back where it came from is a sensible rule and most cashiers of any description operate some version of it. Applied to a substitute reference it has a failure mode that an ordinary card payment does not, and the failure mode is quiet, which is what makes it expensive.

A substitute reference outlives the thing it refers to. Cards get reissued with new numbers, accounts get closed, a wallet gets pointed at something new, and the substitute can go on working across all of that because keeping it working is the point of it. The record a cashier wrote down at the moment of deposit is a different object, and nothing obliges it to keep pace.

So the two sides can disagree. Deposits keep succeeding, because they travel through the substitute. A payout aimed at the recorded instrument can be aimed at something that no longer accepts money. Nobody has done anything wrong, no term has been breached, and the payment has somewhere wrong to go.

One further asymmetry is worth stating. A deposit that fails is visible in about eight seconds, because a reader is standing there watching it. A payout that fails is visible when somebody notices it has not arrived, which on a leg of three working days means at the earliest 104 hours after the request and realistically a good deal later. The slow direction is also the blind one.

6

Two hundred and eighty-eight hours to recover one payment

Pricing the recovery needs four posited lengths on top of the ones already used. Three working days for the first attempt, which fails. Five working days before the failure comes back to the desk. Twenty-four hours of correspondence to name a replacement, and eighteen hours for the replacement to be checked. Then three working days again.

Followed through from a request at nine on a Monday morning, that chain finishes at five in the afternoon on a Wednesday sixteen days later, which is 392 h. Against the clean debit card case of 104 hours from the identical request, the recovery costs 288 hours, and 288 hours is twelve days exactly.

Row three carries the damage. One hundred and sixty-eight hours during which the payment has been sent, has not arrived and is not yet known to have failed is a week in which there is nothing to do and nothing to chase, because the desk believes the money has gone and the reader believes it is on its way. Nobody is lying and the week is lost all the same.

Two hundred and eighty-eight hours is 86,400 presses at twelve seconds to a press, assumed here and never observed, and 785 sittings of twenty-two posited minutes. At an assumed three per cent a year on £200 it is 19.7 pence of carry, against 7.1 pence for the clean case. Twelve and a half pence is the money difference and twelve days is the real one.

Three point seven seven times is what the recovery multiplies the clean wait by, 392 over 104. Set against the mean of part four it is 3.27 times, 392 over 120. Neither multiple is published anywhere, and neither can be, because it depends on a record inside one cashier matching a setting inside one wallet.

7

Which term each answer moves, and by how much

Four questions, four terms, and a different amount of hours riding on each. Setting them out together is the point of this part, because a reader with limited patience should know which question to ask first and the ordering is not obvious.

Seven minutes answers the lot, and the worst case they stand between a reader and is 392 hours. Divided out, that is 56 hours of exposure removed for every minute spent, which is the kind of exchange rate that does not occur anywhere else in a payout chain. The holds cannot be shortened. The legs cannot be chosen once the deposit is made. Only this can be done in advance.

Thirty-six hours of holding is the floor underneath every row and it does not move for anybody. Fifty-six hours is therefore the fastest this chain goes on the posited lengths, being 36 of holding and 20 of leg and calendar, and no amount of resolution beats it. What resolution does is avoid the other three rows.

One warning belongs with that table. The hours riding on a question are not additive, because the questions overlap: a reader who learns the record is stale has also learned something about whether a nominated account would help. The right reading of the last column is as a ranking, not as a sum.

8

A familiar mark at a cashier proves nothing whatever

Substitution runs on recognition. A reader sees a mark they know, infers that somebody respectable has looked at the desk behind it, and taps. That inference is the thing this part is about, and cashiers reached this way hold no British licence, so the inference has nothing underneath it.

The regulator has published the general version of this point about games rather than payments, and the general version transfers exactly. Games from licensed suppliers have turned up live on websites holding none, which means a game somebody recognises tells them nothing whatever about whichever desk is carrying it. A payment mark somebody recognises tells them nothing about the desk under it either, for the same reason: the recognisable party and the unlicensed party are two different parties.

Money has been attached to that. July 2026 produced a settlement of £4.75m from a supplier that does hold a British licence, after its games were found on six websites carrying none, every one of them reachable by consumers in this country, across a stretch to November 2024 in which the visit volumes were described as large. Three findings were made: how it assessed money laundering risk, how it controlled unlicensed routes into this market, and how it ran customer due diligence.

What that costs a reader is set out by the regulator in its own terms. Often absent from sites without that licence, on its listing, are checks on age, the tools a player uses to govern their own gambling, and payment systems that are secure. What consumers meet instead, on the same listing, is unfair gaming practice, losing money, and nobody to turn to once something has gone wrong. Secure payment systems is the entry that belongs on this page, because the resolution step of part two means the reader’s real instrument is recorded at the desk and not hidden behind the substitute.

Not one hour in parts three to seven changes because of any of that. Fifty-six, 104, 200 and 392 are what the posited lengths give. What changes is whether there is anybody to ask when the 392 becomes 600, and whether the card details sitting in that record are somewhere a reader would have chosen to put them.

9

Who should look underneath before the first deposit

Everybody, and it takes seven minutes. That is the short version and the arithmetic behind it is in part seven: 288 hours of worst case against seven minutes of looking, which is 56 hours of exposure removed per minute spent, available only before the deposit and worth nothing afterwards.

Somebody whose wallet points at a current account should carry on without worrying. Fifty-six hours on the posited legs is the quickest total anywhere in this note, 36 of it holding, and a bank leg is the one resolution that cannot be beaten by changing anything else.

Somebody who does not know what their wallet points at is the reader this page was written for, and the honest statement of their position is a 144 hour band with a useless mean in the middle of it. Fifty-three and a third hours is how wrong they will be on average if they guess, and the guess is unnecessary.

Somebody who last changed a card and has not thought about it since is in the fourth row, and the fourth row is 392 hours. The deposits will keep working, which is the reassuring and misleading part, and the first sign of trouble will arrive about a week after the money should have.

One hundred and twenty hours is the number a reader should refuse to accept, because it is the mean of three outcomes and the property of none. Fifty-six, 104, 200 and 392 are the four real answers on the posited legs, and which of them applies is settled by four questions nobody has to ask anybody’s permission to answer.