Pushing Money or Pulling It: Two Ways to Pay Compared

One shape pushes: the payer instructs their own bank to send money out of their own account. The other pulls: the business's side asks the payer's bank for money, and the payer's bank decides and funds it before collecting anything. Everything else that separates the two, including who is exposed and where a dispute starts, follows from which direction the instruction travels.

One question asked of any payment makes the rest of it fall out. Who gave the instruction, and who was being asked? There are only two answers, and they are not variations on a theme. In the first, the payer speaks to the bank that already holds the payer's money, and says send this amount there. Nobody has to be persuaded of anything except whether the balance is enough. In the second, somebody on the business's side speaks to the payer's bank and says please give us this amount, against a permission the payer handed over at some earlier moment. The payer is not in that conversation at all.

Most people have been on both sides of this without naming it. Handing a shopkeeper cash is a push: the money leaves the payer's hand and the transaction is finished at the moment it lands. When the milk seller writes each delivery in a book and is settled with at the end of the month, the seller is the one who comes and asks, and what makes that work is not the money in the payer's pocket but the standing understanding between the two of them. The direction the request travels is the whole of the difference, and every other thing that separates a payment sent from a payment asked for is a consequence of that one direction rather than a second, separate fact.

Two shapes are enough to cover the ground. Every parameter that would tell two named arrangements apart is set by an authority and moves. The shapes generalise. Learning the two makes it possible to place a way of paying that has not been invented yet.

What is actually being instructed in each of the two ways to pay?

The names people use for these two shapes describe the direction rather than the words. Start with what the instruction literally says. In the push, the instruction reads roughly: I am the holder of this account, take this amount out of it and put it into that account over there. The instruction goes from the payer to the payer's own bank, and it goes nowhere else first. The bank receiving it is the bank that has the money, so the question in front of that bank is a question of fact. Is this instruction really from the holder, and is there enough in the account?

In the pull, the instruction reads differently and it starts in a different place. The wording runs roughly: this payer gave a permission earlier, here it is, please give us this amount against it. The request starts on the business's side and travels towards the payer's bank, and a request is not an order. The mandateA permission a payer hands over in advance, so that somebody else can ask the payer's bank for an amount later on without the payer being present when the asking happens. that sits underneath it was given at some earlier moment for a purpose described in general terms. The request arriving now has to be matched against it in a second or two by a bank that has never met the party doing the asking.

One shape sends money and the other asks for it, and once that difference can be heard in the wording of the instruction itself, the words push and pull never need explaining again. A card is one way of handing over that earlier permission and it is not the only one. Any arrangement where the payer agrees once and somebody collects repeatedly afterwards has the same shape underneath, whatever it is called and whatever it is made of.

Who is being asked, and why does that one answer produce every other difference?

Here is the row the entire comparison hangs from. In a push, the party being asked is the payer's own bank, and what it is being asked about is the payer's own balance. Few questions in banking are easier. The bank holds the money, the bank can see the money, and the person asking is the person whose money it is. There is a check to be made about who is giving the instruction, and how that check is done is settled separately and is not rebuilt here, but once it is passed the rest is arithmetic against a balance.

In a pull, the party being asked is still the payer's bank, and readers stop reading right there. The question put to that bank is a completely different question. The bank is being asked to act on a permission given at some earlier time, presented now by somebody it has no arrangement with, on behalf of a payer who is not in the room. The bank has seconds. The bank cannot ring the payer. The bank has to decide whether this request is the sort of request that permission was for, and whether it is willing to put the money up.

The pull shape has to carry a decision that the push shape never has to make, and every extra party, every shared rule and every dispute route in the pull shape exists to make that decision possible. The impression that a card payment is needlessly complicated dissolves right there. A card payment is exactly as complicated as the question it has to answer in two seconds without the payer present.

The household version is close at hand. Handing a neighbour a hundred rupee note needs no machinery at all. Telling a neighbour that the shop on the corner may put things on the household's account until further notice needs a shared understanding of what counts as that shop, what counts as a reasonable amount, what happens if the shop asks for something that was never taken, and who sorts it out when the two of them disagree. All of that apparatus is what a permission given in advance requires, and none of it is required when the note is simply handed over.

Who is exposed in each shape, and to what?

Readers arrive here assuming the two exposures are the same size pointing in opposite directions. The two exposures are not the same size. The exposures differ in kind, and one shape has an exposure the other simply does not create.

The push is short, so take it first. The money was the payer's, it was in the payer's account, and now it has gone. No party in the chain put up anything of its own to make the payment happen. The payer's bank moved a balance it was already holding. The payee's bank received something that had already been paid for. Nobody advanced anything, so nobody is waiting to be repaid. Plenty can still go wrong in a push, and what goes wrong is set out below. Without an advance, though, no party is carrying an open position while a fault is sorted out.

Now the pull, and go slowly. The payer's bank has frequently acted before it has collected anything from the payer. The bank has said yes to a request and stood behind that yes, and the collection from the payer happens afterwards, on a schedule that has nothing to do with the moment the business was told the payment was good. Acting first and collecting later is funding before collectionA bank acting on a request and standing behind its own answer before it has taken anything from the payer. The money it commits at that moment is its own until the collection catches up., and it is the single most under-appreciated fact about the pull shape. Meanwhile the business has often already handed over the goods, on the strength of an answer that is not yet final. The pull shape manufactures an exposure the push shape does not have, two parties are carrying something at the same moment, and almost all the machinery around a card payment that looks unnecessary is that exposure being managed.

Everything in that paragraph turns on finalityThe point after which a payment is not going to be undone by the arrangement that made it. Before it, an answer can still be revisited; after it, the money has stopped being anybody's open question.. Between the moment the business is told the payment is good and the moment it becomes final, somebody is carrying the difference. The earlier that moment of finality falls, the smaller the exposure; the later it falls, the larger. Where exactly it falls is set by the authority named below.

What does a dispute look like in each, and where does it start?

Two questions decide what a dispute looks like. Where has the instruction got to by now, and who does somebody take the matter to? Blame is not one of them: a payment that went somewhere it should not have is a thing that happens rather than a thing somebody did.

In the pull shape, the payer's bank made a decision. The bank looked at a request, matched it against a permission, and said yes. The yes is a decision by an identifiable party, taken on the payer's behalf, and a decision can be questioned as a decision. So the matter is raised with the payer's bank, and because the yes was given inside a shared set of rules that the business's side had already signed up to, there is an established route for taking the money back off the business's side afterwards. The route exists because the decision existed. Where there is a decision made on somebody's behalf, there is machinery for revisiting it.

In the push shape, nobody decided anything on the payer's behalf. The payer's bank carried out an instruction the payer gave. Carrying out an instruction is a different act entirely. So what is in front of everybody is not a decision to revisit but a payment that has landed where it landed, and getting it back means unwindingRecovering money that was correctly sent to a place it should not have gone. Unwinding differs from reversing a decision. No decision by anybody else exists to revisit. rather than reversing. The payer's bank acted, so the matter is still raised there, and the payer's bank then has to approach the bank that received the money. There is a route. The route runs between two banks rather than through a rule book both had already agreed to follow.

A decision that somebody else made can be disputed as a decision, and an instruction the payer gave has to be unwound instead, and that structural difference is the whole of what this part claims. It is not a claim that one route is faster, kinder or more likely to succeed than the other. Every step in both routes, and every period allowed at every step, is set by the Reserve Bank of India, moves when that authority moves it, and is left as an empty row below.

How many parties does one payment pass through in each shape?

Count them rather than asserting a number, and count functions rather than companies. One company can perform two functions, and two companies can share one. A push passes through five functions. There is the payer, who gives the instruction. There is the payer's bank, acting on it. There is the settlementThe later moment at which two banks square up what one of them owes the other because of an instruction. Which arrangements do that squaring up, and in what shapes, is covered separately. arrangement, in which the obligation between the two banks is discharged. There is the payee's bank, receiving it. And there is the payee.

A pull passes through all five of those and two more. The extra pair sit early, before the request ever reaches the payer's bank. One of them put the business on its books in the first place and now presents that business's requests for it, and it exists because the payer's bank will not take a request straight from a business it has no arrangement with. The other is the shared rule book, together with the arrangement those requests travel over, and it exists because the payer's bank and the business's side have never negotiated anything with each other directly and need somebody else's rules to have agreed in advance.

So the count is five against seven, a difference of exactly two functions. More functions means more places a payment can stop and more parties to be paid out of it. More functions does not mean the shape is worse. Those two extra functions are precisely what makes the decision described above possible at all. Strip them out and the pull shape does not become a leaner pull shape. The shape stops working. Nobody is left to present the request, and the payer's bank has no agreed rules to answer it under.

The count is a property of a shape, worked out from what each shape needs. The count is not a reported figure about anybody: the record behind it holds no count of the parties actually involved in any real payment anywhere. A party countThe number of separate functions a payment passes through. A party count describes the shape of the payment rather than measuring any particular payment or any particular set of companies. arrived at this way shows where to look; it does not report what anybody found.

Why do both shapes suit a system whose average payment is small?

Here the comparison meets the one arithmetic fact this guide carries. Setu Payments Limited carried 1,200 crore instructions in the stated year with Rs 3,60,000 crore of value on them. The division runs: Rs 3,60,000 crore, divided by 1,200 crore instructions, comes to Rs 300.00/- on the average one. Three hundred rupees an instruction says more about what such a system must be like than any description of it would.

The average has to mean that everything done once for each payment is nearly free in both shapes. Not cheap. Nearly free. There is no room inside Rs 300.00/- for a person to look at any individual payment, for a step that requires a phone call, or for a check that takes a human minute. Both shapes are built that way, and that is exactly why they feel the same in use.

Then the difference the average does produce. The decision a pull carries, and the two extra functions supporting it, mean the pull shape does more things once for each payment. A falling average payment therefore presses on the pull shape harder than on the push shape. That is a statement about direction and not about size. Which way the pressure moves is established; how far it moves is not. Seven functions have to be paid for out of the same payment that five functions have to be paid for out of.

Now keep the value carried fixed and move the average instead. Holding one and moving the other is the cleanest way to feel what is going on. Suppose the very same Rs 3,60,000 crore rode on payments averaging Rs 3,00,000/- each. The same value works out at 1.2 crore instructions, not 1,200 crore. One figure has gone up by a factor of a thousand while the other has come down by the same factor, and both moves follow from pinning the value in place. Up there, whatever gets done once for each payment sits on a thousand times as much value, so neither shape feels any of this pressure. Three hundred rupees an instruction and three lakh an instruction are not two dials on one business. The two are different businesses.

One warning about the average itself. A reader can be led astray here without noticing it happen. Rs 300.00/- is a quotientWhat remains after dividing one figure by a second. A quotient says how the two figures stand to each other, and says nothing about any single item counted inside either., one figure divided by another. The quotient describes how a year's value stands to a year's count, and it describes nobody's actual payment. Behind it sit exactly two numbers and no third. The record therefore holds no typical instruction, nothing about how widely the instructions scatter, and no biggest or smallest. Picture two very different systems. In the first, nearly every payment really is somewhere around three hundred rupees. In the second, a handful of colossal payments sit among an enormous number of trivial ones. Both arrive at the same quotient, and nobody who could see inside them would mistake one for the other.

How does somebody running a shop actually use this?

What does this comparison not settle?

A comparison is exactly where a reader expects a verdict, and there is none to give. Neither shape is the one a person or a business should use, neither ranks above the other, and neither does any application, arrangement or system built on either one.

The reason is structural rather than evasive: the two shapes differ in who is asked, who is exposed and where a matter starts, and which of those three differences matters depends entirely on the situation. Somebody paying a stranger once for something they will carry away immediately is in a different situation from a household setting up a payment that will repeat for two years. The same three differences point in opposite directions in those two cases. Picking one would be picking a situation and calling it a rule.

There is a second reason, and it is arithmetic rather than principle. Nearly everything anybody would actually decide on sits in the four rows below, and every one of those rows is empty because the values in them are set by an authority and move. Two things cannot be ranked on parameters that are not stated.

A capability remains instead of a conclusion. For any payment at hand, including one made in a way that did not exist when this was written, its shape can be worked out. From the shape follow who was asked, who is exposed, and where a matter starts if something goes wrong. A capability of that kind is durable in a way that a verdict about two arrangements as they stand today would not be.

Who sets every parameter that separates the two in practice?

Four conditions were drawn round above, and each is left with nothing written into it. Setting them belongs to the Reserve Bank of India rather than to any arrangement. First, the value an instruction of either shape may carry, both on one instruction and across a period. Second, when an instruction of either shape may be given, and when the money underneath it settles, including every cut-offA moment in a working day past which an instruction waits for the following batch rather than joining the one already in progress. that decides which batch it joins. Third, the charge that may be made for an instruction of either shape, and every ceiling on it. And fourth, the route by which an instruction of either shape is put right after it has gone wrong, together with what is allowed at each step of it.

Almost everything a reader would use to choose between the two shapes in a real situation sits inside those four empty rows. The shapes hold still while the values move. Knowing that a limit exists, and knowing who sets it, outlasts knowing what it was.

The block below is built so that another market becomes an addition to it rather than a replacement for it. Nothing above the block depends on any value inside it.