KYC: The Process, the Layers and What e-KYC Changed

Know-your-customer (KYC) is the check an arrangement runs before it will open an account. The check establishes two facts, who somebody is and where they can be reached, against evidence the arrangement will accept, and it keeps a record that the check happened. A threshold is all it is, and not a judgement. The electronic route changed how those facts may be established and left what must be established exactly where it was.

An arrangement that holds money for somebody has to be able to attribute the account to somebody, and it has to be able to answer for that account when an authority asks. Neither is possible if it does not know whose account it is. So before it will begin, it establishes two facts and files a record that it did so. Four questions unpack that single sentence: which facts, what evidence establishes them, how that evidence reaches the arrangement, and how long the filed record stays good. All four answers are sentences somebody wrote down, and not one of them is a measurement of anybody.

What is know-your-customer actually for?

The answer lives on the arrangement's side, so ask the question from there. There are two reasons and the second one is the one everybody quotes, so start with the first. An arrangement that cannot attribute an account to somebody cannot return the money in it, cannot act on an instruction about it, and cannot answer a question about it. Every ordinary service a deposit account provides depends on the arrangement being able to say whose it is and where to reach that person. Take that away and the account is not a lighter version of an account, it is an unusable one.

The second reason is the one that gets stated: an authority requires the check, and requires a record that the check was done. The requirement to run the check and to keep the record is real, and it binds. But an account that gives only the second reason leaves a reader thinking the whole thing is paperwork imposed from outside, and it is not. The paperwork sits on top of a working requirement the arrangement would have anyway.

The check is not an assessment of anybody. It produces no score, no rating, no opinion and no ranking. Completing it says exactly one thing: two facts were established against evidence the arrangement accepts. Not completing it says exactly one thing too, set out below, and that one is not about a person either. A locker room that will not hand over a key until it can attribute the locker to somebody is protecting whatever is inside the locker, and the rule is a fact about the locker room. Nobody standing at the desk has been evaluated.

What are the three layers of the check, and which one is which?

Almost every confused argument about this subject comes from treating one thing as three, or three things as one. Separating them once, properly, settles most of those arguments before they start.

Layer one is what must be established: who somebody is, and where they can be reached. Layer two is how it may be established: which evidence the arrangement will accept, and in what form that evidence may arrive. Layer three is how often it is done again. A record established once does not stay established forever, so the arrangement runs a periodic reviewGoing back over a fact that was already established, at set intervals rather than once, so that a record filed some time ago is confirmed as still current. at intervals and sooner if something about the account changes.

The first layer has barely moved in decades and the second has been rebuilt, and almost every argument about this subject is really an argument about the second layer wearing the first layer's name. When somebody says the requirement got easier, listen for which layer they mean. If the same two facts still have to be established, layer one did not move at all. The evidence route moved. The two layers have different owners, change at different speeds, and fail in completely different ways, so the distinction is neither small nor pedantic.

The contents of all three layers belong to the Reserve Bank of India, and they change when it changes them. A list of accepted evidence changes when the Reserve Bank of India changes it. A list that was right two years ago gets carried away and acted on, and that makes it worse than no list at all.

What can a check confirm, and what can it never confirm?

Here is the mechanism, and it is sharper than the word check suggests. A check does not investigate anything. It compares. Something presented is set against something already recorded, and the comparison returns one of exactly three things: a match, a mismatch, or nothing at all. The third return, nothing at all, carries the whole argument, and it repays slow reading.

Readers move past the consequence at speed, so it is worth stating slowly. A check can only ever confirm a fact that was recorded somewhere, by somebody, at some point. Where no record of the fact exists, the comparison has nothing to sit the details against, so it returns nothing rather than returning a wrong answer. The check has not failed. The check has reported, accurately, that this registerSomewhere a fact was set down and kept so that it can be looked up later. A list of who lives at which address is one. So is a list of births. holds nothing to compare against.

So here is the line: a missing record is a gap in a register. The gap is a fact about what somebody once wrote down and filed, and about the completeness of a filing system. The gap is not a fact about the person the record is missing for. A library catalogue with no entry for a book says something about the catalogue. Whether the book exists is a separate question the catalogue was never able to answer, and reading the empty line as evidence about the book is a mistake in reading, not a discovery.

What are the routes by which the same fact gets established?

There are three, named here by function only. Route one is in person against something presented: a document is carried to a counter, looked at there, and a copy is kept. Route two is electronically against a record already held: the details presented are compared against a record kept elsewhere, and the comparison returns its answer without anything physically travelling. Route three is by relianceOne arrangement accepting a check that a different arrangement already completed, instead of running the same comparison again from the start.: an arrangement uses a check already completed somewhere else rather than repeating it.

All three establish the same fact, and they differ only in what has to travel, how far it goes, and how long it takes. The comparison between the three routes ends there. Nothing about who somebody is changes depending on which of the three routes established it, and no route produces a stronger or weaker version of the fact. Logistics is all that differs, and logistics is where the cost lives.

Names are the least durable thing about any of these routes. Names change, the arrangements sitting behind the names change, and a treatment written around a name goes wrong in a way one written around a function does not. Function survives the rename. A reader who has understood route two understands every future version of route two, whatever anybody calls it next.

Two features of that control are worth noting. First, set the register to hold no record and move the route across all three positions. A route decides what travels and never decides what was filed, so the return stays at nothing at all on every one of them. No amount of improvement to the second layer creates a record that nobody ever wrote down. Second, set the register back to holding a record. Layer one is not on the control at all, so the top band stays put at every setting while everything else moves.

What did the electronic route change, and what did it leave exactly where it was?

The electronic route changed the second layer and not the first. Everything below is detail underneath that one line. The same two facts still have to be established. The record that the check happened is still kept. The change is in what has to travel: instead of a document and a journey to a counter inside that counter's opening hours, a comparison runs against a record already held.

Three things moved when that happened, and each of them is one of the barriers a counter puts in the way. The distance stopped being a journey. The time stopped being a queue and an appointment inside somebody else's opening hours. And the cost of delivery stopped being a counter's fixed cost divided by the uses passing through it. Dividing that fixed cost by those uses is the arithmetic that decides whether a point of serviceAnywhere a financial service can be reached from at all: a counter, a machine, an agent, an application. Whether one exists in a given place turns on what it costs to run against how many uses go through it. exists at all in a thinly used location.

Now the counterweight. An account that stops after the good half is selling something. An electronic route works where a record exists to compare against, so it moves the threshold from a journey to a register. That is a genuine move and it is not a small one. But a move is not a removal. Where the register holds nothing, the route returns nothing, instantly and at no cost, and instant is faster without being an answer. The barrier changed shape from distance into filing.

And here is the limit of what the material supports. The material carries one arrangement, one year, and no second period at all. The effect of the electronic route on the cost and the distance is a statement about mechanism; what it achieved would be a statement about outcomes. One arrangement in one year cannot test the second claim, and a claim made anyway would be a conclusion dressed as a finding.

What does the check cost, and who carries that cost?

Start with the absence, out loud. The absence decides how far the cost question can be taken. The cost of running these checks at Suvarna Commercial Bank Limited, an invented bank, sits inside a single line: operating expensesEverything an institution spends to run itself for the year, added into one line: staff, premises, technology, and the running of every process it operates. Reported as one figure unless the institution chooses to break it up. of Rs 5,040 crore for the year. The material carries no split of that line at all. The share of that line the check accounts for cannot be stated. A made-up share gets carried away and repeated as though somebody had measured it, and no made-up share is offered in its place.

Without the split, the shape can still be said, and the shape is the part that generalises. The check is mostly a fixed cost per accountA cost that is the same for every account regardless of how much money is in it, because the work behind it does not get bigger when the balance does. rather than a cost that rises with the balance in the account. Running a comparison takes what it takes whether the account holds a little or a lot. So measured per rupee held, the same cost lands very differently across accounts, and it lands hardest on the smallest ones.

The shape of that curve is a property of arithmetic and of nothing else. A fixed cost sitting on a variable base always weighs most on the smallest base, in every business anywhere, and the curve above would look the same for the cost of issuing a receipt or the cost of posting a statement. The curve is a fact about the cost structure of the check. The curve is not a fact about anybody holding a small balance, and the arithmetic does not need one, so no person is drawn on it. The same arithmetic is the reason a lighter route exists at all: where the cost of the full check is heavy against a small base, an arrangement that wants to serve that base needs a cheaper way to cross the same threshold.

The conditions riding on an account opened under a lighter check, and what such an account may then do, belong to the Reserve Bank of India, and they change when it changes them. A row below carries the name of the body that fixes them and a blank where the answer would go. A blank against a named address is the honest shape for anything that moves.

What can one bank's own figures say about the cost of a check?

Less than a reader would want, and saying so precisely is more useful than filling the gap. The material for Suvarna Commercial Bank Limited runs to one single year, and the period is named because both lines belong to the same one.

The last row of the table is the cost to income ratioWhat an institution spent running itself for a period, set against everything it earned in the same period. Written as a share of income, so the base has to be named or the figure means nothing. for the year. A ratio without its base is a number without a meaning, so note that this one is stated on total income and on nothing else. Now the three absences, named in one place rather than worked around. There is no count anywhere of the accounts those deposits sit in, so a cost per account cannot be worked here at all. The material carries no split of the Rs 5,040 crore operating expense line, so no share of it can be attributed to verification. And it carries one year and no second period, so no before and no after exists against which any change could be tested.

The more useful thing stands in its place: the exact list of what would have to be asked for. A count of accounts, a split of the operating expense line, and two periods. A reader who knows those three names can walk into any real arrangement and ask three specific questions. Three specific questions are worth considerably more than a plausible number nobody could have checked.

What happens when a check does not complete?

Mechanically, two things can happen and both of them are reports about evidence. A comparison that finds no record returns nothing, and a comparison that finds a record which does not agree with the details presented returns a mismatchA record was found, and it does not say the same thing as the details that were presented. Two written things disagree; which of the two is out of date is a separate question the comparison never answers.. Neither of those is a verdict. Neither of them is even an opinion. One reports the absence of a record and the other reports a disagreement between two written things, and the comparison never says which of the two is the one that is out of date.

The mechanical consequence is simple: the account does not open, and the step stays outstanding. And here is the rule: a check that does not complete is a statement about evidence and never a statement about a person. An arrangement that treats it as the second has drawn an inference its own process cannot support. The process it ran was a comparison, and a comparison that finds nothing has reported the absence of a record and nothing else at all.

The route matters more than the diagnosis, and the route is the thing to go and find. A record that does not agree with the details presented is corrected at the register that holds it. The register is a different place from the bank or intermediary that ran the check. And a complaint about the arrangement's own handling of the step has a route, and a time inside which it must be answered. Both belong to the Reserve Bank of India at rbi.org.in, and both change, so a stale version of either would point somebody at the wrong counter.

What goes wrong when a completion figure is read as a conclusion?

The error that costs the most is not made by anybody near the counter. The person who makes it is whoever is handed a completion figure for a verification step and has to explain it before the end of the day.

How does somebody handed a completion figure actually use any of this?

Who sets what the check asks for?

Six conditions have been touched above. An authority fixes each of them, and each one changes when that authority changes it. The sheet below carries the condition, the body that fixes it, and a blank where the value would be.