First Loss Default Guarantee: Where a Loss Actually Lands
A first loss default guarantee is a promise by the party that sourced a pool of loans to bear the losses on that pool up to a stated share of it. The lender still holds the loans, still made the decision to lend, and still records the loss when a repayment fails to arrive. The promise changes who ends up out of pocket, and it changes nothing else.
Where the figures come from. Every rupee below belongs to Rukmini Finance Limited, an invented lender, and each one is worked out from the three or four figures the business starts with. Further down sits a table whose last column has nothing in it. Those six entries belong to a regulator that revises them on a timetable of its own, so each row carries a name to ask and an address to ask it at, and the reading is done on the day it matters.
Most readers have met this arrangement without meeting its name. An application on a phone finds a borrower and hands them a loan in four minutes. Somewhere behind the screen a registered lender put up the money. And somewhere in the agreement between those two businesses is a line saying that if the loans go bad, the party that found the borrowers will absorb the first slice of the damage, up to a point.
The line about absorbing the first slice is a first loss default guarantee. The wording is short and it sounds reassuring. The reassurance also points in the wrong direction, and the arrangement is misread more often than almost anything else in lending. So it is worth slowing right down and being exact about three things: who promises what to whom, what the promise actually pays, and which of the things it might be expected to move does not move at all.
One sentence carries the whole of it. A loss that has happened has happened. An arrangement of this kind reaches only who is finally out of pocket, never whether anything went wrong.
Hold that and everything else follows. A repayment does not arrive. The lender that holds the loan records the charge in its own accounts. Only then, and separately, does a promise made by a different business decide whose money finally absorbs it. Three events, three moments, and this arrangement touches the third one only.
What is a first loss default guarantee, and who gives it to whom?
Three parties stand in fixed places, and almost every confused question about this arrangement turns out to be a question about which of the three is being asked about. Set them down before anything is compared.
The first is the party that finds the borrower. The provider runs the application, reaches people the lender behind it would never have reached, and often handles the collection afterwards. Below it is called the lending service provider rather than given a name of its own.
The second is the lender. The lender is registered to lend, makes the decision on the application, hands over the money, and holds the loan on its own balance sheetA single-date statement of everything a business is holding and everything standing against it. How to read one is set out under the balance sheet itself.. The money it hands over is money it borrowed from somebody else. Borrowed money matters more than it sounds, and the next part is built on it.
The third is the borrower, who owes the money to the lender.
Now the arrangement itself, standing alongside all of that rather than inside any of it: the lending service provider separately promises to bear the losses on the pool of loans it sourced, up to a stated share of that pool. Read the word separately twice. The promise is not a feature of any loan. The promise is not a term the borrower agreed to. The promise is a second agreement, between two businesses, about money that has not yet gone missing.
Three properties follow, and everything after them is downstream of them. The lender holds the loan. The lender made the decision. And the promise is given by a third party rather than being built into anything the borrower signed.
Whether such a promise may be given at all in a particular market, how far it may reach, what form it may take and how long it may stand are settled elsewhere. All of that is decided by the Reserve Bank of India at rbi.org.in, all of it gets revised, and a sentence that fixed such a value in place would be carrying a falsehood the moment the revision landed, not merely an old fact. The name goes in and the number stays out.
A pool of loans is covered by a first loss default guarantee. Whose balance sheet are those loans on?
Why does an arrangement like this exist at all?
The funding is what makes a lender behave the way it does, so trace the arrangement from the funding rather than from the lending.
Rukmini Finance Limited holds assets under managementOne number standing for everything a lender has out on loan and remains answerable for. Here it covers a single business across a single stated year. of Rs 18,000 crore. Behind that book sit borrowings of Rs 14,400 crore, on which it pays 8.50 per cent a year, and its own net worthThe lender's own stake in its own business. Turn everything into cash, clear every claim against it, and this is the amount still standing. Where it accumulates from is settled elsewhere. of Rs 3,600 crore. The borrowings and the net worth added together land back on Rs 18,000 crore to the rupee, with nothing else in the total. Whatever this lender has out on loan arrived from one of exactly two places: somebody it borrowed from, or its own pocket.
The liability decides the business. Those borrowings carry a rate and a repayment date of their own, and neither of them waits to see how the book is doing. Rs 14,400 crore at 8.50 per cent a year is Rs 1,224 crore of interest for the year whether the lender writes one loan or a hundred thousand. So the lender needs somewhere to put the money, and it needs that somewhere at the pace its own funding sets.
Think of a wholesaler who has taken a warehouse on a yearly rent. The rent is the same in a slow month as in a busy one, so the pressure is always to move stock. A lender's borrowings are that rent, and loans are the stock.
Now the other side of the arrangement, and this is what makes it an arrangement rather than a favour: a business that can reach borrowers may have no money to lend them, and a business with money to lend may have no way of reaching those borrowers. One has the shopfront and no stock; the other has a warehouse full of stock and no shopfront. The two businesses are obviously better off combined, and the only question is on what terms.
The guarantee is one of those terms. The party that found the borrowers is saying, in effect, that it believes in what it is passing over, and that it will stand behind the first slice of whatever goes wrong. A reader who has traced the funding first can see why such a promise gets offered, rather than simply being told that it does.
A partner promises to bear the first losses on a pool of loans up to a stated share of it. In a year when losses turn out to be small, what does the promise pay?
What does the cover pay, and when does it stop paying?
Here is the mechanism, and it is worth meeting as a shape before a single rupee is attached to it.
The cover is a stated share of the pool. The losses on that pool are also a share of the pool. Two quantities, both measured against the same thing. Measuring both against the pool is what makes them comparable at all.
The rule is the whole mechanism: the promise pays whichever of those two is smaller. Not the cover. Not the losses. The smaller one.
The same rule is already in use elsewhere without being named. A travel policy that covers up to a stated amount of lost baggage pays the value of what was lost or that stated amount, whichever is less, and a bag worth a tenth of the limit does not pay out the limit.
And here is the consequence readers get wrong: the payment stops for two entirely different reasons. It stops because the cover was used up while losses kept coming. Or it stops because the losses ran out while cover was still sitting there unused. The name of the arrangement points hard at the first reason and says nothing about the second. So the second one gets missed.
Neither reason changes the fact that the losses happened. In the first case the lender absorbs the part above the cover. In the second case nobody needs to absorb anything beyond what the promise reached. In neither case did a rupee of bad lending become good lending.
What does it look like worked on this lender's own figures?
Now put numbers on it, and start with a warning that belongs at the top rather than in a footnote. Rukmini Finance Limited's own figures carry no partnership split of its book at all, so the working below supposes that the whole of the Rs 18,000 crore was sourced under one such arrangement, and that supposition is stated every time it is used rather than quietly assumed. The working is a counterfactualA working that says plainly it is a supposition rather than a report of something that happened, so a reader is never left to guess which of the two they are reading., not a fact about this lender.
Here is the year, built once, and every figure in it is worked on assets under management of Rs 18,000 crore. Charge 14.50 per cent a year across the Rs 18,000 crore book and Rs 2,610 crore comes in. Pay 8.50 per cent a year across the Rs 14,400 crore of borrowings and Rs 1,224 crore goes out. Subtract the second from the first and net interest incomeWhatever the lending side earned, minus whatever the funding side cost, and nothing else taken off yet. Built up elsewhere and spent here. is Rs 1,386 crore, standing at 7.70 per cent of the book. Operating expensesThe bill for running the business itself over the year: people, systems, premises, everything that is neither the cost of funding nor a charge for lending gone wrong. take Rs 540 crore off that, or 3.00 per cent measured on the same Rs 18,000 crore.
And here is the figure that does not move anywhere in this part: the credit costThe charge a lender runs through its own accounts for one year of lending, always quoted against a stated base. The stock of provisions sitting behind it is a separate matter. for the year is Rs 396 crore, or 2.20 per cent of assets under management, and it is what the year actually cost. The cover moves in the table below. The loss does not.
| Cover, as a share of the pool | Cover in rupees | The promise pays | The lender bears | Profit before tax |
|---|---|---|---|---|
| None at all | Rs 0/- | Rs 0/- | Rs 396 crore | Rs 450 crore |
| 1.00 per cent of the pool | Rs 180 crore | Rs 180 crore | Rs 216 crore | Rs 630 crore |
| 2.20 per cent of the pool | Rs 396 crore | Rs 396 crore | Rs 0/- | Rs 846 crore |
| 4.00 per cent of the pool | Rs 720 crore | Rs 396 crore | Rs 0/- | Rs 846 crore |
Work the rows rather than reading them. At no cover the promise pays nothing, the lender bears the whole Rs 396 crore, and profit before taxWhat is left of the year after every cost of running and funding the business and every charge for lending gone wrong, and before any tax is taken. Settled separately. is Rs 1,386 crore less Rs 540 crore less Rs 396 crore, leaving Rs 450 crore. At a cover of 1.00 per cent of the pool, 1.00 per cent of Rs 18,000 crore is Rs 180 crore, that is smaller than Rs 396 crore, so Rs 180 crore is paid and the lender bears Rs 216 crore, leaving Rs 630 crore. At a cover of 2.20 per cent, the cover reaches Rs 396 crore exactly, the lender bears nothing, and Rs 846 crore is left.
Now the fourth row, and it carries the whole lesson of this part: at a cover of 4.00 per cent of the pool every single figure is identical to the row above it. The cover is Rs 720 crore, but there is only Rs 396 crore of loss for it to reach, so it pays Rs 396 crore and stops. Doubling the cover from there changes nothing, and neither would trebling it. Past that point the arrangement has already done everything it can do.
Credit cost for the year is Rs 396 crore on a pool of Rs 18,000 crore. Above what cover does raising the cover further change nothing at all?
Move the cover and watch the line climb, then refuse to climb any further.
One control moves: the cover, as a share of the pool, from 0.00 per cent to 4.00 per cent in steps of 0.10 of a percentage point, each step being Rs 18 crore of the Rs 18,000 crore pool. The realised loss is held at the record's own Rs 396 crore for the year at every setting, and the control never touches it. At its opening setting the control shows the middle row of the table just above: cover at 1.00 per cent of the pool, Rs 180 crore paid across, Rs 216 crore left with the lender, profit before tax of Rs 630 crore.
Read the range for what it is. Nought to 4.00 per cent of the pool is a range of commercial terms two businesses might agree between themselves, drawn wide enough to show the line climb and then go flat. The range is not a statement about what anybody is permitted to agree. The limits are set by the Reserve Bank of India at rbi.org.in and they move, so no mark of any kind is drawn anywhere along the axis. The loss rate stays off the control as well. Moving it would mean sliding a number across the repayments that did not arrive on a whole book of loans. The shape of the line is the lesson, and that shape is perfectly visible with the loss held still. Rukmini Finance Limited's figures cover one year, one credit cost and no distribution of losses of any kind, so every position of such a control would have to be made up.
Educational illustration. Every figure here is Rukmini Finance Limited's own. One year, no forecast. The whole book being sourced under one such arrangement is a supposition and this record carries no partnership split of any kind. The realised loss stays at Rs 396 crore at every setting and the control never moves it. And the price of the promise is not in this record and is not in this drawing, so the profit figure above shows where the loss landed and is not a statement that the lender is better off. No mark on the axis suggests what is permitted.
Profit before tax reads Rs 450 crore at no cover and Rs 846 crore at a cover of 2.20 per cent of the pool. Is the lender Rs 396 crore better off?
What does that arithmetic leave out?
The largest point here, and the reason the working cannot stop at the profit line.
A promise of this kind is not given away. The party that gives it is paid for it. The payment can take several shapes: a fee, a share of what the loans earn, a deposit left with the lender that it can draw on, or some combination of all three. The shapes matter less than one fact. Not one of them is given a figure anywhere in the working above.
So say the consequence plainly rather than burying it. The movement from Rs 450 crore to Rs 846 crore of profit before tax is where the loss landed. The movement is not money made. The price of the promise is not in this record and it would sit against that movement, reducing it by an amount this record cannot supply.
The same error is caught instantly outside finance. A shopkeeper says the roof no longer leaks onto the stock, and it is obviously true, and it is obviously not the whole story until the cost of the roof is known. Nobody would call the dry stock a profit. Set out in a table, somehow, people do.
A profit line rising under a guarantee, read on its own, teaches a free lunch that nobody ever gets. So the higher figure is where the loss went rather than a gain, and whether the arrangement was worth having cannot be settled until its price is known.
What does the arrangement do to the lender's own exposure?
Exposure is the judgement that matters, and it survives after the arithmetic is forgotten.
Before the arrangement, the Rs 396 crore of credit cost for the year arises across a whole book of many separate loans. Thousands of separate situations, thousands of separate outcomes, none of them connected to any of the others. The lender finally bears the sum of a great many independent things going one way or the other.
After the arrangement, the first slice of that same Rs 396 crore depends on one party being able to pay.
So correct the obvious reading. The obvious reading is that an exposure was removed. It was not. The exposure was replaced. A spread-out exposure became a concentrated one, and the thing it is now concentrated in is the lending service provider itself. The amount did not change. The number of things it depends on did.
Everybody understands this in ordinary life and forgets it on a balance sheet. A household with two salaries from two employers and a household with one salary of the same total size are not in the same position, and no arithmetic on the monthly total will show the difference. Ten shops in one mall share one driver: if the mall closes, they all close together, however different their businesses look.
Which raises a question that has to be left open: what stands behind the promise? A promise is worth what the promisor can pay and not one rupee more. And this record carries no figures whatever for the lending service provider. Not its net worth, not its own funding, not what else it has promised to anybody else. The absence of those figures is itself the lesson. A reader meeting one of these arrangements described in a document is very often in exactly that position.
One more routing before moving on. How much of a promise received a lender may count against the capital it holds is set by the Reserve Bank of India at rbi.org.in, and it moves.
Before the arrangement the lender's Rs 396 crore of credit cost arose across a whole book. After it, the same amount is reached by a promise. What has changed about the exposure?
What does not move when a guarantee is put in place?
The list of what does not move is the heart of the whole subject, and it is longer than most readers expect.
The decision does not move. The lender that holds the loan decided to make it and is answerable for that decision. A promise standing behind the pool is not a reason to look at any application differently, and a lender that started treating it as one would be doing something the arrangement never said it could.
The recording of the loss does not move. The charge goes through the books of whoever holds the loan. A claim under the promise is a separate thing, recorded separately. The two do not quietly cancel each other out somewhere off to the side, and the whole of the Rs 396 crore appears in this lender's year at every setting of the control above.
The conduct requirements do not move. What is asked of a party that sources, services or collects a loan on a lender's behalf, and what is asked of a lender that hands part of its lending process to somebody else, is set by the Reserve Bank of India at rbi.org.in and it moves.
And here is the one that matters most: the borrower's position does not move by a single rupee or a single day. The same amount is owed to the same lender on the same date, whatever two businesses agreed between themselves about who would end up out of pocket. Not one line of this arrangement is a fact about the person who took the loan. The arrangement is an agreement about money, made between two companies, over the head of somebody who was never party to it and whose obligation is untouched by it.
A pool is fully covered by a guarantee. What changes for the person who took one of the loans in it?
What is pointed at here rather than printed?
Six requirements sit under everything above, and none of their values is stated here.
Six rows, drawn empty on purpose
| What is decided | Who decides it | The value |
|---|---|---|
| Whether a promise to bear a pool's first losses may be given at all, and any limit on how far it may reach | Reserve Bank of India, rbi.org.in | |
| What form such a promise may take, and for how long it may stand | Reserve Bank of India, rbi.org.in | |
| What a lender may count a promise received for, against the capital it holds | Reserve Bank of India, rbi.org.in | |
| Which of the two businesses carries the loan in its own accounts, and what follows from that | Reserve Bank of India, rbi.org.in | |
| What is asked of a party that sources, services or collects a loan on a lender's behalf | Reserve Bank of India, rbi.org.in | |
| What the two businesses say publicly about the arrangement between them | Reserve Bank of India, rbi.org.in |
Six rows, no values, and not one sentence of the mechanism above depended on any of them. The right-hand cells are empty because whoever is named in the middle cell revises them whenever it chooses, so a value typed in today becomes a false statement rather than an old one the first time it is revised. One extra warning belongs with this subject that the others in the sequence do not need: the range the control above moves through is a range of commercial terms two businesses might agree, and it is not a statement about what is permitted. Those two things are decided in completely different places, and the first is drawn here while the second belongs to the row above.
The control above moves the cover between 0.00 per cent and 4.00 per cent of the pool. Does that range state what a lender is allowed to agree?
What does anybody actually do with this?
Three different people meet this arrangement and each of them does something different with it.
An analyst reading a lender's accounts asks how much of the book carries such a promise and who gave it. Two lenders reporting the same credit cost are not in the same position if one of them is relying on a promise from a single counterparty for a slice of it. The question is not whether the arrangement exists. The question is how much of the book depends on one name, and what that name has behind it.
A lender's own risk function asks a narrower and harder question: what happens to this year if the promise is not honoured? That is a single arithmetic step. Take what the promise was expected to pay, put it back on the lender, and read the profit line again. On the working above, at a cover of 1.00 per cent, that step moves Rs 180 crore back and takes profit before tax from Rs 630 crore to Rs 450 crore.
And somebody being offered the other side of the arrangement asks what the promise costs them and what it commits them to. The party giving it is agreeing to write a cheque out of its own resources in a year that has already gone badly for the loans it sourced. Such a year tends to be the same year that has gone badly for everything else it does. The correlation between those two bad years is the reason a promise of this kind can be worth much less exactly when it is needed most, and no arithmetic on a good year will show it.
The failure: reading the guarantee as a gain
Here is how this goes wrong in practice, and it goes wrong for a reader who does everything the table above asked and then stops one line too early.
The wrong reading takes profit before tax rising from Rs 450 crore at no cover to Rs 846 crore at a cover of 2.20 per cent of the pool, and treats the Rs 396 crore difference as money the lender made by entering the arrangement. There are two separate costs to that reading, and the second one survives even if the first is fixed.
The first cost is an omission. The promise has a price, paid as a fee, as a share of what the loans earn, as a deposit left with the lender or as some mixture of those, and this record carries no figure for it. So the reader has set a benefit that is inside the record against a cost that is outside it. Any comparison built that way returns the answer it was handed, and being handed an answer is not the same thing as finding one.
The second cost is deeper. The arithmetic shows the loss landing somewhere else. The arithmetic cannot show whether the loss will actually land there when the time comes. Landing there depends on the lending service provider being able to pay, and no figures for that party appear anywhere above. The lender has swapped an exposure spread across a whole book for one concentrated in a single counterparty, and the table is silent about that by construction rather than by oversight.
The fix is two habits, and both of them are questions rather than calculations. Never read a profit figure moved by an arrangement without asking what the arrangement was paid. And when a number depends on somebody's ability to pay, ask what stands behind it before believing the number. Here that means asking something the working cannot answer, and saying so rather than filling the silence.
What has to be weighed before anybody calls this good or bad?
A subject like this closes on questions rather than on a verdict, and the three that matter can be asked of any arrangement of this shape. Not one of them has a number in it.
Who holds the loan? Because that answers who carries the loss if the promise does not answer, and it is the first thing to establish rather than the last.
What stands behind the promise? Because a promise is worth what the promisor can pay and no more, and a stated share of a pool is a number on paper until somebody can write the cheque.
What does the promise cost? Because the price sits against the loss it absorbs, and a comparison that leaves the price out is not a comparison at all.
And the point this closes on. Nothing in the arithmetic settles whether such an arrangement is a good one. One year of figures contains no failure, no cycle and no second year against which any such conclusion could be tested. An arrangement is not made sound by an arithmetic worked in a year in which nothing went wrong, and a year in which nothing went wrong is precisely the year that tests a first loss cover least.
An arrangement of this shape is shown with one year of figures in which nothing went wrong. What can be concluded about the arrangement?
The whole subject in one sentence. What does a first loss default guarantee change, and what does it leave exactly where it was?
What this subject does not reach
The subject here is one arrangement for sharing losses in a lending partnership, and it goes no further than that. Whose books a loan lands on under each of the ways it can be made was fixed earlier in the sequence and is used here rather than rebuilt. How a software-run lender earns its living was fixed earlier as well. Borrowing against a pool of loans before that pool is sold on is a genuinely different mechanism, covered separately, and mistaking one for the other is common enough to be worth this single line of caution. Reaching a lending decision, and the journey an application takes before money moves, were both handled earlier. Measuring, pricing or modelling credit risk sits elsewhere, along with how the machinery behind any decision gets built, tested and supervised. Deposits against no deposits, and direct selling against selling through a partner, both arrive at the end of this sequence. Splitting a return into its parts belongs to loan book economics and is covered there. The subject in hand is where a loss comes to rest rather than what a lender takes home, so the working stops short of a tax line. Six questions are pointed at rather than answered: whether such a promise is permitted and how far it may extend, its permitted form and duration, what capital treatment a lender may claim for one it receives, which of the two businesses must carry the loan, what is asked of a party acting for a lender, and what the pair must say publicly. Every one of the six is the Reserve Bank of India's to decide, and rbi.org.in appears in place of each.
Where does any of this get confirmed?
| What somebody would have to look up | Who sets it | Site |
|---|---|---|
| Whether a promise to bear a pool's first losses may be given at all in a lending partnership, and how far it may reach | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| What form such a promise may take, and for how long it may stand | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| What a lender may count a promise received for, against the capital it holds | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| Which of the two businesses carries the loan in its own accounts, and what follows from that | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| What is asked of a party that sources, services or collects a loan on a lender's behalf | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| What the two businesses say publicly about the arrangement between them | Reserve Bank of India | rbi.org.in confirmed 23 August 2026 |
| When a loan sitting on a lender's books is treated as impaired, and how the charge is recorded | Institute of Chartered Accountants of India | icai.org confirmed 23 August 2026 |
Rukmini Finance Limited is invented.
Educational material. Not advice on any investment, tax, budget or market position.
