Retail and Corporate Lending Compared, Side by Side

Retail and corporate lending differ in the size and the number of the promises a lender holds, not in the promise itself. Retail lending is many small exposures on standard terms. Corporate lending is few large exposures on terms drafted for that borrower. The same book value can be thirty lakh borrowers or three hundred, and one loan not being repaid means something completely different in each.

A loan is the same object on both sides of this comparison. Money now, against money later, on terms that were stated before the money moved. Two numbers change: how many copies of that object a lender is holding, and how large each copy is. Almost everything else said about the two, the paperwork, the pricing, the way a lender watches what it is holding, falls out of those two numbers rather than sitting beside them. A set of criteria is fixed first, each side is described in full on its own, and only then are the two put next to each other.

What actually separates lending to a household from lending to a company?

The criteria go down before either side is looked at, and that is not tidiness. Work the criteria out while the comparison is being written and each side ends up described in whatever terms suited it. The result is two portraits under one title, and the reader is left holding two descriptions that never meet. A criterion earns its place here by separating the two sides, and a criterion that returns the same answer on both has taken up room and taught nothing.

Five of them do the work. The first is who the borrower is. The second is how many exposuresOne borrower or one agreement counted as a single item of what a lender could lose, so that a book of many loans is a book of many exposures. there are and how large each one is. The third is whether the terms are standard or drafted for that borrower. The fourth is what stands behind the loan. The fifth is what one loan not being repaid does to the lender's book, and it is the only one of the five that changes the kind of reading a lender can take at all.

Three things are not in the set. There is no criterion about how a lender decides whether to lend, no criterion about which side is safer, and no criterion scoring either against a standard both are trying to meet. The first belongs with assessing a borrower, a subject covered separately. The second and third would need a ranking, and none is produced here.

What is retail lending when it is taken on its own?

Retail lending is lending to households and to very small businesses. A loan to buy a home, a loan to buy a vehicle, a loan taken against an asset the household already has, a balance carried on a card, a small loan to a shop. Different products, one shape. There are very many of them and each one is small, and every other property of retail lending falls out of those two facts.

Because there are very many, one set of terms is written once and copied. The document the borrower signs is a template with the amount, the period and the rate dropped into it. The rate itself arrives off a published schedule rather than out of a conversation, and the schedule exists precisely because there are enough borrowers to make writing one worth the trouble. Building a schedule and standing behind it is work, and that work only pays when it is spread across a very large number of agreements.

Try the everyday shape of it first. The finance version is the same shape at another scale. A tea counter outside a station serves four hundred people a morning. The counter does not know who any of them will be. The owner knows the shape of the morning instead: roughly how many cups, roughly when the rush comes, roughly how many will want it strong. One price list runs the counter, and the shape of the crowd is what the owner plans on. A retail book is managed the same way, as a population rather than as a list of names. With very many small agreements there is no other way to work, and a lender on this side works by the same logic.

The template cuts both ways, and the second edge is worth seeing. Because the terms are standard, the borrower gets the same document as everybody else on that product, so there is nothing to argue about and nothing lost by not arguing. The template also stops the lender quietly pricing one borrower differently without moving the published schedule for all of them. A standard term is a promise made to a population rather than to a person, and that constrains the lender as much as it constrains anybody borrowing. Where a lender does want to separate one borrower from another on this side, it moves them to a different product rather than rewriting a clause.

What is corporate lending when it is taken on its own?

Corporate lending is lending to companies. A term loanA loan drawn once for a stated amount and repaid over a stated period, so the balance falls as the period runs and never rises again. to build or buy something that will last years. A working facilityA borrowing limit a company draws on and repays over and over through the year as money comes in and goes out, rather than taking it once and repaying it once. the company draws on and repays through the year as stock is bought and sold. A loan written against a specific contract the company has already won. There are few of them and each one is large, and that pair of facts drives everything else exactly as the mirror pair drives the retail side.

Because there are few and each is large, no template survives contact with the borrower. The terms are drafted for that company: what it may do while the loan is outstanding, what it must show the lender and how often, what happens if a stated condition stops being met. The rate is arrived at for that loan rather than read off a schedule. And the book is watched name by name, for the plain reason that there are few enough names to watch.

The everyday version sits one step up the same street. The wholesaler who supplies that tea counter has eleven customers. He knows each one's Tuesday by heart, he knows which one pays late in the monsoon, and he does not have a price list because every conversation is a conversation. Eleven relationships are read; four hundred customers a morning are measured. Nothing about the wholesaler is more careful than the tea counter. The number of relationships decided the method for both of them.

The working facility has a shape with no retail equivalent, and it deserves a second look. A term loan is drawn once and falls away as it is repaid. A facility rises and falls all year as stock is bought and receipts come in, so the lender is watching a balance that moves rather than a schedule that runs down. A corporate agreement therefore spends much of its length on what the company may do while the loan is outstanding, rather than on the repayment dates. On a balance that moves, the dates are not the interesting part. The lender wants control over the money between drawings instead.

How do the two read across one set of criteria?

Now run the five in the same order down both columns, so either column can be read on its own. Who the borrower is: a household or a very small business, against a company. How many and how large: very many and each small, against few and each large. Whether the terms are standard or drafted: standard and published, against drafted for that borrower. What stands behind the loan: often one named asset on the retail side and often a mixture on the company side, and whether a loan is secured at all is a separate split covered alongside this one. What one loan not being repaid does: almost nothing on its own, against a great deal on its own.

Four of those five separate the two by degree. A very small business and a company are both borrowers making the same promise, so the first row is a size boundary drawn across a continuous range rather than a break in it. Many-and-small against few-and-large is plainly one quantity at two settings. A template against a draft is one document with more or less of it moving. The backing is a mixture at both ends, with a charge over an assetA recorded claim a lender has on a named asset, which is what lets the lender reach that asset ahead of other people if the loan is not repaid. recorded in the same place whoever the borrower is.

Only the fifth row changes the kind of reading available, and that is why it is the row worth the most. Where one loan is a very small part of a very large number, a lender can work from a rate across the population and the rate means something. Where one loan can be a visible part of the book by itself, there is no population to take a rate across, and the lender is reading agreements. Reading agreements and taking a rate across a population are not two degrees of the same activity.

How can the same book be thirty lakh promises or three hundred?

Start with what is missing, because it decides how this block is allowed to be written. Suvarna Commercial Bank Limited reports gross advances of Rs 1,44,000 crore for the stated year. Nothing in the record behind these figures splits that book by borrower kind at all, so the split is simply absent. A split written to fill the gap would look precisely as reported as the total above it. One thing can be done instead and it is enough: the one published total, built two ways.

Built out of exposures of Rs 4,80,000/- each, Rs 1,44,000 crore is 30,00,000 exposures. Built out of exposures of Rs 480 crore each, the same Rs 1,44,000 crore is 300 exposures. Both are divisions of one published figure and neither adds anything to it. One exposure of the second kind is ten thousand of the first, and the printed total does not move by a rupee between the two. Both counts are counts of loans rather than statements about any household, and both exposure sizes were chosen to make a shape visible.

Now the consequence, and it is the reason the arithmetic was worth doing. A book of three hundred names is read, and a book of thirty lakh promises is measured. Nobody reads thirty lakh loan files, and nobody measures three hundred names as a population and learns anything by it. Which of those two methods is available is decided by the number of exposures and by nothing else. The lender's care, the quality of its systems and its wish to do the other one decide nothing.

The whole apparatus around each kind of lending therefore looks so different from outside. One side builds schedules, templates, scoring and reporting that describes a whole population. The other side builds documents, conditions and a person who knows the name. Both answer one question at two different counts: how to watch what is being held.

One qualification before the picture. The two constructions are the ends of a range rather than a choice between two options. A real book is a mixture: some agreements near one size, some near the other, and a long stretch in between, and it is the mixture that decides how much of the book can be measured and how much has to be read. The two constructions here are drawn separately so that the range between them becomes visible at all. Where the bank's own book sits inside that range is not stated anywhere in the record behind these figures. The absence is the one named above.

What does one loan not being repaid do on each side?

Take the one asset quality figure the record does carry and put it through both constructions. Against gross advances of Rs 1,44,000 crore the bank reports Rs 6,480 crore of gross non-performing advances, and that division gives 4.50 per cent of gross advances. Apply that same 4.50 per cent to the first construction and it is about 1,35,000 exposures. Apply it to the second and it is about 13.5 exposures. The ratio is identical and the two situations have almost nothing in common.

Watch what one exposure does to the ratio in each. On the first construction, one exposure of Rs 4,80,000/- against gross advances of Rs 1,44,000 crore moves the figure by 0.000033 percentage points, invisible at any number of decimal places anybody prints. The figure behaves like a rate, and a rate is a fair description of a population. On the second, one agreement of Rs 480 crore is 0.33 per cent of gross advances by itself. Thirteen agreements print 4.33 per cent of gross advances and fourteen print 4.67 per cent, so on that construction the book cannot report 4.50 per cent at all. The figure is a count wearing the clothes of a rate.

The 13.5 is worth pressing on. The awkwardness of it is the teaching. A count of loans is a whole number, so 13.5 is not a picture of anything that can stand on a single day. The 13.5 is what an average produces when asked to describe a small number of large items, and the average stops answering the moment the question becomes which thirteen or which fourteen. On a book of few large agreements the sizes are unequal as well, so one agreement can be worth a good deal more than the 0.33 per cent an equal share would carry, and the ratio moves by that much when it moves at all. None of that is visible in the printed percentage, on either construction.

An account written only from the lender's desk has described half of a transaction, so here is the same event from the other side of the counter. The borrower does not meet a rate. The borrower meets a route: what happens first, what happens next, what the borrower can do at each stage and by when. The route is the same object whether the loan was one of thirty lakh or one of three hundred, and it is walked properly under the kinds of loan. The route is not obvious from outside, not obvious from the document, and nobody should be told they could have found it easily.

Where the two sides genuinely differ is in concentrationHow much of what a lender is holding sits with a small number of borrowers, so that one of them moving matters to the whole rather than to a corner of it., and that is a statement about the lender rather than about anybody borrowing. The moment a loan stops counting as a performing one, the point at which interest stops counting as income, and what must be set aside from then on, all sit with the Reserve Bank of India, and they move.

Does this record carry the same measure for both institutions?

The record does not, and checking was the first thing worth doing. The shape argument above runs inside one book because that book has the two figures it needs: a gross advances total and a gross non-performing total, both for Suvarna Commercial Bank Limited and both struck on a stated base. Rukmini Finance Limited has neither of them in this record, so no contrast on asset quality can be run across the two institutions.

The near miss is the shape most bad comparisons take, and it is worth naming. Rukmini Finance Limited does carry a charge of Rs 396 crore taken in the year for credit, and that charge is 2.20 per cent of assets under management of Rs 18,000 crore. The 2.20 per cent looks like it could sit beside something. One figure is an amount charged through a single year and the other is a stock standing at a date, so the charge cannot sit beside Rs 6,480 crore of gross non-performing advances, and the two answer different questions. The charge also sits on a different base: assets under management rather than gross advances. Two figures that are not the same measure and are not on the same base are not a comparison, whatever they look like printed next to each other.

What differs on the borrower's side of the counter?

Everything above is written from the lender's desk, and the other side of the counter completes it. For a household the terms arrive as a published schedule and a form. The negotiable surface is small: the amount, the period, sometimes which product, and past that the document is what it is. For a company the terms arrive as a draft, and the negotiable surface is the document itself, clause by clause, over weeks.

The consequence rather than the observation: on the retail side the reading work is finding out what the standard terms actually say, and on the company side it is finding out what the drafted terms have been changed to. The two are different jobs. One is discovery in a document nobody negotiated, and the other is discovery in a document that moved. Neither is obvious from outside, and neither is a matter of anybody being careful enough. Where a route exists to find out, it exists at the lender and at the Reserve Bank of India at rbi.org.in, and finding it is work.

How does somebody reading a lender actually use any of this?

An analyst reading a lender's disclosure uses it as a question rather than an answer. Before comparing any ratio struck on advances, the question is what the advances are made of: how many exposures, at what sizes, in what proportions. Where that composition is disclosed the ratio can be read. Where it is not, the honest note is that the two lenders were not compared on it. The note is shorter than most people write and more useful.

Somebody putting money into a lender uses it the same way one step out. Two lenders with identical gross advances and identical asset quality percentages can be holding books that behave completely differently under the same conditions, and the published pair of figures cannot separate them. The number of exposures is the missing variable, and where it is missing the reading stops there rather than continuing on a guess.

A household uses it closer in. Where the terms offered came off a published schedule, the schedule is public and the work is reading it. Where the terms came as a draft, the work is finding what moved between the standard version and the drafted one, and asking for the standard version to compare against is a normal thing to ask for. Neither of those is a recommendation about what to borrow or from whom, and none is made here.

A lender uses the same fact about itself, one step earlier than anybody reading it does. The count decides what is worth building: schedules, models and reporting that describes a whole population on one side, and drafting, conditions and a person who knows the name on the other. Neither apparatus makes sense at the other count, and a lender doing both usually runs them as two operations with two sets of people. The split is not a preference and not a strategy anybody picked. Thirty lakh promises and three hundred agreements each require it of whoever is holding them.

What does this comparison not settle, and who decides the rest?

The comparison does not settle which of the two is better, and nothing in the criteria set could. Every one of the five separates the two and not one of them scores either against a standard both are trying to meet, so a ranking would have to be smuggled in from outside the arithmetic. Retail and corporate lending are two shapes of the same object, met at two counts.

Seven requirements got named above and every one of them was left where it sits. How exposures must be classified between borrower categories. The limits a lender must observe against one borrower and against one group of connected borrowers. The priority lendingCategories of lending a bank is required to reach stated targets in, so that part of what it lends is directed by rule rather than by choice. targets a bank must reach and how they are measured. The fair practice conditions binding a lender in its dealings with a borrower. The disclosure a lender must make about the composition of its advances. The point at which interest on a loan stops counting as income. The identity checks completed before any loan account is opened.

Every one of those is set by the Reserve Bank of India, and every one of them moves, so a figure copied into a row would be untrue from the morning it moved. Untrue is a worse thing to be than silent. Each row below is drawn with its heading written and its value left blank, and the authority that fills it is printed inside the blank. A reader takes away a form to complete at the source, and the form stays usable after the requirement has moved. A row filled from recollection would look finished and would not be.