How to Map a Swap Cash Flow Exchange onto One Sheet

Nine steps, in order, turn one agreement into a single sheet. Write the notional down and label it as not a cash flow. Record each leg on its own. Record the calendar. Record the counting method. Work the first period and keep the working. Mark the direction in words. Draw the rows nobody can fill. Name whoever sets a condition. Then stop.

A swap agreement is a long document holding a small number of things that matter to anybody tracking the money. The routine below exists to pull those few things out in a fixed order and put them on a surface that comes out the same shape every time. Two arrangements mapped this way can be compared without reopening either document, and a row nobody is able to fill shows up as a blank row rather than as something a reader forgot to go looking for. The blank row is the property most templates throw away.

Chitrakoot Cements Limited, an invented manufacturer, has agreed to hand over a rate that never moves, 7.20 per cent a year, receiving whatever the floating benchmark reads in exchange. Saranga Capital Limited, invented alongside it, stands on the opposite side of that same arrangement. One shared notional of Rs 1,000 crore sits under both sides of it.

One shortage shapes everything this sheet can hold, so it is worth stating before step one rather than discovering it at step seven. Nowhere in the underlying record is there a schedule of what the floating benchmark will read on any future date. Without such a schedule no period beyond the first can be worked at all, and no figure on the sheet can carry a value for the arrangement today. Step seven exists for exactly that reason: the routine deals with the shortage by drawing it on the sheet.

What does a finished sheet look like before anything is filled in?

Knowing where a figure is going to land makes the act of pulling it out of the document much faster, so the shape comes before the steps. The finished sheet is four regions stacked one above the next. At the top, an identification region carrying who pays which leg and the single figure the two legs are worked from. Underneath it, the two legs, each written out far enough that somebody could check either one without looking at the other, with the calendar riding between them. Below that, a settlement region holding what came out of the arithmetic and which way it went. At the foot, a region of rows drawn with their headings and left blank, each printing the reason it stays blank.

Every arrangement mapped this way produces the same four regions in the same order, and that sameness is the whole reason the sheet is worth building. The notional is in the same place on both sheets, the direction row is in the same place on both, and the empty rows are in the same place on both. Somebody who has mapped two agreements can therefore lay them side by side and read straight across. Reading two agreements instead means comparing two authors, two orderings and two vocabularies.

Step one: which figure goes down first, and what gets written beside it?

The notional goes down first, in the top row of the identification region, with the word notional printed beside it. Then, before anything else at all is recorded, the next column on that same row gets the words NOT A CASH FLOW. Only after that does the routine move on. Chitrakoot Cements and Saranga Capital are computing on Rs 1,000 crore. Written out to the last digit that is Rs 10,00,00,00,000/-. Not one paisa of it will ever be paid by either side.

The sheet keeps a row for the headline figure and a separate row for the money, and prising those two apart is the very first job the routine does. Rs 1,000 crore sits in the first row with its label. Rs 12.00 crore of net difference sits in the settlement region much further down, and that lower row is the only one on the sheet describing money leaving a bank account.

A figure written without its label gets copied without its label. The marking therefore happens at the instant the figure is written, never as a tidy-up afterwards. Watch how that goes wrong. Somebody lifts the notional into an email. Somebody else lifts it out of the email into a summary. By the third copy the figure is sitting in a column headed amount, next to figures that really are amounts, and nobody in the chain can now say which column it came from. The label is cheap at the moment of writing and expensive to reconstruct at any point after.

Step two: how are the two legs recorded so either one can be checked alone?

Each leg gets its own block, and each block is written out far enough to stand up to a check on its own. Four things go into a leg block: which party pays it, what rate it uses, where that rate came from, and how often it falls due. For the fixed leg, the rate is copied once out of the agreement and never touched again for the life of the arrangement. Chitrakoot Cements pays 7.20 per cent a year, and it will read the same in period nine as it does in period one. For the floating leg, the block gets a row per period, and only the periods whose resets have already happened carry anything at all. Saranga Capital pays what the floating benchmark read at the relevant reset. For the first period that reading was 6.00 per cent a year, and every later row on that block stays blank.

Never collapse the two legs into their difference on this sheet. The difference gets its own row lower down. Holding all three figures apart is precisely what lets somebody name the leg that went wrong, months after the period closed. A single collapsed figure can say that Rs 12.00 crore was owed. A single collapsed figure cannot say whether the rate was wrong, the fraction was wrong, the notional was wrong or the subtraction was wrong, and those are four different conversations with four different people.

Here is the same idea in a setting with no rates in it at all. Picture two shopkeepers on the same street who have agreed to swap what they charge each other for storage. One of them quotes a flat rate that will not change all year. The other charges whatever the going rate for a shed happens to be that month. At the end of the month nobody carries two bundles of cash across the street. Whoever is short walks over with the difference, and that single walk is the whole settlement. Now suppose the two shopkeepers disagree about the walk. If each has written down only what they handed over, neither can reconstruct the two charges that produced it. If each has written down both charges, the disagreement resolves in a minute.

Step three: which dates go on the sheet, and in what order?

The dates go down in the order they occur, and very often the document lists them in a different order. Four kinds of date matter to somebody tracking the money. The effective dateThe day the first period starts running. Anything before it is paperwork rather than accrual., when the first period starts running. The reset dates, one for each period, each falling at the start of the period it applies to. The payment dates, one for each period, each falling at the end of the period it covers. And the end date, when the last period closes.

Reset dates and payment dates go in two separate columns, even where a particular date looks identical in both. That looks like duplication and it is not. The whole shape of this arrangement is that one kind of date sits before a period and the other kind sits after it, and a sheet that merges the two columns into a single date column has destroyed the exact information it was built to hold. Look at what happens on the day period one ends. The day period one ends is the payment date for period one and the reset date for period two. One column would show one entry there and a later reader could not tell which of the two jobs that entry was doing, or whether the other one had simply been forgotten.

There is a related discipline about how the columns are read. The two legs may fall due at different frequencies, so the two columns need not line up, and on many arrangements they will not. A reader can run a finger down the reset column and account for every reading, then run it down the payment column and account for every transfer, without either exercise contaminating the other.

Step four: where does the counting method go, and why does it need its own row?

The counting method gets a row of its own inside each leg block, one row per leg, and what goes in that row is the method named exactly as the agreement names it, and nothing further. Two reasons it earns a row rather than a footnote. The first is that the two legs need not use the same one, and a footnote covering both would quietly assert that they do. The second is that this is the input which turns a rate into an amount, so it belongs beside the rate rather than at the foot of the sheet in small print.

The row holds no statement about what a given counting method is called in any market, or which one a market usually picks. Both of those are set somewhere outside this sheet and both move. No sheet ever announces that it has gone stale, so a sheet carrying either would drift out of true without anybody noticing. The name the agreement uses is copied across as it stands, treated as a quoted label and nothing more.

Somebody reading this sheet later is entitled to ask what the fraction was for the period in question. The first period runs a full term, so the honest answer is 1.0000. The fraction lives with the working rather than in this row. Keeping the method and the fraction in different places matters: the method is a rule that holds for the life of the arrangement, and the fraction is an output that changes every period.

Step five: how is the first period worked, and how much of it stays on the sheet?

Each leg is worked on its own line, showing the inputs and not only the answer, and then the difference goes on a line beneath the two of them. Three lines, not one. The same relationship is applied on both lines, and writing them one under the other is what makes the sameness visible.

Run it twice for the first period. Rs 1,000 crore carried at 7.20 per cent a year across a full period lands on Rs 72.00 crore gross for Chitrakoot Cements. The same notional, taken at the 6.00 per cent a year the reset produced, leaves Rs 60.00 crore gross for Saranga Capital. Subtract the smaller from the larger, and the sheet keeps Rs 12.00 crore of net difference. The subtraction is legitimate only because one date and one currency sit behind both figures.

Two routes reach the same Rs 12.00 crore. Multiply first: 1.20 percentage points applied to Rs 1,000 crore. Then divide instead: Rs 12.00 crore set against a notional of Rs 1,000 crore returns 1.2 per cent. The distance between the two rates has shown up again from the other end. Both routes belong on the sheet in the sense that either can be used to check the other in about ten seconds, and neither belongs on it as a substitute for the two gross lines.

All three amounts stay on the sheet permanently, and the reason matters more than it first appears: a sheet carrying only the settlement figure is a sheet nobody can audit. The settlement is the one figure on the whole sheet that can come out right for the wrong reasons. A rate typed in too high and a fraction typed in too low can produce a perfectly correct difference. So can a notional entered on one leg and a different notional entered on the other, if the two errors happen to cancel. Every one of those is invisible in the difference and obvious in the two lines above it.

Step six: why is the direction written in words instead of as a sign?

The direction gets its own row, filled in words, naming the party the money went to for that period. On the sheet for the first period that row reads: paid by Chitrakoot Cements to Saranga Capital. Not a plus, not a minus, not a bracket.

A plus or a minus on a settlement figure means whatever the person who wrote it happened to think it meant, and there is no convention anywhere on the sheet that fixes it. Minus might mean the party paid. Minus might equally mean the party received, with the sheet built from the counterparty's point of view. Or minus might mean the figure was entered as a reduction in a working file and never converted. A row reading paid by Chitrakoot Cements to Saranga Capital cannot be read the wrong way round by anybody, including by the person who wrote it, six months after they have forgotten which convention they were using that week.

The direction row is also filled per period rather than once at the top of the sheet. In the arrangement mapped here the fixed rate sits above the first floating reading, and the direction feels settled enough to write once at the top. The direction is not settled. A label on a party says which leg that party pays; it says nothing at all about which way the difference runs in any given period. Suppose the benchmark at some later reset stood above the fixed leg's own 7.20 per cent a year. Money for that period would then travel in the opposite direction, and the row covering it would say so. The routine has no view on whether that will happen, and this record contains no reading that would let anybody form one.

Step seven: which rows can never be filled, and what would fill them?

Two rows on the finished sheet are drawn with a heavy border and left empty. The first row asks the value of this arrangement now. The second asks what settles after the first period. Neither is answerable from anything that exists here, and both are drawn anyway.

Beside each blank row the sheet names the input that would fill it. In both cases the input is a dated schedule putting a reading against each future reset date. Naming the missing input turns an absence into an errand. A reader who later obtains such a schedule, from a source that publishes one and dates it, knows without asking anybody exactly which box it goes into. A reader who does not obtain one at least knows the shape of what is missing.

A deleted row looks like a question nobody asked, and a blank row looks like a question nobody could answer. Only the second of those is true, so the rows are drawn with their headings and left visibly blank rather than deleted. A tidy-minded person deletes those rows on a second pass through the template, so the justification above is the sentence to carry away. The tidier version reads better and tells the reader less. Somebody picking up a sheet with the rows deleted has no way of knowing whether the arrangement was valued and the answer was left out, or whether valuing it was never possible, and those are opposite situations.

Step eight: how is a condition recorded when it cannot be stated?

Two more rows sit at the foot of the sheet and neither carries a value. One asks which details of an arrangement struck in private must be reported, where they go and how soon. The other asks what collateral a counterparty places behind one. Inside each of those rows the sheet prints the Reserve Bank of India and rbi.org.in, and nothing else at all.

The rule that produces those rows is short: write the question and the address, never the answer. Both of these are decided by somebody outside the arrangement, both of them change, and a sheet is a document that gets copied, reused and inherited long after whoever built it has stopped looking at it. A figure entered in one of these rows today becomes a false statement the morning the authority moves it, and nothing on the sheet will put a hand up. An address is an instruction to go and look, so an address in that row cannot go stale.

There is a second reason for drawing the rows at all rather than leaving that region off the sheet. A sheet with no conditions region anywhere on it lets a reader quietly assume there are no conditions. Of the three possible states that one is the most dangerous. So: a stale figure is worse than a blank row, and a missing row is worse than both. The blank row with an address in it is the only version that leaves a reader correctly informed about their own ignorance.

Step nine: when does the routine stop, and what has the sheet still not told anybody?

The routine stops when the fifteenth row has either an entry or a printed reason. There is no tenth step, and resisting the urge to add one is part of the routine.

A completed sheet describes an arrangement accurately. A completed sheet answers nothing whatsoever about whether anybody should be inside one, and the routine refuses to be pushed into answering it. Answering would want the existing obligations of Chitrakoot Cements to hand, along with what its own cash does when a benchmark shifts, over which stretch of time the question is being asked, and how it would stand in a period where the difference ran towards it instead. All of that lives in the records of the company itself.

The refusal is not a formality bolted onto the end. The refusal follows from the shape of the sheet. A party's position is made of everything else the party is holding, and none of that appears anywhere on the sheet, so fifteen rows describing one arrangement cannot say anything about that position. A reader who feels the sheet is close to answering the bigger question should notice how quickly the feeling arrives once the arithmetic works out neatly, and treat that feeling as the signal it is.

What does the finished sheet look like with every row on it?

Here is the routine run once, end to end, on one agreement. Chitrakoot Cements Limited signed it. Saranga Capital Limited signed the other side of it. The whole sheet is reproduced so the shape can be copied straight off it, with the notes following underneath.

Now count. Fifteen rows on the sheet. Eleven of them carry something taken from the agreement or worked out of it. Four of them carry the reason they hold nothing. A reader who expected all fifteen rows to fill in has not yet met the limitation that shapes this arrangement: only the first period has a benchmark reading behind it. The assumptions sitting beside this sheet are worth reading as carefully as the entries: the first period runs a full term at a fraction of 1.0000, the floating benchmark is an illustration rather than any published series, no reading has been obtained for any period after the first, and the sheet stops short of valuing the arrangement.

Who fills in which row once the sheet is actually in use?

A sheet is only worth building if somebody fills it, and the useful way to think about the labour is by field rather than by job title. Take the rows one group at a time and ask who is holding the document that answers them.

Rows one to four and row six come off the documents on the day the arrangement is set up. Whoever is reading the confirmationA short document sent after a deal is struck that repeats the terms back, so both sides can check they wrote down the same thing. against the master agreementA standing document two sides sign once. Later arrangements between them sit underneath it instead of repeating its wording. fills them, and they should never need touching again. If somebody edits row three later in the life of the arrangement, something has gone wrong that is much bigger than a sheet. One caution about sources: the term sheetA short summary of what has been agreed, circulated between the two sides before the long documents are drawn up. circulated before those documents is not a source for these rows. A term sheet is a summary, and summaries drop conditions.

Row five and rows eight to eleven are filled period by period, and they belong to whoever watches the calendar. In most arrangements that is the back officeThe people who record, check and settle what the dealing side agreed, working from the paperwork rather than from the conversation., working on the morning after each reset and again on the morning of each payment. The reading is easy to find on the reset day and awkward to find a quarter later, so the single most valuable habit in the whole routine is filling row five on the reset morning rather than on the payment morning. Row eleven, the direction, is filled by whoever writes the payment instruction. The person writing that instruction has to state the direction anyway and is the last one who can catch it being backwards.

Rows twelve to fifteen belong to nobody inside the company at all, and drawing them is the only way to say so. Rows twelve and thirteen wait on an outside input that has to be obtained and dated. Rows fourteen and fifteen wait on an authority. A completed sheet with those four rows blank is not an unfinished sheet; it is a finished sheet that is telling the truth about the edge of its own knowledge.

Two other pairs of hands read the sheet without filling anything in. Somebody assessing a borrower reads rows one, two and eleven together. The notional tells them the scale of the arrangement, the direction row tells them which way money has actually gone, and the gap between those two facts is often the first thing a hurried reader gets wrong. Somebody keeping the company ledgerA running record of what a party owes and is owed, kept in date order. reads rows eight to eleven and nothing else. Those four rows are the only ones that ever turn into an entry. And a household version of the same discipline is not hard to picture: a household with a loan whose instalment moves and a rent that does not writes down both numbers each month and then the difference, rather than only the difference, for exactly the reasons step five gives.

Chitrakoot Cements Limited and Saranga Capital Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.