The Treasury Policy: Mandate, Limits and Permitted Instruments

A treasury policy is the written mandate: what the treasury may hold, what it may never do, and how much of anything it may have at once. Six permitted instruments, four prohibitions, five limits. The policy exists to separate a treasury from a trading desk, and one clause does most of that work, the clause prohibiting any position taken to profit from a rate or price view.

Everything else in a treasury policy is scaffolding around one sentence, so that sentence comes first. The clause that turns a treasury into something other than a trading desk does not name a single instrument; it names a motive. No position taken to profit from a rate or price view. The clause does not ask what was bought, or how much, or from whom. The clause asks why it was bought, and it is the only line in the whole document that does.

A rule about motive is unusual, and it is unusual for a reason. Every other control in a policy can be checked from a screen. Was the instrument on the list. Was the amount inside the cap. Was the maturity short enough. A motive cannot be read off a screen at all. So the motive clause is paired with a second clause that can be checked: nothing may be entered into unless a real commercial exposure already exists for it to sit against. The motive clause states the intent and the attachment clause makes the intent testable.

The worked group here is Nirjhar Industries Limited, invented for this material: a steel and alloys maker with revenue of Rs 10,800 crore, cash of Rs 540 crore across 26 bank accounts at 4 banks in 3 currencies, and total borrowings of Rs 4,320 crore. Girish Talwalkar is its group treasurer. Its board has approved a treasury policyThe written mandate a board approves, saying what the treasury may hold, what it may never do and how much of anything it may have at once. carrying six permitted instruments numbered TP1 to TP6, four prohibitions numbered TQ1 to TQ4, and five limits numbered TL1 to TL5. Every one of those fifteen items is that invented board's own choice. None is a market convention, an industry standard or a requirement imposed on anybody by anybody.

A treasury manages exposures that other people created. The purchasing team agrees a price in a foreign currency and the exposure arrives at the treasury complete. Exposures arriving from elsewhere are what make a permitted list meaningful in the first place. If a treasury chose its own exposures, the policy would simply be a description of what the treasury felt like doing, and there would be nothing left for it to constrain. The policy constrains precisely because the exposures walk in from elsewhere and the treasury has to decide what to do with them under rules written before any of them existed.

What is a treasury policy actually for?

A policy is not a description of what the treasury already does. The distinction is worth holding. Most documents called policies in most businesses are a description of current practice: somebody wrote it down, circulated it, and got it signed. Anything the treasury does is by construction inside a policy written that way, so such a policy controls nothing.

A board approves instead a mandate written forwards. The mandate says what the function may do in situations that have not happened yet, and it is deliberately narrower than what the treasury could physically do. The gap between the two is the whole point. A policy that permits everything the treasury is capable of has spent its signature on nothing.

Think about a household version before the corporate one. A couple agree that anything above about Rs 50,000/- gets discussed before it is spent. The couple usually spend far less than that, so the rule describes nothing about what they usually do. The rule describes what happens on the day one of them sees something. Both people write the rule in advance and for the exception, at a moment when neither of them is excited. A treasury policy is the same instrument at a different scale, and it is written for exactly the same moment: the day somebody attractive walks in with something clever.

Three separate parts do three separate jobs. The mandateThe section of a policy that says what the function exists to achieve, which is what every limit below it is measured against. says what the function exists to achieve. The permitted and prohibited lists say what may be held and what may never be done. The limits say how much of anything may be held at once. Drop any one and the document loses a dimension it cannot get back from the other two.

What does the mandate section say, and what does it leave out on purpose?

The mandate is the shortest section and it is written without numbers. At Nirjhar Industries Limited the mandate names four things the treasury exists to achieve: that the group can settle every obligation on its due date, that funding is arranged and serviced, that the exposures arriving from operating are managed, and that the position is reported. The mandate names Girish Talwalkar as the person accountable for all four. Then comes the sentence most readers skip past: the mandate states what the function is explicitly not there to pursue.

The mandate leaves profit out. Not because profit is disreputable, but because a mandate that included it would make every limit underneath unenforceable. The moment a treasury is asked to contribute earnings, every prohibition below turns into an obstacle to a target somebody is measured on, and obstacles get argued with. Leave earnings out of the mandate and a prohibition is simply a boundary. Nobody has any incentive to litigate a boundary.

The mandate is the part of a policy that never gets read and always gets quoted afterwards. When somebody proposes something clever, the argument for it is almost never that a rule permits it; the argument is that it is good for the group. A mandate is the sentence held up in reply. The mandate already defined what good for the group means, before anybody had a proposal on the table.

Why is the permitted list a list of names rather than a principle?

The permitted section could be written as a principle. Something like: the treasury may hold low risk, short dated, liquid instruments. The principle sounds better, it is shorter, and it is almost useless. Every word in it is arguable. Low compared with what. Short dated by whose measure. Liquid on a normal day or on the day the cash is needed.

So a policy uses a permitted instrument listA closed list, so that anything not named on it is prohibited without needing a separate prohibition. instead, and the design that matters is not what is on it but the word closed. Anything not named on the list is prohibited by silence. The policy therefore never has to anticipate a product that has not been invented yet. A principle cannot supply that single property. A principle has to be interpreted against the new thing, and a closed list simply does not contain it.

The six at Nirjhar Industries Limited follow, and they read as names rather than as explanations. How each of them works is covered separately, under instruments.

Look down the right hand column rather than the left. Three of the six carry a condition that is not about the instrument at all. TP4 and TP5 are permitted only when something real already exists for them to sit against, and TP6 is permitted only at a grade the board wrote down. An instrument is permitted in this document not for what it is or how it pays, but for what it has been tied to. Attachment is the idea at work here, and it is a control idea rather than an instrument one.

Because the list is closed, every proposal that ever reaches the treasurer collapses into the same short conversation. The collapse is not a side effect. The collapse is the reason the section is written this way, and the section is worth seeing as a sequence of gates rather than as a document.

What does a prohibition list add when the permitted list is already closed?

This is the objection every careful reader raises at this point, and it is a good one. If the permitted list is closed, and anything unnamed is already refused, then a prohibition list is redundant. Four clauses saying no to things the document has already said no to by silence.

The prohibition list would be redundant if instruments were the only thing a policy needed to control, and instruments are not. A prohibition does not refuse an instrument; it refuses a use, and the same named instrument can be used in a way the policy wants and a way it will not tolerate. The instrument being used is on the list, so silence cannot catch the use.

Take TQ1 first, because it is the one that makes the rest testable. A forward contract is on the permitted list as TP4. Nirjhar Industries Limited holds a committed payable of 24 million US dollars (USD) falling due in 90 days. At the group's own contracted rate of Rs 84.00 to the dollar, invented for this material and never a market rate, that payable is Rs 201.6 crore. A forward against that payable is a hedge. An identical forward, same contract, same size, same price, entered into by the same treasurer on the same afternoon with no payable behind it, is a position. Nothing about the contract distinguishes the two, and everything about what it is attached to does.

The clause is written about attachment rather than about the instrument for exactly that reason. A rule saying no forwards would refuse a legitimate hedge. A rule saying forwards are fine would permit an unattached position. Only a rule about the underlying exposureThe real commercial amount a hedge is attached to, without which the hedge is itself the position. separates them, and it separates them cleanly.

Which prohibition separates a treasury from a trading desk?

TQ2. No position taken to profit from a rate or price view.

Everything else in the document limits what and how much. TQ2 limits why, and it is the only clause in the policy that does. TQ2 is therefore the strangest clause to enforce and the most important one to have. The difference between a treasury and a trading desk is not visible in any single transaction. Both can be looking at the same screen, dealing in the same instrument, at the same size, on the same day.

Consider the shape of the temptation honestly. Pretending it is rare is how policies get written badly. Girish Talwalkar has a committed payable of Rs 201.6 crore and a limit that asks for at least 60.0 per cent of it to be hedged. Suppose he forms a view that the rupee will strengthen. Hedging less leaves more of the exposure open, and if he is right the group books a gain. Nothing about that decision is exotic. The decision is made with a permitted instrument, in the right direction, on a real exposure. The reason for the under-hedging was a view about a rate, and TQ2 is what turns the deliberate under-hedging of a real exposure into a policy breach rather than a judgement call.

And that is why the clause has a partner in the limits. Nobody can audit an intention, so a motive clause on its own is unenforceable. Pair it with a minimum hedge ratio and the intention becomes visible as a number: below the floor, and the only available explanation is that somebody decided to be right about something.

Which two clauses survive contact with a salesperson?

TQ3 and TQ4, and they are worth separating from the other two because they do a different job. TQ1 and TQ2 handle the treasury acting on its own. TQ3 and TQ4 handle the treasury being sold to.

TQ3 refuses anything the treasurer cannot value from a source independent of whoever sold it. Feel the ordinary version first. A dealer sells a household a second hand car and also, helpfully, offers to tell them each year what it is now worth. Nobody would accept that arrangement, and the reason is not that the dealer is dishonest. The reason is that a valuation from the seller is not a check on the seller. An instrument that can only be priced by the person who sold it removes the group's ability to know what it holds. Losing that ability is a control failure before it is ever a loss.

TQ4 refuses anything whose worst case the board has not seen written down. Not modelled, not discussed, not described as unlikely: written down, in rupees, in front of the people who approved the policy. TQ4 exists because the structures that hurt a treasury are almost never the ones nobody understood. The structures that hurt are the ones where the good case was quantified in a presentation and the bad case was described in an adjective.

Notice what the two clauses have in common. Neither says a word about the instrument itself. One demands an independent price, the other demands a sized worst case, and both are satisfiable or not before anybody has to form a view about whether the thing is any good. Being answerable in advance is the property that makes both clauses survive a persuasive meeting.

What do the five limits actually cap?

Limits are the third part, and they supply the dimension the two lists cannot. A closed list says a bank fixed deposit is fine. A closed list has no opinion on whether Rs 500 crore of bank fixed deposits at one bank is fine. Size is a different question and it needs a different sentence.

The five at Nirjhar Industries Limited follow. Every number in the table is that board's own internal choice. Not one of them is a requirement, a market norm, a benchmark or a cap anybody imposed from outside, and that is stated beside the table.

Read the last column before the third. Three ceilings and two floors, and the mixture is not decoration. A ceiling stops the treasury doing too much of something and a floor stops it doing too little, and a limit set carries both because the two failures are different failures. Holding Rs 540 crore of cash against a TL5 floor of Rs 120 crore leaves the group Rs 420 crore of headroom above its own minimum. The headroom is a comfortable number, and it says nothing at all about whether TL1 is satisfied. TL1 asks about a completely different quantity.

How much of the payable is left open at each hedge ratio?

A limit that has never been worked is just a sentence, so the limits now go onto a real exposure. E3 Nirjhar Trading FZE, the group's overseas trading entity, agreed to buy in US dollars. The purchase decision left the group with a committed exposureAn amount already contracted, so the only uncertainty left is the rate. of USD 24 million falling due in 90 days.

The arithmetic runs in three steps and none of them needs anything from outside the group. At the group's own contracted rate of Rs 84.00 to the dollar, 24,000,000 times 84 is Rs 201.6 crore. A 5.0 per cent adverse move on that payable, the group's own scenario rather than a forecast, costs 201.6 times 5.0 per cent, being Rs 10.08 crore. And the rupees left open at any hedge ratioThe share of an exposure that has been covered, expressed as a percentage of the exposure. are simply 10.08 times whatever share is left uncovered.

Rs 1.01 crore is what those ten percentage points are worth on this one payable, and it is worth pausing on the size of it. The whole group holds Rs 540 crore of cash, so the payable itself is 37.3 per cent of every rupee the treasury has in a bank account anywhere. The gap between the two limits is not a rounding difference; it is a real amount of money that the policy has decided nobody may sit in.

Why is one hedge limit a floor and the other a ceiling?

Now the sharpest thing in the whole document. One of the two hedge limits sets a minimum of 60.0 per cent. The other sets a maximum of 50.0 per cent. Same group, same treasurer, same instruments available, opposite shapes. Most readers meet that and assume a drafting error.

Before the reason, sit with the question. The intuition almost everybody brings to it is backwards.

Move the marker into the shaded strip and watch both lanes refuse it at once. The shaded strip is the answer to the drafting error question: there is no error, and the strip is deliberate. A committed exposure must be hedged 60.0 per cent or more and a forecast exposure 50.0 per cent or less, so the two permitted bands sit on opposite sides of a ten point gap and never meet.

The reason is one sentence, and it is the sentence worth carrying away. A forecast can be wrong. A committed amount is already contracted, so the only thing still uncertain about it is the rate. A hedge deals with exactly that, and the policy therefore pushes coverage up and sets a minimum. A forecast amount may simply never happen. Hedged, and with the underlying purchase falling away, the hedge does not disappear with it. The hedge is then the only position left standing, pointing in the opposite direction to nothing at all. TQ1 exists to prevent exactly that position arriving by the front door.

Feel the household version. Somebody buys travel insurance for a trip already booked and paid for, a committed exposure. Somebody else buys the same insurance for a trip they are thinking about, then does not go. The first person covered something real. The second person is holding a policy on nothing, and if it were a financial contract rather than an insurance one, the second person would now have an exposure that only exists because they tried to cover an exposure that did not.

So the asymmetry is not caution against boldness. The asymmetry is a rule about certainty. Where the amount is certain, cover most of it. Where the amount is a guess, cover only part of it. Being wrong about the guess then costs a fraction rather than the whole.

What does a cap on one bank have to do with a cash sweep?

TL1 is the limit that looks least interesting and catches the most surprising thing. TL1 caps exposure to any one bank at Rs 180 crore, that board's own number. Most readers file that away as a housekeeping rule about where deposits go.

Then look at what the treasury was doing at the same time, entirely separately and entirely correctly. Eighteen of the group's twenty six bank accounts, all belonging to E1 Nirjhar Industries Limited and E2 Nirjhar Alloys Private Limited, run in a daily sweep into one header account. The sweep exists because scattered balances cost money, and the group had measured what it was costing. Nobody was being careless. The sweep was the fix.

After the sweep completes, the single net cash position across those eighteen accounts is Rs 324 crore, and it is sitting in one account at one bank. Hold that figure carefully. Rs 324 crore is also this group's profit after tax for the year, and the two are entirely different objects that happen to share a number. TL1 measures the first of the two: cash sitting at a single bank. Physical concentration converts many small bank exposures into one large one, and the limit that notices is the counterparty limitA cap on how much the group may have with any one bank, which cash concentration runs straight into. rather than anything in the pooling arrangement itself.

324 less 180 is Rs 144 crore. Rs 144 crore is the amount which, on the day the sweep works exactly as designed, the group's own policy will not let stay where the sweep put it.

So where does the Rs 144 crore go. The destination is itself a policy question, and the policy answers it: into TP1 to TP3, inside the TL2 cap of 12 months on any investment. Notice what has just happened to the shape of the work. A pooling decision has become an investment decision, and the investment decision is constrained by a limit set for a completely different reason.

Nobody in the case had computed the capacity arithmetic, and it is worth doing once to see how tight the group's own rule actually is. Nirjhar Industries Limited banks with four banks. Four times the TL1 cap of Rs 180 crore is Rs 720 crore of total permitted bank exposure. The group holds Rs 540 crore of cash. The whole of that cash therefore occupies 75.0 per cent of every rupee of bank capacity the policy allows, and Rs 540 crore divided by the Rs 180 crore cap is exactly three, so the group's cash needs a minimum of three of its four banking relationships simply to be allowed to sit somewhere.

Who actually reads a treasury policy from outside the group?

Most treatments of this subject stop at the board approving the document, and that leaves out the people who read it hardest. Three of them are worth naming. Each reads it for a different reason and each looks at a different part of it.

A lender reads it as a covenant on behaviour. Vindhya Commercial Bank Limited, invented, carries Nirjhar Industries Limited as its single largest exposure. When that bank's credit team asks for the borrower's treasury policy, it is not curious about deposits. The credit team wants to know whether the group can take a position that could hurt it, and it will go straight to two clauses: whether derivatives require a documented underlying exposure, and whether anything may be entered into on a view. TQ1 and TQ2 are, from a lender's side of the table, the two sentences that decide whether a borrower's earnings can be surprised by something the lender never lent against.

An analyst reads it for what the numbers will not show. A group that hedges committed exposures to a stated floor produces earnings whose currency effect is bounded, and a group with no such floor produces earnings where a currency move can arrive without warning. Neither policy is better; they are different, and only one of them is legible from outside. A policy is the document that tells an outsider which parts of a group's reported earnings were the result of a decision and which were the result of a rate.

And a household reads a version of the same document without ever calling it one. A couple who agree that the emergency fund stays in something they can reach in a day, that nothing gets bought from a person who called them, and that any single institution holds no more than a stated amount, have written a mandate, a prohibition and a counterparty cap. The couple wrote it in advance and while calm, about situations that have not happened. Writing it then is the whole design, and the scale is the only difference.

Who approves the policy, and what has to be in an exception request?

The board approves it, and the reason that matters is not ceremony. A limit approved by the person it constrains is not a limit. Girish Talwalkar proposes, the board approves, and the treasurer then operates inside something he cannot himself relax. Every one of the fifteen items at Nirjhar Industries Limited sits at that level for exactly that reason.

Which raises the practical question that decides whether a policy survives contact with a real business. Something will eventually come up that the policy refuses and that somebody has a genuinely good argument for. A policy with no exception route gets ignored the first time that happens, and a policy with a casual exception route was never a policy. So the exception request is itself a defined artefact with fixed parts, and the parts are chosen so that a request missing any of them cannot be answered by anybody.

Look at the third row for a moment longer. The third row is the same idea as TQ4 wearing different clothes. TQ4 refuses a structure whose worst case the board has not seen written down. Part 3 of the exception form asks for the worst case in rupees. Both make the same demand: somebody must put a number on the bad outcome before anybody is asked to say yes. The demand is uncomfortable to make and it is the single most useful thing in the whole document. Everything else can be argued and a blank box cannot.

One last property of a good exception route: it expires. An exception granted for one deployment ends with that deployment, rather than quietly becoming the new practice that the next policy review writes down as current behaviour. A mandate written forwards turns back into a description of what the treasury already does in exactly that way, one unexpired exception at a time.

Which subjects sit outside a treasury policy?