Indemnity, Covenant and Undertaking: What Each Does

An indemnity promises to pay for a named loss when a named thing happens. A covenant promises to do something, or to refrain, across a stretch of time. An undertaking promises the same kind of thing, and any difference between the two words is one this document made. The three promises part company on what sets the payment off, and one Rs 8 crore loss on this transaction produces three separate answers.

Underneath that sits a small, unglamorous fact about how transaction paper works. A document does not carry one promise; it carries dozens, sorted into kinds, and the kind decides the machinery a claim has to run through before anybody writes a cheque. A warrantyA statement in the paper that something was true at a stated moment. If it was not, the party who stated it can be asked for money. It is worked in full under warranties. is a statement about the past with a price attached. A condition precedent is not a promise at all but a gate. A disclosure scheduleA separate list attached to the paper that carves stated exceptions out of what the seller promised. How one is read is covered separately. carves exceptions out of what was promised. The three remaining words a reader keeps meeting are indemnity, covenant and undertaking, and one fact run through all three is what separates them.

Summarising a document is a different job from drafting one. A summariser has to place each promise rather than choose it. Which rule of law applies, and which of these promises either side should have asked for, are separate questions. Both belong to a lawyer and to the two sides.

Where do the three filtering figures come from?

Three numbers have to be on the table before any of the promise words matter. The three numbers are what turns a loss into a payment or into nothing. Meeting them as rupees first and as fractions afterwards helps.

Rs 1.14 crore. Rs 11.37 crore. Rs 227.40 crore.

The first is a de minimisA floor under the claims process. A loss smaller than the floor is never entered on the list at all, so it cannot join the running total that decides payment later.: a loss below it is not even written down as a claim, so it can never join anything. The second is a basketA running total that the counted claims have to get past before any money moves. Below it the paperwork exists and the payment does not.: the claims that were written down have to add past it before a rupee moves. The third is a ceiling on the total the seller can be asked to pay under the promises it reaches: nothing above it is paid at all, and further loss stays with the buyer. Alongside them sits an escrowPart of the price parked with an outside holder for a stated stretch instead of reaching the sellers at completion, so money is already sitting somewhere if a claim lands. of Rs 113.70 crore. An escrow is not a filter but a funding arrangement, and how it compares with the alternatives is covered separately.

Where did those three come from? From one figure and three fractions of it. Rs 1,137 crore is where the price finished. The price got there from Rs 1,140 crore, an enterprise value of Rs 1,320 crore with the target’s Rs 180 crore of borrowings net of its cash taken off. The Rs 1,320 crore had been struck at ten times Rs 132 crore of earnings before interest, tax, depreciation and amortisation (EBITDA)A profit measure struck before four deductions are taken, those being interest, tax, depreciation and amortisation. Used here only to size the purchase; how the measure is built belongs to the valuation material.. The last Rs 3 crore is the two completion adjustmentsTwo corrections run once the transaction closed, one on working capital and one on net debt. Here they move in opposite directions and net to Rs 3 crore. settling against each other: Rs 12 crore added for working capital, Rs 15 crore taken away for net debt. A tenth of a per cent, a whole per cent and a fifth of that finished figure give the three numbers above.

Two things follow that a reader should refuse to be impressed by. The basket is exactly ten times the de minimis and the cap is exactly twenty times the basket. All three were struck as fractions of one figure, so neither of those multiples reveals anything. A ratio between two fractions of the same base is arithmetic, not evidence: it says the drafter wrote 0.1, 1.0 and 20.0, and it says nothing about what the drafter feared. The same goes for the escrow being exactly half the cap. A package shows what it was worried about in the promises the limits reach rather than in the ratios inside it, and the limitation clause below is where that list sits.

The second thing is smaller and sharper. Struck on Rs 1,140 crore instead, the basket would be Rs 11.40 crore, a difference of Rs 3,00,000/-. Three lakh looks like nothing. Now take a loss of Rs 11.38 crore, chosen deliberately because it lands in the gap: on the correct base it has passed the basket and the whole Rs 11.38 crore comes back, and on the wrong base it has not passed and nothing comes back. A three lakh error in a threshold decided eleven and a third crore of recovery, and it did so because a threshold is a cliff rather than a slope.

What does an indemnity actually promise?

An indemnity is a promise to pay for a named loss if a named thing happens. The promise stops there. An indemnity is written against a risk both sides already know about and have already talked about. Naming a known risk is exactly why an indemnity exists as a separate promise: nobody is pretending the risk is not there, so there is nothing to state as true and nothing to turn out false.

The household version has the same shape. A buyer takes a second hand scooter. The seller says the engine has never been opened up. The seller’s sentence is a statement about the past, and if the engine turns out to have been rebuilt twice, the statement was untrue and the buyer has something to complain about. Now suppose both sides already know there is an unpaid challan sitting against the vehicle from before the sale, and neither of them knows how much it will be. Nobody is lying about anything. The seller offers a promise instead: whatever that challan turns out to be, the seller pays it. The first promise is a statement that can be false. The second is a payment promise for something that was never in dispute.

A warranty ends in money because a fact turned out to be untrue, and an indemnity ends in money because a named event happened, and that is why a document carrying both writes them as two separate promises. The reader who conflates them will look for the untruth under an indemnity and will not find one, then conclude the clause is unusable. There is no untruth to find. There is a trigger, meaning the event the paper names as the one that has to happen before the obligation bites, and that trigger has either fired or it has not.

On the transaction in question, the risk that both sides could see was ordinary enough. Harivansh Packaging Limited, an invented buyer, was buying Sundarban Polymers Private Limited, an invented target, and two of the target’s contracts could not simply travel with the business: each needed the other side’s permission first. Everybody knew that before the paper was drafted. The permissions were not in hand, so nobody was going to state that they were. So the risk had to be handled by a promise about what happens rather than by a statement about what is, and that is the space an indemnity occupies.

What does a covenant promise, and what does breaking one produce?

A covenant is a promise to do something, or to refrain from something, over a period. Deliver the audited accounts each quarter. Do not sell the plant. Seek the two permissions. Keep insurance in place until completion. The subject matter is enormously varied and the shape never changes: somebody is on the hook for conduct, across time, and either the conduct happened or it did not.

Notice what that does to the direction of the promise. A warranty looks backwards at a fact and a covenant looks forwards at behaviour, and breaking one produces a claim in the same way a false warranty does, except that the trigger is conduct rather than a fact. The machinery downstream can be identical. The proof required upstream is not identical: a warranty requires showing that the world was not as described, and a covenant requires showing that somebody did not do what they said they would do.

The everyday version is a tenancy. The tenant signing it states that they are employed where they said they were. Employment is a fact about now. The tenant also promises to pay on the fifth of every month and not to sublet. Both of those promises run across the next eleven months. Both sit in the same document. Both can end in the landlord asking for money. Nobody would confuse the two. The first is about who the tenant is and the second is about what the tenant will do, and a transaction document is exactly the same distinction written at a much larger scale.

A covenant also carries a period, and the period is doing real work. A promise to keep insurance in place until completion is finished the moment completion happens. A promise not to solicit the target’s employees for a stretch after completion is only starting then. On a fast reading of a document, the period is the quickest way to see what a covenant is actually protecting: promises that run to completion are protecting the thing being bought, and promises that run after it are protecting what the buyer expects to do with it.

On this transaction, the seller promised to seek the two permissions. The promise ran across the nine week conditions period, so it is a promise about conduct across time. A promise about conduct across time is a covenant, and if the seller simply did not ask, there is a claim.

Covenant vs Undertaking: is there a difference to find?

One question gets asked more than any other in this corner of a document, and the honest answer is not the one most readers want.

Both words describe a promise to act or to refrain. In ordinary use they name the same kind of obligation. There is no general legal difference between the two words, and any difference inside a particular document is one that document created for itself. A reader who arrives expecting one word to be heavier than the other is looking for a distinction no general rule supplies, and will therefore find whatever they were already expecting.

The place to look instead is the defined termsThe section near the front of a document that fixes what particular words mean inside it. A word can mean something narrower or wider there than it does in ordinary speech.. The defined terms section is where a document sets out what its own words mean, and if it has made the two words do different jobs, it will have said so there. Failing that, the position of each one matters: the same word can behave differently depending on whether it is inside the operative clauses, in a schedule, or in a side letter, and position is a fact that can be checked.

Other documents are already read this way. On a loan sanction letter, the word charges might be defined at the top to mean four specific line items, and everything the letter later says about charges then means those four and nothing else. The ordinary English meaning of the word stops being an argument once the letter has defined it. A transaction document works the same way and simply does it for many more words.

There is one useful consequence for a person summarising a document. A memo that treats undertakings as a separate and lesser category from covenants, where the document never made that separation, has invented a structure and then reasoned from it. Inventing a structure is worse than being unsure. The memo now reads with confidence about something it made up. The safe line is short: the document uses both words, here is where each is used, and here is what the defined terms say about them.

Covenant vs Condition Precedent: what does a failure cost?

The contrast between a covenant and a condition precedent is the sharpest in the set, and it is sharp because the two can be written about exactly the same subject matter.

A covenant is a promise that can be broken. Breaking it produces a claim, and a claim is a request for money that then has to survive whatever machinery the document puts in its way. A condition precedent is not a promise at all but a gate. The gate either opens or it does not, and what it produces is a completion or no completion. Nobody breaks a condition. A condition is satisfied or it is unsatisfied, and there is no misconduct in the second case.

The same subject matter can be written either way, and the choice decides whether a failure costs money or costs the transaction. Read that twice. Most readers of a document have never had it put to them plainly. The permissions from the two contract counterparties could have been drafted as a promise by the seller to go and get them, or as a gate that has to be open before anybody completes. Same permissions, same counterparties, same commercial risk. Two entirely different consequences when the permission does not arrive.

Look at what actually happened here. Completion on this transaction hung on three gates. Had the regulator cleared it? Had the business stayed recognisably the one that was agreed on, with no material adverse changeA condition asking whether the business is still recognisably the one that was agreed on. What counts as bad enough belongs to the deal execution material.? And had both counterparties whose contracts were moving across said yes? Only the third gate does any work here, and it does a great deal. The counterparty permissions could equally have been a covenant.

The everyday version is a flat purchase. The agreement can be written so that the sale only happens if the buyer’s home loan is sanctioned, in which case an unsanctioned loan means no sale and the advance comes back. Or it can be written so that the buyer promises to obtain a sanction, in which case failing to get one is the buyer’s breach and the seller can ask for something. The house is the same house. The loan is the same loan. The difference is whether the paper turned the risk into a gate or into a promise, and it is felt entirely in what happens next.

One more distinction is worth carrying, and it catches people out. A condition that fails does not automatically mean anybody has done anything wrong. If the regulator simply takes a view, nobody breached anything. So a document that turns a risk into a gate very often adds a covenant beside it: the gate handles whether the transaction proceeds, and the covenant handles whether somebody made a genuine effort. The two are not alternatives so much as two different questions asked about one risk, and a document can ask both.

Why does one Rs 8 crore loss give three different answers?

Now put the machinery down and take one fact through it.

One of the two permissions is refused. The contract does not travel. The buyer loses the trade the contract was carrying, and when the loss is worked out it comes to Rs 8 crore. Nothing about that sentence changes across the three routes below. The refusal is the same refusal, the contract is the same contract, and the number is the same number.

Written as a condition precedent, the gate does not open, completion does not happen, and there is no money question at all in either direction. The buyer does not pay the price. The seller does not pay a claim. The business the loss would have fallen on was never bought, so the Rs 8 crore never gets computed.

Written as a covenant sitting inside the package, the claim is Rs 8 crore and the recovery is nothing. Work it through. Rs 8 crore is 0.70 per cent of the finished price, well clear of the de minimis, so the claim is counted and written down. Then the counted claims have to pass Rs 11.37 crore, and Rs 8 crore does not. So the claim exists, it is real, it is provable, and it pays nothing. The package has not failed. The package is doing precisely what it was drafted to do.

Written as a standalone indemnity that the document places outside the limits, the seller pays Rs 8 crore in full. The sentence that says which promises the filters reach does not name this one, so no filter touches it.

One event. One document. One loss. The answers are no transaction, nil, and Rs 8 crore, and the loss decided none of them.

The discipline matters more than the preference. Each route has been named for what it does. Which route is preferable is a negotiation between the two sides. The conclusion available is narrower and more useful: the amount was decided by drafting, so the drafting is what a summary has to describe.

Are those three answers even on the same scale?

The three answers just set out carry a problem worth stopping on, and a trio is exactly how this contrast usually gets written down.

No transaction. Nil. Rs 8 crore. Read as a list, that reads like three points on one axis, running from worst to best or from most to least. It is not. Two of those answers say how many rupees come back. The first answers a completely different question, whether the transaction happens at all. A yes or no about completion cannot sit beside two rupee figures and be called a comparison, and the temptation once it has been done is worse: somebody divides Rs 1,137 crore by Rs 8 crore, gets a large number, and announces that the condition route puts a hundred and forty times more at stake. The ratio is meaningless. The division has a price on top and a loss underneath.

So rebuild the trio on a measure all three routes actually publish. Ask each of them the same question: on this one refused permission, how many rupees is the buyer down against what it bargained for?

On one shared measure, the route that sounded like the disaster costs the buyer nothing on this loss, and the route that sounded like a mild disappointment is the only one that leaves a hole. The reversal is the whole reason for doing the rebuild. The unrebuilt trio invites a reader to treat no transaction as the heaviest outcome because it is the most dramatic sentence, and on the measure the other two are speaking, it is not an outcome of that kind at all.

Two honesties are owed before anybody uses this table. The first is that the condition precedent row is incomplete and cannot be completed here. Not buying the business is obviously not costless to a buyer that wanted it; there is a value to the purchase that simply does not happen, and this transaction record carries no figure for what Harivansh Packaging Limited expected the purchase to be worth to it. The value of a purchase that never happens is absent rather than zero, and writing a zero into the row would invent the most important number on the table. So the row says Rs 0 on this loss and says nothing at all about the transaction.

The second is that the two rupee rows are complete, and their completeness is exactly why the comparison between them is worth making. Both start from the same price actually paid. Both concern the same Rs 8 crore. The only difference between them is one sentence in the limitation clause, and the next block is about that sentence.

Which promises do the limits actually reach?

The cap, the basket and the de minimis do not float above the document applying themselves to everything. The three filters apply to a list, and the list is written down. Somewhere in the limitation clause there is a sentence saying which promises the limits apply to, and that sentence is doing more work than any heading in the document.

A promise inside the limits and the identical promise outside them are different amounts of money, so find the sentence that lists which promises the limits reach. On these figures the difference is the whole Rs 8 crore. The difference is total. Same loss, same wording in the promise itself, and the recovery moves between nothing and everything according to whether that promise appears on a list fifteen printed sides away.

The list has more than two settings, and a promise sits at exactly one of them. A promise can be inside every filter. Another can be inside the cap but outside the basket, so it pays from the first rupee and stops at Rs 227.40 crore. A third can be outside all of them. A fourth can be inside a filter with its own separate ceiling. A document is free to arrange this however the two sides agreed, and the only way to know which arrangement a particular document carries is to read it.

The limitation list is also where a reader’s instinct is least reliable. Indemnity sounds like a stronger word than warranty, so the instinct says that a promise called an indemnity must be the strong one. The instinct is reasoning from vocabulary, and the document is not written in vocabulary; it is written in cross references. A promise labelled indemnity that sits on the filtered list behaves, in money terms, exactly like a warranty. A promise labelled warranty that the list leaves off pays from the first rupee. The label is a filing convenience.

What does a reader check to tell them apart?

Four questions place any promise in a document, and all four can be asked in under a minute once the reader knows what to look for.

The four are what triggers it, over what stretch it runs, who has to show what, and whether the limits reach it, and the heading a promise sits under answers none of them. Ask them in that order. Each one narrows the next. A promise whose trigger is a stated fact being false is a warranty however the clause is titled. A promise whose trigger is conduct across a period is a covenant or an undertaking, and which of those two words the document used is a matter for the defined terms. A promise whose trigger is a named event, with no assertion behind it, is an indemnity. And a thing that produces no claim at all when it fails, only a transaction that does not close, was never a promise in the first place.

The fourth question is the one people skip and it is the one that changes the number. The fourth is also the only one whose answer lives somewhere other than the clause being read. The first three are answered by the promise itself; the fourth is answered by a list elsewhere. The asymmetry is exactly why the fourth gets forgotten: everything else needed was in the clause already open.

A word on the second check. The document plainly has periods; a promise to seek permissions ran across the nine week conditions period, and an escrow was held for eighteen months. But eighteen months of holding money is not the same measure as a window during which a claim may be brought, and this transaction record publishes no such window at all. The two durations answer different questions in the same way the three answers above did, and a summary that reports the escrow hold as though it were a claim window has made the same category error twice over. The second check stands: the document itself will answer it, and this transaction record does not.

What happens to the two routes as the loss gets bigger?

The block above worked one loss. The control below covers every loss between nothing and Rs 40 crore, and what it shows is not what most readers expect before moving it.

Watch the arrow between the two bars rather than the bars themselves. The arrow is what the one sentence in the limitation clause is worth at each setting, and the arrow does not simply grow steadily. The gap widens all the way up to Rs 11.37 crore, reaching its widest exactly at the basket, and then collapses to nothing the moment the basket is passed. The sentence deciding whether a promise is filtered is worth the most on claims that sit just under the basket and worth nothing at all on claims comfortably above it. The lesson is worth carrying into any document. The drafting choice matters most in the range where a reader is least likely to be paying attention, and the amounts there look small.

Two readings are worth taking from the control before leaving it. The first is that the filtered bar is flat on the floor across the whole first quarter of the travel, and it is flat for two different reasons that produce the same nothing: below Rs 1.14 crore the claim is not counted, and between there and Rs 11.37 crore it is counted and still unpaid. Both zones give nil. With a single claim, the de minimis changes no rupee at all. The basket standing ten times above it has already produced the same answer. The de minimis earns its place when there are many small claims. The floor decides which of them are allowed to join the running total that has to pass the basket. On one claim, it is a formality.

The second reading is about the jump. Take the control to Rs 11.37 crore and the filtered route pays nothing. Move one step further, to Rs 11.38 crore, and it pays Rs 11.38 crore. A step of Rs 1,00,000/- in the loss moved the recovery by more than eleven crore. A threshold does exactly that, and it is why the earlier point about the Rs 3,00,000/- base error mattered so much more than its size suggested.