Most contracts are read carefully at the front, price, scope, term, and skimmed at the back, where the lawyers put the clauses about what happens when things break. That is exactly backwards. The risk-allocation clauses near the end decide who carries the cost of a failure, and they are often the only part of the document that ever gets tested. This is a plain-English map of the three that matter most: warranties, indemnities, and the limitation of liability that caps them both.

The quick version

  • A warranty is a promise that something is true (the software works; you own the rights you're selling). Break it, and the other side can sue for damages, but they have to prove their loss and try to reduce it.
  • An indemnity is a promise to pay for a defined loss if it happens (we'll cover any third-party claim that our code infringes a patent). It's a faster, fuller route to money, often paid as a debt, without the usual hurdles of a breach claim.
  • A limitation of liability clause caps how much anyone can recover, and excludes certain losses entirely. It is usually the single most valuable clause in the contract, and the one negotiated last and fastest.
  • The law won't let you cap everything. In the UK and elsewhere you cannot exclude liability for death or personal injury caused by your negligence, and many caps must be "reasonable" to stand up.

The idea in depth

Strip away the jargon and these clauses answer one question: when this deal goes wrong, who pays, how much, and how easily? Each of the three answers a different part of it, so a leader who can tell them apart can read a contract's real risk in a couple of minutes.

Start with the warranty–indemnity distinction, because it is the one people blur most often. A warranty is a contractual statement of fact, "the goods are fit for purpose," "we own the intellectual property we're licensing." An indemnity is a promise to reimburse a specified loss if a specified thing happens. As the UK firm LegalVision puts it, warranties "promise that the good or service that you are providing is of a certain quality," whereas indemnities are "a contractual promise that one party will reimburse the other party if a particular event happens" (LegalVision, "Warranties and Indemnities"). The practical gap shows up when something breaks. Claim under a warranty and you sue for breach of contract: you must prove the breach caused your loss, that the loss wasn't too remote, and, crucially, you have a duty to mitigate, to take reasonable steps to limit the damage. Claim under a well-drafted indemnity and much of that falls away: the indemnity often lets you recover the defined loss more directly, as a debt, without the same mitigation and remoteness hurdles.

Which is why every indemnity you give deserves to be read as a serious commitment, not boilerplate, by granting it, you have signed away some of the natural defences a breach claim would otherwise have handed you. Ask three things of each one: what event triggers it, is it capped, and does it sit inside or outside the liability ceiling? An uncapped indemnity is the clause that can swallow a company.

flowchart TD
  A(["Something goes wrong"]) --> B{"Is there an indemnity
for this loss?"}
  B -->|"Yes"| C(["Claim the defined loss
often as a debt; fewer hurdles"])
  B -->|"No, rely on a warranty"| D(["Sue for breach of contract"])
  D --> E(["Prove causation + remoteness
and mitigate your loss"])
  C --> F{"Is it inside the
liability cap?"}
  E --> F
  F -->|"Yes"| G(["Recovery limited to the cap"])
  F -->|"No / carve-out"| H(["Recovery can exceed the cap"])
The same failure travels a different road depending on which clause covers it, and the cap sits at the end of both. Leaders Loop

Why the liability cap is usually the clause that matters most

Warranties and indemnities decide what you can claim. The limitation of liability clause decides how much, and it almost always wins. A typical cap ties your maximum exposure to a number: the fees paid under the contract, the fees from the last twelve months, or a multiple of contract value. Alongside the cap sits a list of excluded losses, usually "indirect" or "consequential" damages such as lost profits and business interruption, which the contract removes from recovery altogether.

That second part rests on a piece of law old enough to predate the telephone. In Hadley v Baxendale (1854), an English court set the test for recoverable losses in two limbs: losses that arise naturally from the breach, and losses the parties could reasonably have contemplated at the time of contracting because special circumstances were made known. A mill owner whose mill sat idle waiting for a delayed crankshaft lost his claim for lost profits, because he had never told the carrier the mill would stop. When a modern contract excludes "consequential loss," it is drawing a line around that second limb, the special, far-reaching losses, and saying: not our problem.

Warranties and indemnities decide what you can claim; the liability cap decides how much, and it almost always wins.

The practical rule is blunt: find the cap first. Before you argue about a single warranty, read the limitation clause and ask whether the cap is large enough to make the rest of the contract mean anything. A beautifully drafted indemnity is worthless if the cap sits below the loss it was meant to cover. Then watch the carve-outs, the items deliberately placed outside the cap (often the indemnities, breaches of confidentiality, and IP infringement), because those are the exposures that can run to the full extent of the damage. The clause that feels most technical is the one that sets your true downside.

An honest limitation. A cap is only as good as the law lets it be. In the UK, the Unfair Contract Terms Act 1977 makes it impossible to exclude liability for death or personal injury resulting from your negligence, void however it's worded, and other negligence-based exclusions only stand if they pass a statutory test of reasonableness. Other jurisdictions police gross negligence, fraud and wilful misconduct in similar ways. None of this is legal advice, and the rules differ by country and by contract type: the principle is durable, but whether any specific cap is enforceable is a question for a qualified lawyer in the relevant jurisdiction. The takeaway is humbler and still useful, a one-sided cap that tries to exclude the un-excludable can be struck down, leaving you more exposed than a fair one would have.

A worked example

Take a mid-sized retailer, call it Harbour & Co, buying an inventory platform from a software vendor for an illustrative £200,000 a year. (Illustrative figures throughout; this is a teaching example, not legal advice or real contracts.) The draft the vendor sends back looks generous on its face. It warrants that the software will perform materially in line with the documentation, and it indemnifies Harbour against any third-party claim that the software infringes a patent or copyright. Reassuring, until you reach the back.

There, the limitation clause caps the vendor's total liability at "the fees paid in the preceding twelve months", £200,000, and excludes all indirect and consequential loss, including lost profits. Now run a real failure through it. The platform miscounts stock during the Christmas peak; Harbour over-orders, under-sells, and loses an estimated £1.2m in margin and emergency logistics. Harbour reaches for the performance warranty, but its lost profits are the textbook Hadley second-limb loss the contract has explicitly excluded, and even its direct losses are capped at £200,000. The warranty was real; the cap made it small.

flowchart LR
  A(["Vendor warrants performance
+ indemnifies IP claims"]) --> B(["Cap: 12 months' fees
~£200k (illustrative)"])
  B --> C(["Excludes indirect /
consequential loss"])
  C --> D{"Peak-season
miscount: ~£1.2m loss"}
  D -->|"Lost profits"| E(["Excluded, not recoverable"])
  D -->|"Direct loss"| F(["Recovery capped at ~£200k"])
  G(["Negotiated fix:
carve IP indemnity out of cap;
raise cap for data/availability"]) -.-> B
The same warranty is worth a lot or a little depending entirely on the cap behind it, and on which losses are carved out of it. Leaders Loop

The fix isn't to demand unlimited liability, no vendor signs that. It's to negotiate the cap and its carve-outs deliberately. Harbour pushes the cap to a multiple of annual fees for failures that touch availability and data accuracy, carves the IP indemnity out of the cap, and narrows the consequential-loss exclusion so a defined band of operational loss survives. Same three clauses, a very different risk profile, and the conversation that produced it took an hour, because someone read the back of the contract first.

Frequently asked questions

What is the real difference between a warranty and an indemnity?

A warranty is a promise that a statement is true; if it isn't, you sue for breach of contract and must prove your loss and mitigate it. An indemnity is a promise to pay for a specified loss when a specified trigger occurs, often recoverable more directly and without the same duty to mitigate. In short: a warranty is "this is true," an indemnity is "if this happens, we'll cover it." The indemnity is usually the stronger protection for whoever receives it, and the heavier obligation for whoever gives it.

What does "consequential loss" or "indirect damages" actually mean?

Broadly, it's loss that doesn't flow directly and naturally from the breach but from your particular circumstances, classically lost profits, lost business, or wasted management time. The distinction traces to the two-limb rule in Hadley v Baxendale (1854). Because courts in different jurisdictions interpret these words inconsistently, good contracts define exactly which losses are excluded rather than relying on the label alone, an undefined "all consequential damages" exclusion is a common drafting trap.

Can a contract just cap all of one side's liability?

No. Caps and exclusions are limited by law. In the UK, you cannot exclude liability for death or personal injury from your negligence at all, and many other exclusions only hold if they are "reasonable" under the Unfair Contract Terms Act 1977. Fraud, and often gross negligence or wilful misconduct, are typically un-excludable too. A cap that overreaches can be struck out, which is why fair, clearly drafted caps tend to be the durable ones.

Should I worry more about the warranties or the liability cap?

Read the cap first. The warranties and indemnities define what you can claim; the cap defines the ceiling on all of it, and a low cap quietly devalues every promise above it. Then look at the carve-outs, the items placed outside the cap, because that short list is where your largest exposures actually live.

What's a "hold harmless" clause, is it the same as an indemnity?

They travel together, often as "indemnify and hold harmless," and in many contracts are treated as one obligation. The usual reading is that "indemnify" is the promise to reimburse loss, while "hold harmless" adds a promise not to hold the other party responsible. Interpretation varies by jurisdiction, and commentators like contract-drafting specialist Ken Adams have long argued the doubling adds little. Treat the phrase as a single risk-transfer mechanism and check what loss it covers and whether it's capped.

Related in the Toolkit

These clauses live inside a contract, so the place to start is contract fundamentals, what makes the agreement binding in the first place, and the loss you most often indemnify is third-party intellectual property infringement, which is why that indemnity is usually the one carved out of the cap.

Where to go next