The seller says the machine is sound. The buyer cannot open it, cannot judge the workmanship, and cannot turn the seller’s confidence into knowledge merely by listening harder. Then the seller adds a promise: if it fails in the specified way, I will repair it or pay. The words have acquired a consequence. Failure, once wholly the buyer’s misfortune, will now reach the seller’s account. A warranty does not make every hidden part visible. It gives the buyer a claim against a later event, and gives the person making the assurance a reason to consider what that claim may cost.
The same promise can cost different sellers different amounts. Suppose the probabilities of a covered failure are one-tenth for a sound machine and four-tenths for a weaker one. A promise to pay one hundred on failure then carries expected claims of ten and forty. The arithmetic, retained from the earlier treatment, assumes that failure is verifiable, payment collectible, and other claim costs equal. If offering the guarantee raises a seller’s receipts by twenty-five relative to selling without it, the expected gain is fifteen for the sound machine and a loss of fifteen for the weaker one. Under those particular alternatives, one seller wants to offer the promise and the other does not. The buyer may have found a way to let hidden workmanship speak through a willingness to assume liability.
But different expected costs do not establish that only the sound seller can afford a warranty. A sufficiently high price can finance expected repairs to a poor product. A seller may offer cover as insurance, and a buyer may purchase it for protection rather than for the information it conveys. The price, coverage, exclusions, remedy and source of payment form one arrangement. Reading the length of the promise alone leaves out much of what determines its cost. What makes imitation unattractive is the entire comparison between available courses, not the mere existence of a liability. A guarantee can be a signal; the word “guaranteed” is not itself proof that the separating comparison holds.
Protection can succeed where revelation remains incomplete. Grossman made this distinction particularly clear. If a product provides a benefit worth one hundred and twenty when it works and twenty when it fails, a payment of one hundred on failure equalizes those benefits. At a price of ninety the buyer receives a net thirty in either state. He need not know whether failure has probability one-tenth or four-tenths to compare that fully protected payoff with his alternative. Under the model’s assumptions, the warranty can make the undisclosed probability irrelevant to his decision. It has not necessarily told him the quality of the machine; it has removed a specified financial consequence of not knowing.
The conditions carry the weight of that result. Grossman starts with an observable failure, known buyer benefits, risk-averse buyers, seller payouts that can be diversified, and quality fixed before the contract. He initially excludes buyer conduct that changes the failure probability. A real promise may restore the purchase price without restoring lost time, interrupted production, pain, or an object’s irreplaceable value. If those consequences remain, the buyer still cares how likely failure is. Nor does a promised payment remove uncertainty about whether the seller will perform it. The warranty replaces one question with a more specific set of questions: what is covered, what will arrive, when it will arrive, and what harm will remain after it does.
A warranty can also change the quality it seems merely to describe. Suppose an additional step in manufacture costs fifteen and reduces the chance of a covered failure from four-tenths to one-tenth. With a payment of one hundred at stake, the seller saves thirty in expected claims by spending the fifteen. The promise gives him a financial reason to improve production. This is an action incentive, distinct from revealing a quality already fixed. The distinction matters because the buyer may benefit from a guarantee even when no seller began with private knowledge of a superior machine. A future obligation can help create the reliability on which the present sale depends.
The buyer’s conduct belongs in the same account. A maker can take care over a motor while its owner can take care over its use. Full protection against breakdown may strengthen the maker’s incentive and weaken the user’s financial reason to maintain the machine. This does not tell us to abandon the warranty. It tells us to ask which failures reveal whose action, which precautions can be specified, and which risk can be transferred without spoiling a useful incentive. A maintenance requirement, a limited remedy, or a retained share of loss may answer part of that question. Each also narrows the protection, and each can become a source of dispute. The bargain must govern a relationship between actions, not merely announce confidence in a product.
The financial hostage must be within reach. A warranty creates an obligation; a funded bond places assets behind an obligation. Collateral identifies property whose loss can be made contingent on a specified failure, while a third-party guarantee adds another promisor whose own ability and duty to pay must be assessed. These devices are related without being interchangeable. Money stated in a contract is not money already held for the beneficiary. An asset is not a usable remedy until the relevant event can be established and the agreed enforcement can operate. The credible stake is the loss that can actually follow a breach, not the most impressive number written beside it.
The distinction is visible in the first example. If a seller expected to pay only one-quarter of the promised claims, its expected costs would fall from ten and forty to two point five and ten. The supposed twenty-five advantage from offering the promise would then exceed the cost for either quality. The earlier separating comparison would be gone. The figures show why the buyer’s inference depends on performance as well as words. A larger nominal guarantee can be less informative than a smaller promise with a dependable remedy, because the burden of the former may never reach the person who offered it.
A promise gives assurance through the consequence that can follow its failure, whether that consequence reveals quality or makes uncertainty bearable.
Reputation can help supply the missing connection. A seller expecting valuable future custom may honor an obligation because refusing it would damage the name under which he must trade again. The Collected Witness treated that future business as another form of stake. But a continuing name and an enforceable claim answer different questions: one makes performance attractive, the other defines what may be demanded. A buyer can sensibly want both, just as he can want an independent inspection alongside a warranty. No single device needs to carry the whole burden of trust. Their usefulness lies in the different gaps they can bridge.
The seller answers the buyer’s doubt by making something happen if the assurance fails. That consequence may reveal a private quality, insure a loss without revealing it, or encourage the care that prevents it. The task is to determine which of these things the particular bargain can accomplish. The promise has value where the covered event is intelligible, the remedy useful, and the obligation credible. Its limits begin where those connections break. A market need not wait until the buyer knows everything the seller knows; it can arrange for the remaining ignorance to have a different cost. The next witness is another person, whose seal offers a judgment where the buyer cannot make his own.