Extended Warranty and Service Contract Research: Testing What Buyers Think They Bought

Answer first: an extended warranty is usually not a warranty. Under 16 CFR 700.11(c), an agreement that would otherwise be a written warranty becomes a service contract the moment it requires consideration beyond the purchase price or is entered into after the sale, and in many states the same document is regulated as insurance. Buyers do not perceive any of this. They perceive a longer version of the coverage they already have, bought under thirty seconds of checkout pressure, and they routinely pay for a first year that the manufacturer warranty already covered. Every one of those beliefs is testable in an interview that takes a buyer six minutes. Platforms like Koji run it as an AI-moderated conversation where "I think it covers everything" is met with an immediate follow-up asking what the buyer expects to happen when the screen cracks, which is the answer that actually predicts the claim dispute.

Two documents, one impression

The Magnuson-Moss Warranty Act recognises two instruments that can look identical to a shopper. 15 U.S.C. 2306(b) permits a supplier to enter into a service contract "in addition to or in lieu of a written warranty if such contract fully, clearly, and conspicuously discloses its terms and conditions in simple and readily understood language." That is the same undefined comprehension standard the Act applies to warranties in 2302(a), examined in warranty comprehension research, and it is undefined here for the same reason: the statute names the outcome and supplies no test.

What separates the two instruments is not what they promise. It is 16 CFR 700.11(c):

An agreement which would meet the definition of written warranty in section 101(6)(A) or (B) ... but for its failure to satisfy the basis of the bargain test is a service contract. For example, an agreement which calls for some consideration in addition to the purchase price of the consumer product, or which is entered into at some date after the purchase of the consumer product to which it applies, is a service contract.

Read that carefully and the classification rule turns out to be about the transaction, not the promise. Identical coverage language becomes a warranty when it is bundled into the price and a service contract when it is charged separately. 16 CFR 700.11(b) puts the same point positively: for a written warranty, the undertaking must be conveyed at the time of sale and the consumer must give no consideration beyond the purchase price.

There is a third possibility on top of that. 16 CFR 700.11(a) acknowledges that some agreements meeting the statutory definitions are "sold and regulated under state law as contracts of insurance," and points to the McCarran-Ferguson Act, 15 U.S.C. 1011 et seq., under which most federal laws are not construed to supersede state insurance regulation. Automobile breakdown policies are the rule cited example. So the product a shopper thinks of as "the extra warranty" may sit under warranty law, contract law, or insurance law, and the shopper has no cue at the register telling them which.

The research question follows directly. If the legal identity of the product turns on facts the buyer cannot see, what does the buyer think they bought?

The overlap nobody discloses and everybody sells

Start here, because it is the single fastest finding to produce and the hardest to argue with.

Most extended service contracts sold at the point of sale are described in months or years of total term. A three-year plan on a product carrying a one-year manufacturer warranty delivers, in practice, two years of incremental coverage and one year of duplicate coverage. Whether that duplication is meaningful depends on the terms, and sometimes the service contract genuinely adds something in year one. Often it does not.

The measurable fact is not whether the overlap is justified. It is what the buyer believed. One structured question settles it:

When does the plan you just purchased begin providing coverage that the manufacturer warranty does not already provide?

Offer a single_choice with the day of purchase, after the manufacturer warranty expires, and not sure. The gap between the share who answer "day of purchase" and the actual answer in your terms is the overlap blindness rate, and it is a number your merchandising team has almost certainly never seen. It matters commercially for a reason beyond fairness: a buyer who believes coverage started today is a buyer who will file in month three, be told to go to the manufacturer, and experience that redirection as a denial. You paid for the acquisition and inherited the complaint on someone else coverage.

The second measurable belief is the deductible or service fee. Plans frequently carry a per-claim charge, and buyers who describe the plan as covering the product "for free" after purchase have formed a belief that guarantees a bad first claim. Ask what the buyer expects to pay when they file, before you tell them.

The upsell has a legal consequence buyers never see

This is the finding most product and legal teams do not have on their radar, and it changes the risk profile of the checkout attachment itself.

15 U.S.C. 2308(a) prohibits a supplier from disclaiming or modifying any implied warranty to a consumer where either the supplier makes a written warranty, or where "at the time of sale, or within 90 days thereafter, such supplier enters into a service contract with the consumer which applies to such consumer product." Under 2308(c), a disclaimer made in violation of that section "shall be ineffective for purposes of this chapter and State law."

So the act of selling a service contract in the first ninety days removes your ability to disclaim implied warranties on that product. The upsell is not risk-neutral. It converts a sale that might have been made "as is" into one carrying the full implied warranty of merchantability, the topic of implied warranty research. 2308(b) leaves a narrower door open, allowing implied warranties to be limited in duration to the duration of a written warranty of reasonable duration, but only where the limitation "is conscionable and is set forth in clear and unmistakable language and prominently displayed on the face of the warranty."

Count the empirical standards in that one sentence: conscionable, clear and unmistakable, prominently displayed. Three tests, none defined, all of them facts about how a real person reads a real document. A legal reviewer can tell you the sentence exists and is in bold. Only a buyer can tell you whether it was unmistakable.

The decision happens in thirty seconds under pressure

The other reason this category is under-researched is that the purchase context is hostile to every traditional method. The extended warranty decision is made at the end of a transaction the buyer has already mentally completed, often while someone waits, with a price anchored against a purchase that just made the number feel small. It is a decision made in seconds, and the buyer cannot reconstruct it accurately a week later.

Three things follow for study design.

Reconstruct the moment, do not ask about it in the abstract. Show the actual offer screen or script and ask what the buyer would do, rather than asking whether they generally find extended warranties worthwhile. Attitudes toward the category and behaviour in the moment are different measurements and they disagree.

Interview the decliners, not only the buyers. The people who said no hold the information about the offer, and they are entirely absent from a post-purchase satisfaction programme that only contacts plan holders. A study drawn from plan holders will conclude that buyers understood the plan, because the ones who did not understand it walked away.

Get to the claim, not just the purchase. The comprehension gap is realised at the first claim, not at the register. Two studies, one at purchase and one after the first claim attempt, produce a before-and-after on the same beliefs and turn a comprehension score into an outcome.

Building the study

Keep it under eight minutes and use the six structured types so beliefs are scoreable rather than impressionistic. The full set is documented in structured questions in AI interviews.

Two design notes worth following. First, write the covered-failures list from your actual exclusions, not from imagination. Accidental damage, wear parts, cosmetic damage, batteries, and software faults are the usual boundary cases, and a buyer who sorts all five into covered has told you exactly which claim conversation is coming. Second, ask who they would contact first. Buyers who name the retailer for a manufacturer-warranty fault, or the manufacturer for an accidental-damage claim under the plan, will experience one redirection before anything is resolved, and redirections are where satisfaction with a plan collapses.

Why an AI interview fits this problem

The measurement you want is a belief plus its reasoning, at a volume large enough to be segmented by channel, price band, and whether the buyer accepted or declined. Traditional methods force a choice between the two. A moderated interview gets the reasoning from twelve people and cannot cover the segments. A survey covers the segments and returns "it covers most things," which is not an answer you can act on.

An AI-moderated interview removes the trade. Koji asks the structured question, scores it, and then generates the follow-up from what the buyer actually said, so every respondent gets the probe that a good moderator would have asked. Ask what covered means and a buyer says "everything that goes wrong"; the follow-up asks whether that includes dropping it in water, and now you have the exclusion boundary rather than a slogan. Interviews run by voice or text, in the buyer own time, with no scheduling, and the report assembles itself as responses arrive.

Practically that means a plan-comprehension study can be fielded to buyers and decliners across every price band in a couple of days, at a cost that makes it repeatable each time the plan terms or the checkout script change. That repeatability is the real unlock: this is not a one-time study, because the offer changes and the answer changes with it.

Reading it against the terms, not against intuition

Score comprehension against your own contract, the way warranty comprehension research scores against the warranty document. Build the answer key first, from the terms, and have someone who did not write the questions confirm it. If your own team cannot agree on whether a scenario is covered, that disagreement is the finding, and no amount of buyer research will fix a term your own staff read two ways.

Report confidence and accuracy together. A buyer who is wrong and unsure will call before filing and can be redirected cheaply. A buyer who is wrong and certain files, gets denied, and disputes, which is where the coverage misunderstanding turns into the pattern described in chargeback research. The confident-and-wrong quadrant is the one to size and the one to fix, and it is fixed in the offer script, not in the claims department.

Finally, remember what the whole category is: a promise about a future failure, sold to someone who is currently happy. The honest version of that transaction requires the buyer to understand what they are covered for. The measurement of whether they do is not expensive, and it is the difference between a profitable attachment and a deferred complaint.

Frequently asked questions

Is an extended warranty legally a warranty?

Usually not. Under 16 CFR 700.11(c), an agreement that would otherwise qualify as a written warranty is a service contract if it requires consideration beyond the purchase price of the product or is entered into after the purchase. Since extended coverage is by definition paid for separately, most of what is marketed as an extended warranty is a service contract, and in many states it is sold and regulated as insurance instead, which 16 CFR 700.11(a) acknowledges through the McCarran-Ferguson Act.

Does selling a service contract change our other warranty obligations?

Yes, in one significant way. 15 U.S.C. 2308(a)(2) prohibits disclaiming or modifying implied warranties where the supplier enters into a service contract with the consumer at the time of sale or within 90 days after it. 2308(c) makes a disclaimer that violates the section ineffective under both federal and state law. Limiting implied warranties to the duration of a written warranty remains possible under 2308(b), but only where the limitation is conscionable, in clear and unmistakable language, and prominently displayed.

What is the fastest useful thing to measure?

The overlap blindness rate. Ask buyers when the plan they just purchased starts providing coverage the manufacturer warranty does not already provide, offer day of purchase, after the manufacturer warranty ends, and not sure, and score against your actual terms. It takes one question, and the share who believe coverage begins immediately predicts the volume of early claims that will be redirected to the manufacturer and experienced as denials.

Should we interview people who declined the plan?

Always. Decliners hold the information about how the offer was understood, and they are structurally invisible to any programme that only contacts plan holders. A study of buyers alone will systematically overstate comprehension, because the buyers who did not understand the offer are the ones who walked away from it.

How do we test this without biasing the answer?

Show the real artefact and ask what the buyer expects before you correct anything. Free-text first, correction last. If you present the accurate explanation and then ask whether it was clear, you have measured agreement with your explanation rather than comprehension of your document. Koji supports this by letting you upload the actual plan text and offer script into the study and reveal them in a controlled order.

How often should this study be rerun?

Every time the plan terms, the price, or the checkout script changes, and at least annually otherwise. Comprehension is a property of a specific presentation, not a permanent property of the product, so a result from a previous version of the offer tells you very little about the current one. AI-moderated studies make the rerun cheap enough that this is realistic rather than aspirational.

---

Ready to find out what your plan buyers think they bought? Sign up for Koji and get 10 free credits to test your actual plan terms with real buyers and decliners. Upload the offer script, add your covered and excluded scenarios as structured questions, and read a scored comprehension report in hours.

Related Resources

• Structured Questions in AI Interviews - the six question types that make coverage beliefs scoreable
• Warranty Comprehension Research - the same undefined-standard problem in the underlying warranty
• Implied Warranty Research - the implied warranties a service contract sale prevents you disclaiming
• Lemon Law Research - what happens when repeated repairs fail under either instrument
• Right to Repair Research - the repair-access questions that sit behind every service plan
• Money-Back and Satisfaction Guarantees - testing a refund promise made at the same moment
• Chargeback Research - where a denied plan claim usually ends up