The Item Broke After Eleven Months. Who Pays for It?

The message is always the same shape. It stopped working. I bought it from you last year. The tone is usually reasonable, the facts are usually incomplete, and the answer depends on three things the customer has no reason to know: when exactly they bought it, what the fault actually is, and which of several overlapping sets of rules applies.

Sellers lose money here in a specific way. Not by paying claims they should refuse, but by spending forty minutes on each one, applying a different standard every time, and never writing any of it down.

Is this a warranty claim or a legal right?

Usually a legal right, and the distinction decides everything else. A warranty is something you or the manufacturer chose to offer. A statutory right exists whether anybody mentioned it, and it cannot be signed away by a line in your terms.

In the UK the relevant machinery sits in the Consumer Rights Act 2015. Section 19 gives a consumer whose goods do not conform three routes: a short term right to reject, a right to repair or replacement, and a right to a price reduction or a final right to reject. The detail that matters most to a seller is in subsection 14. A fault that appears within six months of delivery is treated as having been present when the goods were handed over, unless you can show otherwise or the presumption does not fit the nature of the goods. The burden sits with you for half a year.

In the United States the structure is different. The Magnuson-Moss Warranty Act does not force anybody to offer a written warranty, but the FTC's guide to federal warranty law is clear that once you do offer one, you cannot disclaim or modify the implied warranties that come with the sale. A limited warranty may restrict them to its own duration. It cannot remove them.

How long does the customer actually have?

Three different clocks, and they overlap rather than run in sequence. Confusing them is the most common error in a support reply.

The first is the rejection window. Section 22 of the same Act sets it at 30 days, beginning the day after the consumer has ownership or possession, the goods have been delivered, and any installation you were responsible for is done. It pauses while goods are away for repair, and restarts with at least seven days left if the replacement is also faulty.

The second is the repair and replacement period, which extends well beyond 30 days and is where most of these claims live. The third is whatever written warranty exists, which may be shorter or longer than either and which is a commercial promise rather than a statutory one.

Reading across the rows shows why a single company policy rarely survives contact with a second market. A seller shipping from the UK into the EU is operating two sets of obligations on the same product, and a US seller with European customers is operating three. The workable answer is not three policies but one written to the most demanding standard that applies to you, since a customer who gets more than the minimum never complains about it.

Notice the row that changes next year for anyone selling into the EU. Under Directive 2024/1799, sellers must inform consumers that they can choose between repair and replacement, and choosing repair extends the liability period by a further 12 months on top of the two years already provided. Member states had to transpose it by 31 July 2026 and may go further.

What should you ask the customer for?

Everything you need, once. The single biggest cost reduction available in this whole process is replacing a sequence of questions with one structured request.

Three items cover almost every case. The order number or purchase date, because the date decides which clock applies. Photographs or a short video of the fault, because a description of a noise is worth very little and a recording of it is often conclusive. And a plain account of what happened immediately before the failure, which separates a manufacturing defect from damage, misuse or an accessory problem in about half of all claims.

Ask for the serial or batch number too where the product carries one. It costs the customer nothing and it is the field that later lets you tie six scattered claims to one production run, which is the difference between an unexplained cost and a supplier credit.

Put those three in a form rather than in an email. A form produces consistent records, which is what lets you spot the pattern later, and it feels less like an interrogation than three separate replies asking for one thing each. The same logic drives the way we set up returns decisions on real data, and a fault claim is simply a return with a diagnosis attached.

Where does software genuinely help?

In classification and drafting, not in deciding. The decision about whether something is a defect or damage carries legal and financial weight, and it should have a person's name on it.

Classification is the useful part. A model reading the customer's description along with a photograph can route a claim into likely categories, flag which ones name the same component, and tell you within a week that eleven of the last forty claims involve the same hinge. That pattern is the expensive thing to miss, because it turns a stream of individual costs into either a supplier conversation or a decision to stop selling the line. We covered the second half of that in how to decide which products leave the range.

Drafting is the second win. Warranty replies are formulaic, repetitive and easy to get wrong in tone, and the wrong tone is what turns a refused claim into a public review. A model writing the first draft against your own stated policy, with a human approving it, is a reasonable use of the technology. A model deciding the outcome is not, both because the rules differ by jurisdiction and because a wrong refusal costs far more than the part would have.

One caution specific to this area. Do not let an assistant quote warranty terms it has inferred rather than read. A promise your support channel makes can bind you even when it was wrong, which is the lesson from a tribunal that held a company to what its bot said, and warranty language is exactly the sort of detail these systems smooth over.

What can you require before honouring a warranty?

Less than many sellers assume. Conditions that look reasonable in a terms page can be prohibited outright.

The clearest example is the tie-in rule. Under the FTC guidance, you generally cannot require a consumer to buy or use an item or service from a particular company in order to keep warranty coverage. You may provide such items free under the warranty, and you may decline cover for damage genuinely caused by parts or servicing you did not supply, but requiring your own branded consumables as a condition of cover is the thing the rule exists to stop.

The full versus limited distinction is similarly mechanical. A warranty counts as full only if implied warranties are not limited in duration, service is available to any owner, service is free, the consumer may choose replacement or a refund after a reasonable number of failed repairs, and the only condition on service is telling you about the problem. Miss any of those and the document has to be titled limited, for products costing more than $10. The disclosure and pre-sale availability rules apply above $15.

There is a quieter trap in the pre-sale availability requirement. Warranty terms have to be available before the purchase, not only in the box, which for an online seller means on the product page or one click from it. A warranty that first appears in the paperwork after delivery has not been disclosed in the way the rules contemplate, and that is a surprisingly common arrangement among small sellers who simply pass on whatever the manufacturer supplies.

Proof of purchase deserves its own line. Requiring it is normal and defensible. Requiring the original packaging, a receipt rather than any evidence of purchase, or registration within a short window after sale are all conditions that tend to be either unenforceable or commercially unwise, and they generate complaints out of proportion to what they save.

Should you repair, replace or refund?

Cheapest total cost, with the customer's choice respected where the law gives them one. The instinct to replace everything is expensive, and the instinct to repair everything is slow, and neither is a policy.

Run the comparison honestly for each product class. A replacement costs you a unit at landed cost plus outbound shipping, and it closes the matter the same day. A repair costs parts, labour, two legs of shipping and a wait during which the customer is not happy and may escalate. For low value items the replacement is almost always cheaper once your own time is priced, and for high value or bulky goods the repair usually wins by a wide margin.

The legal overlay changes the calculation in Europe. Because the new repair rules require you to tell the consumer they can choose between repair and replacement, and because choosing repair adds 12 months to your liability, a repair you would have preferred on cost grounds now carries a longer tail. That is not an argument against repairing. It is an argument for knowing the failure mode before you agree to it, since repairing a symptom of a design fault buys you a second claim with an extended clock attached.

Refunds deserve a mention as the underrated option. Where a product has a claim rate you already distrust, refunding and removing the line costs less than servicing a queue of claims on stock you should not have bought. The decision belongs with your range review rather than with support, but support is where the evidence for it accumulates.

What about goods you did not manufacture?

You are still the consumer's counterparty, and sending them to the manufacturer is advice rather than an answer. That is the distinction small sellers get wrong most often, usually in good faith.

Statutory rights run against the trader who sold the goods. A manufacturer warranty sits alongside them as an additional, voluntary promise, and a customer may use either. When you tell someone to contact the maker, you are offering them a route that may be faster, not transferring your own obligation. If the manufacturer refuses or disappears, the claim comes back to you with your credibility already spent.

The practical handling is to do both at once. Accept the claim, deal with the customer yourself, and run your own recovery against the supplier in parallel on whatever terms your purchase agreement gives you. This is the same two track structure that applies when a carrier loses a parcel: resolve with the customer on your timetable, recover on the other party's.

Which makes your supplier terms worth reading before you need them. A supplier who will not accept defective stock back, or who caps returns at a percentage of volume, has moved a cost onto you that belongs in your margin calculation. If you only discover that during your first bad batch, the discovery is expensive. Checking it during supplier verification costs one question.

How do you keep this from eating your week?

Decide the policy once and spend your attention on the hard ten percent. Most claims divide cleanly into three buckets, and only one of them needs thought.

The first bucket is clear faults inside the early window, where the answer is a replacement or a refund and arguing costs more than the item. Approve those automatically below a value you set. The second is claims plainly outside any route, such as accidental damage with no warranty covering it, where a polite refusal with an offer of paid repair or a discount on a replacement is both honest and often profitable. The third is the genuinely uncertain middle, which is where your time belongs.

Track two numbers per product. The claim rate per hundred units sold, and the share of claims you uphold. A line with a rising claim rate is a supplier conversation. A line with a high claim rate and a low upheld rate is usually a listing problem, where the page is setting expectations the product cannot meet, and that is cheaper to fix on the page than in the warehouse. The same signal drives the work in reducing returns from the product page.

Review both numbers quarterly rather than reacting to individual claims, because a single bad unit says nothing and three in a month on the same component says everything. Quarterly also matches the rhythm at which supplier conversations can realistically be had.

Write the policy where it can be linked. A customer who can read what you will do, before they write to you, asks a better question and accepts the answer more readily. Keeping that page, the claim form and the order record in one place you control rather than spread across three tools is worth the small amount of work it takes in a storefront you can shape.

One last piece of arithmetic. Returns ran at 15.8% of US retail sales in 2025 according to the National Retail Federation, and fault claims are a small slice of that number. If you are spending more time on warranty claims than on the rest of your returns, something is wrong with either the product or the policy, and the claims themselves will tell you which if you keep them in a form you can count.