How to Set a Warranty Period You Can Live With

Why this matters

Most shops pick their warranty period the way they pick a font: somebody used twelve months once, it looked normal, and it stayed. Then a competitor advertises two years and the number gets changed in an afternoon by whoever is editing the proposal template. Neither the original choice nor the change was made against any evidence about when this shop's work actually fails, which means nobody knows whether the term covers the failures the shop causes, stops short of them, or extends well past them into failures somebody else caused.

The period is a cut point on a distribution you already own. This walks through building the term as a document, field by field, with the duration derived rather than borrowed.

Field 1: The covered population

Name the work the term applies to before you name the length. A single shop-wide period applied to everything is the most common source of an unaffordable warranty, because it prices the riskiest job type and the safest one identically.

Write it as the specific class of work: full installations of a given type, repairs on customer-owned equipment, maintenance visits, diagnostic-only visits. Diagnostic-only work in particular deserves its own answer, because what you are warranting there is a conclusion rather than a physical result.

Skip this field and every later field becomes ambiguous, since "twelve months" means nothing until you know twelve months on what.

Field 2: The start trigger

Substantial completion, commissioning, first use, or final payment. Pick one and write the words, because the difference between them on seasonal equipment can be three months of extra live exposure on a twelve-month term. The trade-offs belong to the sibling on which warranty terms drive cost; the field itself just has to be filled in rather than left implied.

One practical note: whichever trigger you choose, the date has to be recorded somewhere your office can retrieve it two years later without asking a tech. A trigger nobody can evidence is not a trigger.

Field 3: The duration, derived

This is the field the article exists for. Four steps.

Pull your event history with two dates on each event. The completion date of the original job and the date the warranty event opened. The age of the job at failure is the difference, and it is the only variable that matters here. Twenty-four months of history and at least thirty events is the minimum sample worth cutting against; below that, see the last section.

Attribute each event. Separate events caused by your own work from events that arrived under your warranty but were caused by something else: wear, a third party, misuse, a manufacturer defect, or a condition that was there before you. This attribution is the whole exercise, because the term should cover your causation and stop before the rest.

Build the cumulative curve of your own workmanship events by job age. Not calendar months. Age of the job at the time of the event.

Apply the cut rule. Set the term at the job age by which at least 90 percent of your workmanship-attributed events have arrived, measured over at least 24 months of history with at least 30 events, and then round up to the nearest conventional term of 3, 6, 12 or 24 months. Round up rather than to the nearest, and move only in those conventional steps: a term of 9 or 15 months is unquotable, unmemorable, and reads to a customer as though it was reverse-engineered from something.

Field 4: The remedy

What you will do, in one sentence, in verbs. Repair the defective work. Replace the component. Return to correct. The verb choice moves per-event cost by a multiple, far more than the duration does, and a longer term with a tight remedy is usually cheaper to carry than a short term with an open-ended one.

Two things to keep out of this field unless you have priced them: any promise to make good damage to other property, and any promise about system performance rather than the workmanship of your installation.

Field 5: The exclusions pointer

The exclusions belong in their own block, and what a fair one has to say is its own subject. In the term sheet this field just points to that block and does one job: it confirms that every exclusion listed there is testable at the time of a claim. An exclusion you cannot test at the counter is not a boundary, it is an argument, and arguments are the expensive part of a warranty book.

Field 6: Notice and access

Two conditions, both reasonable, both frequently missing.

Notice. The customer tells you within a stated window of discovering the problem. This exists so a small failure does not become a large one while nobody calls, and the window should be short and human: promptly on discovery, and in writing for anything involving water, heat, or a burning smell.

Access. You get a reasonable opportunity to inspect and to correct the work before anyone else touches it or before the customer hires someone else and sends you the bill. This is the single most valuable line in the whole document, because it protects the one thing you cannot recover after the fact, which is the evidence of what actually failed.

Field 7: Transfer

Whether the term survives a change of owner. If yes, say whether the new owner has to register the transfer and how, because a promise to a person you have never met, on a job you cannot pull a record for, is where warranty administration goes badly.

Field 8: What happens at expiry

State it. Coverage ends, and here is what we offer afterwards. A term with no stated ending is read as an implied ongoing relationship, and the call that comes at month fourteen is much easier when the document already named the boundary and the alternative.

The filled-in sheet, and the data behind field 3

A shop pulls 24 months of warranty events on one job type: 41 events, each dated against the completion date of its job.

Attribution comes next, and it changes the denominator. Both of the late events were traced to causes outside the shop's work: one was ordinary wear on a component the shop did not supply, and one was damage from a third party who worked on the same system afterwards. So the workmanship-attributed population is 39 events, not 41.

Against 39, the cumulative shares are 27 at month 3 (69 percent), 35 at month 6 (90 percent to the nearest whole number, 89.7 exactly), and 39 at month 12 (100 percent).

Now run the shop's own rule. The gate is at least 90 percent of workmanship-attributed events. Month 6 sits at 89.7 percent, which does not clear it. So the cut point falls between month 6 and month 12, and the round-up-to-conventional rule lands on 12 months.

That is a close call and it should be read as one. A single event moving between bands flips it, which is what a 39-event sample buys you. The tie-break that makes 12 the comfortable answer is the marginal cost: months 7 through 12 carried 4 events across six months, which is a low incidence band, so the extra half-year is cheap to cover and easy to advertise. Had those six months carried 15 events, the same rule would have produced the same cut point while telling a completely different story about the shop's workmanship, and the right response would have been to fix the work rather than to shorten the term.

What the two late events settle. They are the argument against extending to 24 months. Neither was caused by the shop, so a 24-month term would not have prevented a single failure; it would have bought two conversations in which the shop explains why a covered-looking failure is not covered. That is the hidden cost of a long term: not repair hours, argument hours, and they land on the office rather than on the crew.

What would flip this. If both late events had traced to the shop's own work, the workmanship population becomes 41, the month-12 cumulative is 39 of 41 or 95 percent, and the gate is met at 12 anyway - but the shape of the tail would now be telling you that your failures keep arriving after a year, which is a different diagnosis entirely and usually points at a specific method or material rather than at a term length.

The filled sheet for this job type, then: full installations of that type; term starts at first use, with the commissioning date recorded on the job record; 12 months; remedy is repair or replacement of the defective work at our option; exclusions per the attached block, each testable at the time of claim; notice promptly on discovery and in writing where water or heat is involved; we get first opportunity to inspect and correct; transferable once on written notice within 30 days of sale; at expiry, coverage ends and a maintenance agreement is offered.

Where the law sets the floor rather than you

The number you derive is a commercial choice sitting on top of legal duties you do not get to choose, and the two are easy to confuse.

The Magnuson-Moss Warranty Act, 15 U.S.C. 2301 and following, governs written warranties on consumer products, and where it applies it requires the warranty to be designated "full" or "limited," with federal minimum standards attaching to a full one at 15 U.S.C. 2304. Whether it reaches you is the first question and it is not obvious for a service business: a warranty on services alone is generally outside it, one covering parts and labour on a consumer product may be inside it, and the designation duty carries a low dollar threshold set in the statute and the FTC rules. Get that answered for your work by your own attorney before you pick a word.

There is also a floor underneath your written term that you did not write. Implied warranties can arise by operation of your state's law whether or not you say anything; whether they can be disclaimed and how far differs by state and between consumer and commercial customers, and where the federal warranty scheme applies to a written warranty on a consumer product it separately restricts a warrantor's ability to disclaim them. This is the part shops most often do not know exists, and it is a one-time question for your own attorney rather than a per-job one.

Beyond that, some states set statutory warranty periods for particular categories of work and some licensing boards impose their own minimums as a condition of the licence. State statutes of limitation and, for improvements to real property, statutes of repose set outer limits on claims regardless of what your document says. Both the length and the event that starts the clock vary substantially between states, so get the number for your states rather than assuming a typical one.

None of that is resolvable in an article, because the answers depend on your state, your licence, and whether your customer is a consumer or a business. Ask your own attorney once, in advance, and set your commercial term above whatever floor comes back.

When your sample is too thin to cut against

Below about thirty attributed events you do not have a distribution, you have anecdotes, and cutting a term against them will produce a number with false authority.

Three things to do instead, in the meantime. Adopt the conventional term for your trade and market, which is a defensible default rather than a derived answer, and say so internally so nobody mistakes it for evidence. Start recording the two dates and the attribution on every event from today, because the sample only exists if somebody codes it. And review at a fixed interval, annually is enough for most shops, rather than in response to a single bad quarter, since a term changed reactively after two ugly events will get changed back after two clean ones and the customers who saw both versions will never trust either.

References

• Magnuson-Moss Warranty Act, 15 U.S.C. 2301 et seq., including the full-versus-limited designation requirement at 2303 and the federal minimum standards for a full warranty at 2304
• See related: What Your Own Warranty Is Actually Underwriting, The Warranty Terms That Decide What It Costs You, What a Warranty Exclusion Has to Say to Be Fair