How to Decide Whether a Callback Is Warranty Work

Why this matters

Almost every shop makes this decision in the worst possible place: standing in a customer's house, by the technician, under time pressure, in front of the person the answer costs money. The tech has no access to the original job record, no idea what the term document says, and a direct personal stake in the answer, because one classification writes off their afternoon and the other bills it. Then the office reverses the call two days later on the invoice, which is the version the customer remembers.

Three of the four determinations in this sort can be made from the record before anyone drives out. This is how to move them there, and what to do with the one that genuinely cannot move.

Step 0: The hazard screen runs before any classification

Before the sort starts, the person taking the call screens the reported symptom for a hazard, because a hazardous callback is dispatched on its hazard and classified afterwards.

If the caller reports a gas odor, tell them on that call: everyone leaves the building immediately, no switch is touched on the way out, no light turned on or off, no phone used inside, and they call the gas utility from outside. If they report smoke, burning smells, a hot panel or scorching, tell them to stop using the affected equipment and keep away from the panel, and dispatch. If they report water near energized equipment, tell them to keep everyone away from both and not to operate anything. If there is a main disconnect outside the building, or a breaker they can reach without approaching the water and without standing on a wet floor, they may turn it off; if there is any doubt at all, they leave it alone and call the utility. Then dispatch.

None of those calls involve the word warranty. Whether the visit is covered is decided later and it is decided at leisure. A shop that makes a customer wait for a coverage answer while a hazard is live has failed at something much bigger than a callback.

Step 1: Same work, or different work

Pull the original job record and compare what is being reported against what you actually did.

If the reported symptom is on a system, component or area you did not work on, this is a new service call. Say so on the phone, quote it as new work, and stop; the sort is over. This resolves more callbacks than most shops expect, and resolving it on the phone rather than on the driveway saves the customer an argument they were never going to win and saves you a visit you would have had trouble billing.

The failure mode here is confusing proximity with causation. Work in the same room, or on the same system a week earlier, is not the same as work on the failed component, and a tech who concedes coverage because the timing looks bad has given away a billable call. The counterpart error is worse: a shop that denies on this ground without reading the original scope will eventually deny something it did cause.

Step 2: Inside the term, on the recorded trigger date

Read the term and the trigger from the document you actually issued, not from memory, and get the start date from the job record rather than from the invoice date, which is a different date under most triggers.

Outside the term, the sort ends and the visit is billable, and that is a decision to state cleanly rather than apologetically. Goodwill remains available as a separate, deliberate choice, and the sibling articles on when goodwill beats the letter of the warranty cover when to reach for it. What matters procedurally is that goodwill is chosen with the classification known, not used to avoid making it.

This step is where shops discover that their trigger date was never recorded anywhere retrievable. If the office cannot answer "when did the term start" in under a minute, that is the finding, and it is fixable only on future jobs.

Step 3: Does a testable exclusion apply to the reported symptom

Run the reported symptom against your exclusion list, but only against the exclusions that turn on an observable fact. If the customer's own description already establishes the excluded condition, the sort can end here.

Only observable exclusions can be applied at this stage, and that is the point. An exclusion that turns on a judgment about the customer's behaviour cannot be resolved from a phone call and should not be attempted on one, because a denial issued over the phone and then withdrawn on site costs more relationship than the visit was worth.

Step 4: The causal path, which needs the site

Whatever survives steps 1 through 3 is a genuine causation question and it requires a technician at the equipment. Did the failure arise on the path of the work you performed, or somewhere else?

Two rules make this survivable at the customer's house.

The tech arrives with a provisional classification, not a blank page. The office states it in the dispatch note and in the confirmation to the customer: based on our records this looks like covered warranty work, and the tech will confirm on site. The customer now has an expectation set by the office, and the tech is confirming a decision rather than making one.

Overturning is asymmetric on purpose. A tech may overturn toward covered on their own judgment, immediately, without a call. Overturning toward billable requires a call to the office and a photograph of the evidence before the customer is told. That asymmetry is deliberate for two reasons: the toward-billable direction is the one that damages the relationship if it turns out to be wrong, and it is the direction in which the tech's own incentive is misaligned, because their afternoon gets billed instead of written off.

The written response, and why it is not optional

Whatever the classification, put it in writing to the customer the same day, in three sentences: what was found, which term or exclusion it falls under, and what happens next.

There is a legal reason on top of the commercial one. A number of states have notice and opportunity to repair statutes, under which a consumer's construction-defect claim may require written notice to the contractor and a chance to inspect and cure before a suit can proceed, and whether your work falls inside such a statute depends on your state and on how the work is characterized. Your side of that is simple and worth doing regardless: respond in writing, and ask in writing for the opportunity to inspect and correct. Whether the statute applies to you is a question for your own attorney, once, not per claim.

Worked example: one callback, twelve minutes of triage

The call comes in on a Tuesday. Reported symptom involves no odor, no smoke, no water near energized equipment, so the hazard screen passes and the sort proceeds.

Step 1. The original job record shows the shop installed the affected system. Same work, so the sort continues.

Step 2. Term is 12 months from commissioning. The commissioning date is on the job record, and the callback is at month 9. Inside the term, so the sort continues.

Step 3. The reported symptom does not match any of the shop's five testable exclusions on its face. Nothing to apply, so the sort continues.

Total office time: about 0.2 hours. Provisional classification: covered. The customer is told that in the appointment confirmation, in one sentence.

Step 4. The tech finds that the failure originated in a component upstream that the shop did not supply and did not touch, and that the symptom the customer reported is the downstream consequence. The causal path is not the shop's work.

This is an overturn toward billable, so the tech photographs the failed component in place, calls the office, and the office authorizes before anything is said to the customer. The customer is told on site, with the photograph on the tech's screen, before any further work: this is not our installation and the cause sits upstream of it, so this visit is billable, and here is what repairing it involves. The visit bills at 2.5 hours and the shop quotes the upstream repair as new work.

Two things to notice about how that landed. The customer never received two different answers, because the only answer they got before the overturn was "this looks covered, the tech will confirm," which is what actually happened. And the tech never had to argue, because the office authorized the change and the photograph carried the explanation.

What the triage step is actually worth

The shop above ran 36 callbacks in the twelve months before it added the office triage step. Eleven of them were classified on site by the tech and later reclassified by the office, which is 31 percent, and each reclassification cost roughly 1.0 office hour in corrected invoices, calls and apologies, so about 11 office hours in the year, plus three complaints serious enough that the owner got involved.

In the twelve months after, on 34 callbacks, reclassification fell to 2, which is 6 percent. The triage itself cost about 0.2 hours per callback, so roughly 7 office hours across the year.

Set the two against each other honestly: about 7 office hours spent to avoid about 11 office hours of correction, both in the same currency, so a net saving of around 4 office hours a year. That is a real but small number, and it is not the reason to do this. The reason is the three complaints. A classification the customer heard first from the office and saw confirmed by the tech does not produce the reversal call that turns a routine warranty visit into a grievance, and no hours figure captures what that is worth on a book built on referrals.

What would change the recommendation. A one-tech shop where the owner takes the calls and does the work has no office to move the decision to, and adding the step is just adding paperwork to the same brain. There the useful version is narrower: record the trigger date and the exclusion list somewhere retrievable from the truck, so steps 2 and 3 can be answered from the driveway rather than from memory.

Watch one number

Reclassification rate: callbacks whose coverage classification changed after the customer was first told, divided by total callbacks, on a rolling twelve months with at least 20 callbacks in the window. Anything over about 10 percent means the sort is not being run before dispatch, and the diagnosis is almost always one specific missing input rather than a general sloppiness.

Check the three inputs in this order when the number is high, because each one alone explains most cases: the trigger date is not recorded on job records, so step 2 cannot run; the original scope is written too loosely to answer step 1; or the exclusion list contains clauses that are not testable, so step 3 gets attempted on site and fails there.

And code the outcome honestly on every event, including the ones resolved as goodwill, because the same records feed the incidence data your term length is derived from. A goodwill visit coded as a warranty failure inflates your own failure curve, and a genuine warranty failure written off quietly as goodwill deflates it. Either way you end up setting next year's term against a number your own office invented.

References

• See related: Deciding Whether a Callback Is Warranty or Billable, When Goodwill Beats the Letter of the Warranty, What a Warranty Exclusion Has to Say to Be Fair, How to Set a Warranty Period You Can Live With