The warranty year is the most neglected stage of the job — the crew is gone, the money is (mostly) collected, and the PM is three projects deep into something else. Then the GC calls about a failed unit, the call bounces around for a week, and a relationship you spent two years building erodes in the gap. Here’s the uncomfortable math: to the GC, your warranty response is your reputation, because it’s the only part of your work they experience after the leverage is gone.

The short version

  • The warranty clock starts at substantial completion — know the date, know the expiry, and track both per project, per phase if the contract splits them.
  • Respond to every callback fast, even when it isn’t yours. Speed of first response is the thing GCs remember; the entitlement argument can come second.
  • Sort every callback into workmanship (you), manufacturer (the equipment), or neither (damage, maintenance, misuse) — and register equipment warranties at closeout so the manufacturer path is one lookup, not a filing-cabinet dig.
  • Track callback cost — labor hours and materials per visit. Warranty work is unbilled; untracked, it’s invisible margin leakage and lost estimating intelligence.
  • Recurring failures are data: the same root cause showing up three times is a message to estimating and purchasing, if anyone’s writing it down.

Know your clock — and don’t let it start late or run long

The warranty period (typically one year on workmanship) runs from substantial completion, not from final payment or the end of punch. Two edges of the same rule cut for you: a substantial completion date that drifted late during a slow closeout silently extended your free-repair window; and a callback that arrives in month thirteen is a service call, not a warranty obligation — but only if you know the dates cold. Track, per project: warranty start, warranty end, and any separate periods (phased turnover, extended equipment warranties you passed through). When the owner occupies early or in phases, get the warranty-start consequences in writing at the time — occupancy fights make terrible anniversary arguments.

Triage: fast first, entitled second

The instinct on a callback is to establish whose problem it is before rolling a truck. Reverse it. Acknowledge same-day, get eyes on it within the response window the contract sets (or a couple of business days if it doesn’t), then sort entitlement. Reasons: a fast look usually costs one visit; a slow argument costs the relationship; and half the time you can’t classify the failure without looking anyway. The warranty-stage GC phone call is also the cheapest sales call you’ll ever make — the super who watched you show up in two days on a job with no money left in it is the same super scoring your bid next quarter.

Fast response doesn’t mean free work. It means the triage visit is prompt and documented, and the classification comes with evidence.

The three buckets — and the paperwork that sorts them

  1. Workmanship — your install, your fix, your cost. Do it promptly and completely; this is the bucket the warranty exists for.
  2. Manufacturer / equipment — the unit failed, not the installation. Your job is claims management: model, serial, install date, failure mode, and the manufacturer’s warranty terms. This bucket is only efficient if you registered the equipment warranties at closeout — every unit with its serial, warranty start (which often runs from startup, not shipment — worth confirming at commissioning), duration, and the claim contact. Do that once at closeout, and a compressor failure is a lookup and an RMA number; skip it, and every equipment callback begins with an archaeology dig through submittals.
  3. Not warranty — owner damage, missed maintenance, misuse, another trade’s work. Decline with the evidence: photos, the O&M requirement that wasn’t followed, the maintenance log that doesn’t exist. Offering to fix it as billable service work converts an awkward “no” into revenue, and plenty of owners take the offer.

The classification only sticks if each callback carries a record: reported by whom, date, issue, visit log, root cause, and disposition. That record is also what protects you when a “warranty” pattern is really a maintenance pattern.

Track the cost, mine the pattern

Warranty labor is real labor — unbilled, usually invisible, absorbed by whoever was free that week. Log hours and materials against each callback and two things surface. First, the true warranty cost per job and per GC, which belongs in your overhead assumptions — if warranty runs half a point of contract value, your margin math should know that. Second, and more valuable: patterns. The same valve failing on three jobs, the same finish problem behind one crew, the same manufacturer eating a week of claims time per incident — each is an estimating and purchasing decision waiting to be made, sitting unread in your callback log. A quarterly half-hour on the pattern report — repeat root causes by manufacturer, by crew, by detail — is how the warranty stage stops being a cost center and starts feeding the front of the funnel: fewer callbacks on the next job, tighter numbers under the next bid, and a GC who remembers what happened the last time something broke.