How to Document a No Fault Found Visit Defensibly

Why this matters

A no-fault visit is the highest-risk record your shop writes. If the system fails a month later, that record is the only evidence of what you saw, and "operates normally, no fault found" is not evidence of anything. It does not show what you checked, what target you compared against, how long you ran it, or what you told the customer. In a warranty dispute, an insurance claim, or a serious incident investigation, a bare verdict with no supporting observation reads exactly like a tech who did not look.

The same record has a second job that pays more often: it is the baseline the next visit compares against. A no-fault visit documented properly makes visit two a comparison. Documented badly, it makes visit two a fresh start, and the shop eats the difference every time.

Step 1: Lead the record with the safety action you took

If any real hazard was in play, the first line of the record is what you did about it, before any diagnosis. Gas or combustion products, water plus electricity, stored energy, height, or a customer reporting a burning smell all qualify.

Write the action, not the intention. "Evacuated occupants and shut off supply at the meter before entering" is a record. "Followed safety procedures" is not. For electrical work record that you verified dead with live-dead-live, meaning the meter was proved on a known live source, the circuit proved dead, and the meter proved live again. For anything capable of automatic restart, record that you locked it out rather than switching it off.

This ordering is not just liability hygiene. When an investigator or a manager reads the file later, the safety action is the fact they need first, and burying it under three paragraphs of readings makes the record look like an afterthought even when it was not.

Step 2: Record the complaint verbatim, then the observable

Two lines, not one. The customer's own words go in quotation marks exactly as they said them, including the vague parts. Then your translation of that complaint into something observable.

The customer's words matter because a complaint changes shape over time. On visit three, the customer may describe a completely different symptom and genuinely believe it is the same one. The verbatim quote from visit one is the only anchor.

Your observable matters because it defines what you actually tested. "Not working right" cannot be tested and cannot be closed. "Intermittent acrid smell, reported 3 to 4 times over two weeks, always during operation, strongest in the utility room" can be tested, and a reader can judge whether your test addressed it.

Step 3: Write the target and its source before you write any reading

State the number the system is supposed to produce and where that number came from. Nameplate, installation documentation, commissioning record, published operating range, or your own reconstruction from design intent, clearly labeled as a reconstruction.

Order matters here for the same reason it matters in the field. A record that lists readings first and a target second reads as a tech who measured and then went looking for a target the readings would satisfy. It is also how it usually happens, and any reviewer who works in the trades knows that.

If no target could be established, say that, plainly, and say what you compared against instead. Comparison to an identical peer under the same load is a legitimate and often stronger basis than a nameplate, but only if you say that is what you did.

Step 4: Record at least two reading sets, each stamped with time in run

One reading set is a snapshot, and a snapshot cannot detect the entire family of faults that only appear after heat soak, which is the state where component temperatures have stopped climbing. Record a cold set and a soaked set, each with the elapsed runtime written next to it, in the same unit.

Write the elapsed time even when it feels obvious. A record that says "amp draw 5.4" is not comparable to anything. A record that says "amp draw 5.4 at 72 minutes continuous run" is.

Use identical measurement points across both sets, and say where they were. A comparison between two different measurement points is not a comparison, and it is invisible to a later reader unless you name the points.

Step 5: Record what you ruled out and what ruled it out

This is the step that converts a no-fault record from a claim into an argument, and it is the step almost nobody does.

For each theory you tested and eliminated, write one line: the theory, and the specific observation that killed it. Not "checked wiring, OK." Instead: "Intermittent termination ruled out: three occurrences spaced within a few percent of the same interval, which is control-driven regularity, not thermal-mechanical."

Three of these lines tell a reader more about the quality of your visit than any list of readings, and they are what make the record survive scrutiny. They also do the practical job: the next tech does not repeat your dead ends.

Step 6: End in one of three named verdicts, never a bare "no fault found"

"No fault found" is a statement about you, not about the system. Replace it with one of three findings.

The third is critically different from the first two and gets conflated with them constantly. "I could not reproduce it" is not "it is fine." Recording an unreproduced intermittent as conforming is the single most damaging documentation error in this category, because it converts an open fault into a closed one in the file, and the next tech inherits a false clearance.

Step 7: Write the escalation trigger in the customer's language

The record needs a line that a non-technical person could act on. Specific, observable, and framed as a threshold rather than a judgment.

Bad: "Call if it gets worse." Worse is a feeling and they will call for anything or nothing.

Good: "Call back if the pause between running periods drops below about five minutes, if the smell returns and lasts more than one full cycle, or if it does not restart on its own." Each of those is something the customer can observe without any training, and each maps to a real change in system state.

Put the same wording in the customer's copy and in your record, so there is no daylight between what you told them and what you wrote.

Step 8: Capture what cannot be re-created

Photograph anything you cleaned, moved, or corrected, before you touched it. Corrosion, staining, discoloration at a termination, the installed consumable you are about to replace, the position a valve was found in. Once altered, that evidence is gone permanently, and a photograph taken in ten seconds is the entire defense of your recommendation months later.

For a noise complaint, record audio or a short video with sound. A recording is the only way a later reader can evaluate a sound, and it also gives you something to compare against if the customer calls back claiming the sound changed.

The record as written

Complaint (verbatim): "There's a burning smell every so often when it's running. It's happened maybe three or four times in two weeks."

Safety action: Customer reported burning smell. Confirmed no active smoke or heat on arrival, no occupant symptoms. Isolated and locked out at the disconnect, verified dead at the control terminals with live-dead-live before opening any enclosure.

Observable: Intermittent acrid smell during operation only, 3 to 4 occurrences over 14 days, no correlation reported to time of day.

Target and source: Nameplate full-load current 5.0 amps. Manufacturer documentation states a permissible imbalance limit between legs. No commissioning record on file for this system.

Readings, set 1 (4 minutes, cold start): leg A 4.6 amps, leg B 4.7 amps, measured at the load terminals. No thermal rise at any of the four terminations relative to conductor.

Readings, set 2 (72 minutes continuous, heat soak): leg A 4.8 amps, leg B 5.1 amps, same terminals. Roughly a 6 percent imbalance around the average, inside the manufacturer's stated limit. Termination 3, on leg B, reads measurably warmer than the three peer terminations under the same load.

Cycle behavior: 3 complete cycles observed over 66 minutes, 22 minutes each, running 12 and off 10, a 55 percent duty cycle against an expected band of 50 to 70 percent at the conditions on the day.

Ruled out: Off-spec customer-supplied consumable ruled out, element measured on spec for pressure drop at the operating point. Protective limit trip ruled out, no lockout or fault record in control history and system never stopped during the observation.

Verdict: Conforms with an observed trend. All readings inside tolerance. Trend finding is the peer asymmetry at termination 3 combined with the higher-reading leg it sits on, consistent with an early high-resistance joint and consistent with the reported intermittent smell.

Action taken: Termination 3 remade to specified torque. Post-repair readings at 40 minutes continuous: leg A 4.8, leg B 4.9 amps, all four terminations at conductor temperature.

Escalation trigger given to customer, verbatim in their copy: "Call back if the smell returns at all, even once, or if anything near the panel is warm to the touch."

What changes the answer

A hazard-adjacent complaint raises the bar. Anything involving combustion, a burning smell, water plus electricity, or a report of a shock is documented as if it will be read by an investigator, because it might be. The unreproduced-intermittent verdict is mandatory rather than optional in those cases, and closing one as conforming is indefensible.

A commercial customer with a service agreement usually specifies the format. Contracts frequently require readings, elapsed times, and named verdicts, and a bare no-fault line can put an invoice into dispute on its own terms.

A system under a manufacturer warranty needs the target sourced to the manufacturer. A warranty claim built on a peer comparison rather than a published specification is much easier to deny.

Third visit on the same complaint changes the deliverable entirely. At that point the record's job is not to close the visit, it is to build the case that the fault is real and needs a different approach. Document the aggregate across all three visits, not just this one.

How to verify you got this right

Hand your record to someone who was not there and ask them three questions. Can you tell what was tested, and against what? Can you tell what was eliminated, and why? Can you tell what the customer was told to watch for?

If a colleague cannot answer all three from the record alone, it is not defensible, and it will not help the next tech either. The failure mode is concrete and common: the system fails six weeks later, the customer says "your guy was here and said it was fine," and the file contains one line with no target, no runtime, and no trigger. There is no way to show you checked, and no way to show the condition changed after you left. You will pay for that failure whether or not you caused it.

References

• OSHA recordkeeping and hazard-communication expectations for documented safety actions on service work
• NFPA 70E live-dead-live verification practice, recorded as taken
• Manufacturer documentation for nameplate ratings, permissible imbalance, and warranty claim substantiation requirements
• See related: How to Verify a System Is Operating as Designed; Documenting an Unsolved Fault for the Next Tech; The No Fault Found Visit SOP