How to Show a Customer Their System Is Working Correctly

Why this matters

Telling a customer their system is fine almost never ends the complaint. They watched a stranger walk around for twenty minutes, they heard a verdict that contradicts what they have been living with, and the natural conclusion is that you missed it. The call comes back, often to a competitor, and the second shop sells a part to make the customer feel heard. You paid for the truck roll and handed away the relationship.

Showing is a different act than telling. A demonstration the customer participates in, where they predict the outcome before it happens and watch it happen, converts a disputed verdict into a shared observation. That is the difference between a visit that closes and one that repeats.

Step 0: Set the hazard state before you demonstrate anything

Demonstrations tempt you to run equipment with covers off and a civilian standing next to you. Decide the exposure first. If a panel has to come off to show something, de-energize and verify dead using live-dead-live (prove the meter on a known live source, prove the circuit dead, prove the meter still live) before anyone approaches, and put the customer at a fixed distance you name out loud before you re-energize. If gas, combustion products, stored pressure, or height are in play, the customer does not get a front-row seat at all, and the demonstration happens at the readout or the register rather than at the equipment.

You can show someone a number safely. You cannot un-burn their hand.

Step 1: Pin the complaint to something observable before you touch the system

"It's not working right" cannot be demonstrated either way. Push the complaint down to a specific observable: a sound they can point at, a time of day, a room, a duration, a visible thing. Ask what they expected to happen instead, because half the time the expectation is the actual finding.

Then ask the question that saves the most time: "What changed around when you first noticed it?" You are looking for the trigger, not the fault. A recent service visit, a season starting, a new person in the house, a neighbour's failure. If a trigger exists, you now know you are probably explaining a machine rather than repairing one, and you can plan the demonstration accordingly.

Skipping this step is what produces a demonstration of the wrong thing. You prove the compressor is fine while the customer's actual complaint was a noise in a bedroom two floors away.

Step 2: Agree on the test before you run it

This is the step that does the work, and it is the one almost everyone skips. Before you start anything, say what you are about to do, what you expect to see, and what result would mean they are right.

"I'm going to run it for an hour and time every cycle. If it's working the way it should, it'll run somewhere around ten to fifteen minutes at a time on a day like today. If it runs three minutes and quits, or runs the whole hour without stopping, you're right and something's wrong."

Two things happen. You have made the test falsifiable in front of them, which is the entire basis of trust in a verdict. And you have pre-committed yourself, which stops you from quietly moving the goalposts if the result is inconvenient. A customer who agreed to the criterion before the test has a much harder time rejecting the result after it, and more importantly, they should not have to take your word for the criterion afterward.

Step 3: Reproduce the behavior on demand

A behavior you can make happen and stop at will is a behavior you understand. Nothing convinces a skeptical customer faster than watching you produce the exact sound or event they called about, on cue, three times running.

If the behavior is tied to a cycle transition, start and stop the cycle. If it is tied to load, change the load. If it is tied to temperature, wait for the heat soak and call the moment before it arrives. Say what is about to happen a beat before it happens: "It's going to tick about four times in the next thirty seconds as that section warms up, then go quiet." When it does, you have demonstrated that you know the mechanism, not just the outcome.

If you cannot reproduce it, say so plainly and do not substitute a different demonstration. An intermittent you could not force is a genuinely different outcome from a normal behavior you explained, and blurring them is how a real fault gets closed out.

Step 4: Show one number against its target

Resist the instinct to show your whole reading set. A customer cannot evaluate six numbers and will read the display of expertise as an attempt to overwhelm them. Pick the single measurement that most directly answers their complaint, show them the target it is being compared against, and show them the reading next to it.

Point at the nameplate or the documentation while you do it. The credibility comes from the target having an independent source, not from your meter. A reading with no stated target is just a number you produced.

Step 5: Show the contrast case

One reading proves nothing to a layperson because they have no scale for it. Two readings, where they can see the difference, teach in seconds.

Induce the contrast only on a path whose worst realistic outcome is a changed reading. Never induce a fault on a combustion path, a pressurized path, a refrigerant-bearing path, or anything that serves a relief or protective function: restricting a suction line, deadheading a pump, or obstructing combustion air or a relief path can cause damage in seconds and a hazard almost as fast, and none of that is recoverable by reopening what you closed. If the only way to create a contrast on this system crosses one of those lines, skip this step and use the recorded baseline from the last visit instead.

Inside that boundary, the usual safe contrast is an airflow or a low-consequence flow path: block a return, close one supply branch, cover a filter face. Let them hear the sound change and watch the reading shift, then restore it and let them watch it return. Now they have seen what wrong looks like on the same instrument in the same house, and their own system's reading has a scale.

The contrast case is also your defense against the accusation you were going to get anyway: that of course the meter reads fine, meters always read fine.

Step 6: Make them predict

The demonstration is complete when the customer can forecast the behavior without you. Before you leave, ask what they expect the system to do tonight, and on the first genuinely hot or cold day of the season. If they can answer with roughly the right shape, the explanation landed. If they cannot, you explained the machine and not the behavior, and you should go back one step.

The worked example, carried through

Complaint: "It runs way too much, it never shuts off, I think it's on its last legs."

Observable, after Step 1: continuous-feeling operation, noticed over the past two weeks, trigger is a season starting plus a household member now working from home during the day for the first time.

Pre-committed criterion, agreed in Step 2: on a day like today, expect roughly 10 to 15 minutes of run time per cycle with a real off period between cycles. Anything under about 3 minutes per run, or no off period at all in an hour, means the customer is right.

Result: across 60 minutes, 3 complete cycles at 20 minutes each, running 12 minutes and off 8 minutes. That is a 60 percent duty cycle, and every run sits inside the 10 to 15 minute band that was agreed before the test. The system satisfied demand and shut off, three times, in front of the person who said it never shuts off.

Contrast case: close off two supply paths, restart, and let them watch. Run time on the next cycle stretches while the same amount of output has fewer places to go, and the flow noise they can now hear is louder than anything the system was making before. Open them back up and the following cycle returns to the band.

The prediction: on the hottest day of the year, expect that 60 percent duty cycle to climb toward continuous, and that is the system working correctly at its design point, not failing. Naming that in advance is what stops the July call.

The finding that closes the visit is not "no fault." It is: system meets the cycle criterion agreed before the test, 3 of 3 runs inside the 10 to 15 minute band with a real off period each time, at present conditions; complaint attributable to a household occupancy change that put a person in the house during the hours the system has always run. Report against the criterion you pre-committed and nothing else. The 60 percent duty cycle is worth recording as a baseline for next season, but it is an observation, not the test you agreed, and quoting it as a pass against some band you never named in Step 2 is the same goalpost move this whole procedure exists to prevent.

When you can never get there: the same method by voice

Sometimes there is no visit. The customer is remote, the property is inaccessible, the call came in after hours, or a truck roll cannot be justified for the complaint. Their eyes and ears are your only instrument, and the method still holds if you adapt three things.

Give them a measurement they can actually take. Not a meter reading. A stopwatch on the phone, a count, a duration, a photograph, a short video with the sound recorded. "Start a timer when it kicks on, stop it when it goes quiet, do that three times, text me the three numbers" produces real data from an untrained person.

Pre-commit the criterion the same way, out loud, before they collect anything. Tell them what number would mean it is fine and what number would mean you are coming out. Otherwise you will be accused, fairly, of interpreting their data to suit your schedule.

Ask for the observation, never the diagnosis. A remote customer will offer you conclusions: "the motor's straining," "it's overheating." Those are guesses wearing technical vocabulary, and acting on them is how a remote call goes wrong. Convert every one back to an observable. What do you hear, when does it start, does it change, how long does it last, is anything warm to the touch that was not before.

The remote failure mode to guard against: you accept the customer's framing, talk them through a fix for the fault they named, it does not help, and now you have burned your credibility before you ever see the equipment. Collect observations, state the criterion, and be willing to say the honest thing, which is that some faults cannot be closed out at a distance and you need eyes on it.

How to verify you got this right

The visit is done well when three things are true in the record. The complaint is written as an observable, not an adjective. The criterion is written down and was stated before the test, with the actual result next to it. And there is an escalation trigger in plain language that the customer can act on: what specifically to watch for, and what would justify calling back.

The failure mode shows up about a month later. The customer calls back with the identical complaint, the record says "demonstrated normal operation to homeowner, no fault found," and there is no criterion, no number, and no trigger. You cannot tell whether the behavior changed or the customer's tolerance did, so the second visit starts from nothing. Write the criterion down and the second call becomes a comparison instead of a repeat.

References

• NFPA 70E live-dead-live verification practice before exposing energized parts, especially with a bystander present
• Manufacturer operating documentation for expected cycle behavior and duty at stated load conditions
• Standard trade practice on keeping a non-employee bystander clear during equipment operation and testing. Note that OSHA's duties run to your employees, not to a customer standing in their own home; your obligation toward that customer is the ordinary duty of care, which is why the boundary in Step 0 is yours to enforce rather than something a regulation enforces for you
• See related: How to Verify a System Is Operating as Designed; Teaching the Difference Between Normal and a Problem; How to Document a No Fault Found Visit Defensibly