How to Diagnose After the Customer Already Replaced a Part

Why this matters

The call comes in as "I already replaced the part and it still does it." You are walking into a job where the most useful piece of evidence has been thrown away, the system has been opened by someone who did not document what they did, and a brand-new component now sits in the circuit that you have no reason to trust and every reason to assume is fine. Handled well, these calls close cleanly and the customer becomes loyal because you did not repeat their mistake. Handled by habit, you spend two hours proving a fault that a wrong-spec part was guaranteed to produce.

First: what a prior attempt can leave behind

Do this before anything is energized, pressurized, or lit, every time, no exceptions for a job that looks simple.

• Electrical. Assume a conductor is landed wrong, a ground is off, or a cover screw is through insulation. De-energize at the disconnect, lock and tag it, then prove dead with a meter you tested on a known live source immediately before and immediately after the check, which is the live-dead-live sequence in NFPA 70E-2021, 120.5. The requirement to de-energize and lock or tag before working on a circuit that could be energized sits at 29 CFR 1910.333(b)(2) for general industry and at 29 CFR 1926.417 when the same work is done under construction rules. Discharge capacitors and other stored electrical energy before your hands go in.
• Stored mechanical energy. A spring, an accumulator, a lifted or counterweighted assembly, or a pressurized vessel that somebody partially released and walked away from is the classic prior-attempt injury. Isolate, block, and relieve stored energy before disassembly under 29 CFR 1910.147, and confirm relief by a gauge or a physical check rather than by the fact that a valve handle is turned.
• Gas. If you smell gas at the door, everyone leaves the building immediately, nobody touches a switch, nobody turns on a light, nobody uses a phone inside, and you call from outside. If there is no odor but the gas train was opened, leak-test every joint that was disturbed with a listed leak-detection solution or a combustible-gas detector, and do that before any attempt to light or fire the appliance.
• Water plus electricity. Standing water anywhere near an energized enclosure means the circuit gets killed and proven dead before the water gets touched.

Say plainly to the customer that you are checking their work for safety before you diagnose. Almost nobody objects, and it sets the tone that this is an inspection, not a rescue.

How the rest of this is ordered

Not chronologically. These steps are ordered by what you permanently lose if you skip them, because on a prior-repair call the losses are unrecoverable and they compound. Evidence that walks out to the curb in a trash bag does not come back, and a reading taken after you have adjusted something is a reading of your own work.

Skip this and every later reading is meaningless: identify what the new part actually is

Before you test anything, establish three facts about the installed part: what it is, what it should have been, and whether it is configured.

Read the rating off the part itself, not off the box, not off the receipt, and not off what the customer says the counter guy told them. Compare it to the equipment's own requirement from the data plate or the documentation. Then check whether the part has settings: a switch bank, a jumper, an adjustment screw, a stop, an orifice, a spring tension. A correct part shipped in a default configuration that does not match this application will produce a convincing, repeatable, completely fake fault.

What you lose by skipping it: everything downstream. If the part is wrong or misconfigured, every reading you take is a reading of the wrong system, and you will burn the whole visit building a theory about a system that does not exist.

Skip this and the best evidence is gone forever: recover the old part

Ask for it before you ask anything else, and ask in a way that gets a real answer: "Do you still have the old one, or the box it came out of?" Many customers keep it in the garage, the truck bed, or the trash can that has not gone out yet.

A failed part read properly tells you whether it died of its own accord or was killed by something still present. Heat concentrated at one terminal, one lobe, or one side of a seal is a directional clue. Uniform wear is age. Contamination is an upstream problem. A part that failed young with damage pointing at one location is the strongest single argument that the cause is still in the system.

What you lose by skipping it: the difference between "it wore out" and "something is still killing these," which is the difference between a finished repair and the same call in four months.

Skip this and you cannot read the symptom change: anchor the swap to an event

Get the timeline in the customer's own reference frame. Dates are unreliable and people guess. Events are reliable.

Ask: what was it doing before you changed it, what was it doing right after, and what is it doing now. Anchor each to something memorable: before the holiday, after the storm, the week the other trade was here. Then ask the question that separates a partial repair from no repair at all: did the symptom get smaller, or did it change shape? A fault that is now less frequent but harder to recover from has not improved.

What you lose by skipping it: the ability to tell whether the customer's swap did something. If the symptom changed at the swap, the new part is part of the story. If it did not change at all, the replaced component was almost certainly never involved, and you can spend your time elsewhere.

Skip this and your baseline is not the original baseline: find what else was touched

Almost nobody replaces exactly one thing. They also cleaned something, moved a setting to see if it helped, put the wires back from memory, left a filter out, propped a damper, opened a bleed, or turned a bypass and forgot it.

Ask directly and without judgment: "What else did you end up adjusting while you were in there?" Then verify independently. Walk every access panel, look for fresh tool marks on fasteners, check every adjustable setting against its normal position, and confirm every removable element is present and oriented correctly.

What you lose by skipping it: you diagnose a system that no longer matches its own documentation, and you can chase a setting for an hour.

Skip this and you will eliminate the real cause on the first pass: test the new part as an unknown

This is the discipline that separates this call from an ordinary one. A component only counts as eliminated when you verified it, in circuit, under the condition where the fault appears. New certifies nothing. Parts arrive dead, get damaged during a first-time install, and get sold as equivalents that are not.

Test it as you would test a suspect part of unknown age. If it passes under the fault condition, it becomes a constant and you move upstream. If it does not, you have found the cause and it was not the one anyone expected.

What you lose by skipping it: the candidate list. Treating the new part as automatically good removes the true cause from consideration at step one, and no amount of later work brings it back, because you will never go looking there again.

Skip this and the visit turns into an argument: scope and price out loud, before you dig

Say the shape of the job before you are elbow deep. Something close to: "This takes longer than a normal diagnostic because I have to establish what changed before I can test anything. My block on these is two hours. If I have not found it by then, I stop, tell you what I have ruled out, and we decide together."

That sentence does three jobs at once. It sets expectations, it makes the extra time a stated service rather than a surprise, and it gives you a clean stopping point instead of a slow slide into unpaid work.

What you lose by skipping it: the boundary. Prior-repair calls run long by their nature, and a shop that does not price for that either eats the hours or has an uncomfortable conversation at the end.

Worked example: the part was right and it was going to die anyway

A customer replaced a control component after watching a video that matched their symptom. The complaint persisted, so they called.

The safety check found the enclosure cover reinstalled with one screw missing and a conductor's insulation pinched. That got corrected before anything else, with the circuit locked, tagged, and proven dead first.

Identification came next and cost 0.3 hour. The installed part was the correct rating and the correct configuration. Useful, and it also meant the easy answer was gone.

The old part was still in the garage. Reading it cost 0.2 hour and it was the whole job: heat damage concentrated at one terminal, not spread across the body. Uniform overheating means the part was working too hard everywhere. Concentrated damage at one point means the energy went in at that point, which puts the cause in what feeds it.

The timeline interview took 0.2 hour and gave a clean anchor. The customer's answer was that the fault was identical before and after the swap, same trigger, same recovery, no change in frequency. That is the signature of replacing a component that was a casualty rather than a cause.

The what-else-was-touched audit took 0.3 hour and turned up one adjustment moved from its factory position, which was restored and noted.

Testing the new part in circuit under the fault condition took 0.4 hour. It performed correctly, so it became a constant.

The last 0.4 hour went to the supply feeding it, measured at the equipment's own terminals under the load that triggers the complaint rather than at the source with the system idle. The data plate states a tolerance of 10% below nominal. The measured value under that load sat 12% below nominal, which is outside what the equipment is built to accept and exactly consistent with concentrated heating at the point where energy enters.

Total: 0.3 plus 0.2 plus 0.2 plus 0.3 plus 0.4 plus 0.4, which is 1.8 hours against a standard diagnostic block of 1.0 hour. The shop's median on flagged prior-repair calls is 1.7 hours, so they now quote a 2.0-hour block on these and re-quote if the cause is still open when it runs out. That is a 70% median adder over an ordinary diagnostic, which is a real cost that has to be priced rather than absorbed.

The customer's part stayed in service. The repair was upstream. Had the tech treated the new component as automatically good and skipped the in-circuit test, the finding would have been the same here, but the reverse case is the one that hurts: a defective new part, assumed good, moves the entire diagnosis onto healthy equipment and produces a recommendation to replace something expensive that was never at fault.

References

• 29 CFR 1910.333(b)(2), and 29 CFR 1926.417 for the same work under construction rules, on de-energizing and locking or tagging before work on circuits that could be energized
• NFPA 70E-2021, 120.5 for the live-dead-live instrument proving sequence
• 29 CFR 1910.147 on isolating and relieving stored mechanical energy before disassembly
• See related: Confirming a Part Is Actually Bad
• See related: Reading a Failed Part for the Real Cause