The Original Diagnosis Was Wrong, Not the Repair: Decision Tree
Why this matters
When your own repair comes back with the same symptom, techs reflexively re-do the repair: replace the part again, re-seal the joint, re-make the connection. Half the time that is the wrong move, because the repair was never the problem, the diagnosis was. You replaced a good part while the real cause kept working on its replacement. This tree separates a bad diagnosis from a bad repair, because they demand opposite fixes: one means look elsewhere, the other means do it again right.
Safety first
If the returned fault presents as a hazard - gas odor, burning smell, arcing, water at a live component, no heat in a hard freeze - control the hazard before diagnosing: clear people, shut off gas or water, de-energize and verify dead. Only then work the tree below.
Start here: the three links in your own chain
A callback on your own work failed at one of three links. Name which before you touch a tool.
- Diagnosis - you identified the wrong cause and fixed a part that was never broken.
- Execution - you had the right cause but the repair itself was short: incomplete, wrong part, poor workmanship.
- Neither - your work was correct and a genuinely separate fault returned the same symptom.
The tell that sends you down each path is what the replaced part looks like now.
If the part you replaced is already failing again
Pull or test the exact component you serviced. If it is out of spec, leaking, or dead again after days, the part is not the story, the reason it keeps dying is.
This points at a bad diagnosis. A healthy new part that fails fast almost always has an upstream cause you did not find.
- A replacement capacitor cooked in days: suspect a failing motor dragging it down, not a bad capacitor.
- A new water heater element that burns out: suspect scale, a stuck thermostat, or a dry-fire, not the element.
- A fresh igniter that fails again: suspect low gas pressure or a control problem, not the igniter.
- A pump seal that weeps again: suspect misalignment or high pressure, not the seal.
Stop replacing the symptom. Diagnose the load, the pressure, the supply, or the control that is killing it, and fix that.
If the part you replaced is still good
Test it and it holds to spec. Then your diagnosis of that part was fine, and either your execution or a separate fault is the issue.
Suspect execution first. Re-inspect your own work for the incomplete step:
- A valve left half-seated, a line not fully purged of air, a connection not torqued, a setting not restored.
- A fitting that seals on the bench but was cross-threaded or under-tightened in place.
- A wire made up loose, a clamp not fully closed, a panel gasket not reseated.
If you find the gap, that is a workmanship callback: right diagnosis, short execution. Correct it and own it.
If your diagnosis and your repair both check out
The replaced part is good, your workmanship is sound, and it is still down. Now treat it as a separate fault that happens to share the symptom.
- Diagnose the whole system fresh, as a first visit, and prove the new fault is independent of your work.
- This is the concurrent-fault case: two problems, one visible symptom, and you only fixed one because the other was hidden behind it.
See the related article on separating a failed repair from an unrelated new problem for how to prove independence.
Why this order matters
Techs default to re-executing (link 2) because it is the least ego-bruising: "I just need to do it again." But a returned fault on a replaced part usually means link 1, the diagnosis, was wrong, and re-executing a wrong diagnosis burns a third visit. Checking the replaced part first tells you in one measurement whether you are chasing a cause you never found or a step you never finished.
Ordered recap
- Make any hazard safe first.
- Name the three links: diagnosis, execution, or separate fault.
- Test the part you replaced. Already failing means bad diagnosis - hunt the upstream cause.
- Still good means check your own execution for the incomplete step.
- Both sound means diagnose fresh for an independent concurrent fault.
- Never re-do a repair before you know the diagnosis was right.
References
- Trade-standard practice for root-cause diagnosis and callback analysis
- See related: Reopening a Diagnosis You Thought You Closed; Separating a Failed Repair From an Unrelated New Problem
- See related: Is This the Same Fault or a New One on Your Callback (decision tree)