It Worked but You Don't Know Why: Decision Tree

Why this matters

The most dangerous words in the trade are "well, it's working now." You touched something, the symptom vanished, and you do not actually know what fixed it. A fault that disappears without a confirmed cause has not been repaired; it has gone quiet. It will almost certainly come back, often right after the truck leaves, and now the customer trusts you less because the thing you "fixed" broke again. The temptation to declare victory and move on is strong, especially with a customer watching and a schedule waiting. This tree is about what to do when the system started working and you cannot explain why.

Start here: is the fix understood or accidental?

Be honest with yourself about which just happened. An understood fix has a known mechanism: you found the bad part, replaced it, and the symptom cleared because that part was the cause. An accidental fix is the symptom going away while you were poking around, with no confirmed link between what you did and why it recovered.

If you can name exactly what failed and why your action addressed it, this is a real repair. Verify under the triggering conditions and you are done.

If the symptom cleared but you cannot point to the cause, you have an accidental recovery. Do not stop. Work the rest of this tree.

Ask: did you disturb something without knowing it?

The most common accidental fix is reseating a bad connection without realizing it. Wiggling a harness, removing and reinstalling a part, or cleaning a contact can temporarily restore a failing joint that will fail again as soon as it is disturbed by heat, vibration, or time.

If the symptom cleared right after you moved, unplugged, or reseated something, suspect an intermittent connection at that exact point. That is now your prime suspect, not a mystery. Inspect the connection properly: corrosion, a loose terminal, a cracked solder joint, a stressed wire.

If you did not physically disturb anything and it still recovered, the cause is more likely a condition that changed (it cooled down, the load dropped, the timing passed). Move to the condition question.

Ask: did a condition change that you can recreate?

Many faults are conditional and recover on their own when the condition passes. The system "fixed itself" because it cooled off, the load lifted, or the moment passed, not because anything is repaired.

If the symptom appeared under heat, load, or a specific cycle and cleared when that condition ended, the fault is still there, hiding. Recreate the condition deliberately and the fault should return, giving you the diagnostic opportunity you need.

If you can make it fail again by recreating the condition, you have converted a mystery into a testable fault. Diagnose it properly now while it is present.

Ask: can you make it fail again at all?

The decisive test for an accidental fix is whether you can reproduce the failure on demand. A fault you can summon is a fault you can diagnose.

If you can reliably trigger the symptom again, do so, instrument it, and find the real cause. The recovery was temporary and you now know it.

If you cannot reproduce it no matter what you try, you are in genuine intermittent territory. Replacing a part you never saw fail is a guess; be careful here.

The trap of the confident handoff

The pressure to tell the customer "all fixed" when you do not know why is exactly when you should not. A confident handoff on an unexplained recovery is the setup for the worst kind of callback: the customer was told it was solved, so its return feels like incompetence rather than the nature of an intermittent fault.

If you cannot explain the fix, do not claim you solved it. Tell the customer honestly: the symptom is currently absent, you suspect a specific intermittent cause, and here is the plan if it returns. Honesty here protects the relationship; false confidence destroys it.

The decision

Declare a repair done only when you understand the mechanism and have verified it under the triggering conditions. When the recovery was accidental, chase the most likely explanation: a disturbed intermittent connection if you moved something, a passed condition if one changed, and reproduce the failure to confirm before committing to any part. If you genuinely cannot make it fail again, do not install a part on hope and call it fixed. Document the suspected cause, leave monitoring if you can, and frame the next occurrence as the diagnostic opportunity. An honest "could not confirm, here is what to watch for" beats a confident guess that returns next week.

What to document

Record exactly what you did right before the symptom cleared, because that is the strongest clue to an accidental fix. Note whether you suspect a disturbed connection or a passed condition, what you tried to reproduce the fault, and whether it returned. Be explicit that the cause was not confirmed if it was not. This record keeps the next tech from re-walking the same dead end and keeps you honest about what was actually repaired.

References

• Manufacturer service documentation, intermittent-fault and connection-integrity guidance.
• Electronics Technicians Association International (ETA), intermittent-fault diagnostic methodology.
• See related: Recreating the Conditions to Find the Fault; Intermittent Fault: Cannot Reproduce Decision Tree; The Obvious Cause Is Wrong Decision Tree.