The Symptom List That Points Two Ways: Decision Tree
Why this matters
You have a set of symptoms, and they fit two completely different causes equally well. Both stories explain everything the customer told you. Pick one and start replacing parts and you have a coin-flip dressed up as a diagnosis - half the time you fix it, half the time you are back next week having spent the customer's money on the wrong thing. The skill is not to guess better. It is to find the one test that the two causes disagree about, and let that test split them. This tree is how you turn an ambiguous symptom list into a decision.
Start here: does either branch involve a hazard
When two causes both fit, and one of them is dangerous, you do not get to treat them as equally likely for safety purposes. If one possible cause is a safety issue - a gas, electrical, pressure, or structural hazard - handle the equipment as if the dangerous cause were the real one until you have ruled it out. De-energize, depressurize, shut off, or keep clear as that hazard requires before you run the tests that distinguish the two. You can confirm the harmless cause later; you cannot un-injure someone you exposed by assuming the safe branch. The dangerous possibility sets the handling, even when it is only one of two.
For everything below, assume both candidate causes are non-hazardous.
First move: write both stories out in full
Ambiguity hides when you keep the candidates vague in your head. Force them into the open.
- State cause A and cause B as complete explanations. Each one should account for every symptom on the list, or you are not really stuck between them - one already fits worse.
- List what each cause predicts. Not just the symptoms you see, but everything that cause would also produce. This is where they start to differ.
- If a third cause is lurking, name it too. "Two ways" is sometimes three. Get them all on the table.
A symptom that both causes explain tells you nothing. You are hunting for the symptom or test where the two predictions split.
Second move: find the discriminating test
This is the entire technique. Stop testing things both causes agree on; find the one thing they disagree on.
| If both causes predict... | The test is useless because... | Look instead for... |
|---|---|---|
| The same reading | Either way it reads the same | A point where their readings diverge |
| The same symptom under load | Both produce it | A condition where only one acts up |
| Failure in the same area | Both point there | An upstream or adjacent check that separates them |
Ask: "What would be true if A but not B?" That difference is your discriminating test. One clean measurement on a point where the two causes predict different results collapses the ambiguity. Spend your effort finding that point, not re-confirming the shared symptoms.
Third move: run the cheapest discriminator first
Often there are several tests that could split the two causes. Order them by cost and risk, not by what is most interesting.
- Cheapest and least invasive first. A reading you can take without disassembly beats one that requires tearing into the unit. If a simple check separates A from B, take it before anything bigger.
- Reversible before irreversible. A test that does not commit you - no parts replaced, nothing cut - comes before one that does.
- One variable at a time. Change one thing, observe, and you keep the result interpretable. Change two and you reintroduce the ambiguity you are trying to kill.
The first discriminating test that comes back clear ends the analysis. You do not need to run them all.
Fourth move: if no single test splits them, consider both are present
Sometimes the symptom list points two ways because both causes are actually there. This is the case people miss while insisting it must be one or the other.
- Treat a truly even split as a possible double fault. If every discriminator comes back ambiguous, the most likely reason is that A and B are both contributing (see related: the fix worked for the wrong reason).
- Address the one you have confirmed, then retest. Fix the cause you can prove, and see what symptoms remain. The leftover symptoms reveal the second fault cleanly.
- Do not replace both on suspicion. Confirm, fix, retest, confirm again. Shotgunning two parts because you could not decide is the expensive habit this whole method exists to prevent.
Recap
- If either candidate cause is a hazard, handle the equipment as if the dangerous one were real until you rule it out.
- Write both causes out as full explanations and list what each one predicts.
- Find the discriminating test - the point where the two causes predict different results - and ignore tests they agree on.
- Run the cheapest, most reversible discriminator first; the first clear result ends it.
- If nothing splits them, suspect both are present: fix the confirmed one, retest, and let the remaining symptoms name the second fault.
The judgment to bank: when symptoms point two ways, you do not win by guessing the more likely cause. You win by finding the one question the two causes answer differently, and asking only that.
References
- Trade-standard practice for differential diagnosis and discriminating tests
- Manufacturer documentation for component test limits and expected readings
- OSHA guidance on defaulting to the hazardous assumption under uncertainty
- See related: The Fix Worked for the Wrong Reason (decision tree); The Tie-Breaker When Two Pros Disagree