Everything Tests Good But It Fails: Decision Tree
Why this matters
Every component passes. Voltages are right, pressures are in range, the part ohms out fine, the board reports no error. And the thing still does not work. This is the diagnostic puzzle that humbles experienced techs, and the temptation is to declare the meter a liar or the customer confused. Neither is usually true. When good readings sit next to bad behavior, the fault is almost always in something a static test does not see: a condition you have not applied, a connection between two good parts, or a measurement you took at the wrong moment. This tree is how you find it.
Start here: make sure "it fails" is not a hazard
If "it fails" means it trips a breaker, overheats, smells of gas, or puts water near live parts, the testing question waits. De-energize, shut off the fuel or water, and verify it is safe before you keep probing. A system that tests fine but fails dangerously is locked out until the danger is understood, not run repeatedly to recreate the failure with you next to it. Everything below assumes the failure has no safety dimension.
The core insight: a static test is not an operating test
Most meters read a component at rest, with no real load, at the temperature it happens to be right now. The fault lives in the gap between that snapshot and reality. Walk these five gaps in order from most to least common:
Gap one: you did not test under load
A part that reads perfect at idle can collapse the instant it is asked to work. This is the number-one reason for "tests good, fails anyway."
- A connection with high resistance passes a continuity check (which uses tiny current) and then drops massive voltage under full load. The wire is "good" until amperage flows.
- A component holds its rating cold and at rest, then sags, overheats, or drops out when it carries the real demand.
- A supply measures fine with nothing running and then sags below spec at peak draw.
Test it doing its actual job, at full demand, and watch the reading under that load. A reading that is fine at rest and wrong under load has named your fault.
Gap two: you tested at the wrong temperature
Thermal faults are invisible to a cold test:
- A hairline crack opens when the part heats and expands, then closes when it cools and tests fine.
- A marginal connection makes contact cold and loses it hot, or the reverse.
- A component drifts out of tolerance only at operating temperature.
Heat the suspect with a heat source or chill it with freeze spray while watching the reading. A value that jumps with temperature is the fault you could not see cold.
Gap three: the fault is between two good parts
Both components test fine because the problem is not in either one - it is in the connection, the harness, or the interface:
- A loose, corroded, or back-stabbed connection drops voltage that neither part on its end will show.
- A ground or neutral that is intact for a meter but high-resistance under load.
- A mechanical interface (a coupling, a seat, a seal) that each side passes individually but that fails when assembled and working.
Wiggle-test connectors under load. Measure across the connection, not just at each end. The reading you want is the drop the joint introduces while current flows.
Gap four: you measured the wrong thing or at the wrong moment
- A reading taken before the fault stage of the cycle misses a failure that only happens at one point in the sequence.
- An average reading hides a momentary sag, spike, or dropout. Use a meter that captures min/max or a brief event.
- The "good" reading is at the wrong test point - upstream of the actual break, or on a parallel path that masks the open.
Map the sequence and measure at the exact stage the customer says it fails, with a tool fast enough to catch a transient.
Gap five: an external condition you have not applied
The unit is genuinely fine and something outside it causes the failure:
- A shared circuit or supply that only sags when a neighbor load runs.
- An environmental trigger (humidity, rain, vibration, a door position) you have not recreated.
- A demand on the system that only the customer's real-world use produces.
The branches
| Symptom | Most likely gap | Test |
|---|---|---|
| Fine at idle, fails when working | Load (gap one) | Apply full demand, read under load |
| Fails hot or only after running | Temperature (gap two) | Heat/chill the suspect, watch the reading |
| Two parts both pass, system fails | The connection (gap three) | Measure the drop across the joint under load |
| Fails at one cycle stage only | Wrong moment (gap four) | Measure at that stage with a fast meter |
| Fails only sometimes, no internal cause | External condition (gap five) | Recreate the outside trigger |
Recap
Work the gaps in order: load, temperature, the connection between parts, the timing of your measurement, then external conditions. The judgment to bank: a static test reads a part at rest and at the wrong temperature, and faults hide in load, heat, joints, timing, and the world outside the box. When good readings meet bad behavior, you have not measured the fault yet - you have measured the part standing still.
References
- Trade-standard practice for load testing and voltage-drop measurement
- Manufacturer documentation on operating tolerances and cycle sequencing
- OSHA energized-work and lockout/tagout guidance for hazard branches (29 CFR 1910)
- See related: The Fault That Only Happens When You're Gone; Deeper Meter and Gauge Reading (related diagnostics)