Device Works on Test Leads but Not Once Installed: Decision Tree

Why this matters

A device that energizes and operates perfectly when jumpered to test leads, then goes dead or behaves erratically the moment it is landed in the box, is one of the most common "the part is bad" misdiagnoses in the field. The part is almost never bad. The bench condition gave the device a short, fat, low-impedance source and a solid return. The installed condition adds branch-circuit length, terminations, possible open or high-resistance neutrals, switch legs, and load-side faults the bench never saw. Chasing this correctly saves a needless second part swap and surfaces the real fault, which is frequently a connection that will burn later if left alone.

Symptom presentation

The technician confirms the device functions on the bench or on flying leads off a known-good outlet. Reinstalled into its permanent location, one of several things happens: it is fully dead, it operates weakly or intermittently, it works but trips protection, or it works until load is applied and then sags out. Smart devices and electronics show the cleanest version of this because they need a stable neutral and steady voltage; a resistive load like a heater may "sort of" work on a marginal circuit and mask the fault.

The tell is that the variable is the wiring path, not the device. Whatever changed between the two tests is the suspect.

Quick checks

Before pulling the device back out, confirm the supply the installed location actually delivers.

• Measure hot-to-neutral at the device location under no load. You want a tight 114 to 126 V on a nominal 120 V circuit. A reading well below that, or one that wanders, points upstream.
• Measure hot-to-ground and neutral-to-ground. Neutral-to-ground should read near 0 to a couple volts. A high neutral-to-ground reading flags a poor neutral or a borrowed neutral.
• Verify the device is landed on the same conductors you metered. A switch leg, half-hot outlet, or load terminal of a GFCI/AFCI can put the device downstream of something that opens.
• Reseat every termination by hand. Backstabbed connections and a neutral that only catches a strand of the conductor are the classic culprits.

Isolation tree

Work the path from the device back toward the panel, splitting the run in half each step.

Branch 1: Dead when installed, good on leads. Re-meter at the device terminals as wired. If voltage is present but the device is dead, the device terminals are not the same node your meter touched: a backstab released, a wire-nut spun off a stranded neutral, or the device is on a switched/load-controlled point. If voltage is absent at the device terminals but present at the box's incoming conductors, the connection between feed and device is open. If voltage is absent at the box entirely, the upstream splice, switch, or GFCI/AFCI parent has opened.

Branch 2: Weak or intermittent. This is voltage drop or a high-resistance connection. Load the circuit (plug in a known load near the device) and watch the voltage. A drop of more than about 5 V under modest load on a short residential run means a loose or corroded connection or an undersized/long conductor. Walk each junction box from device to panel; thermal at a connection rises with load, so a connection warm to the touch under load is found.

Branch 3: Works, then trips. The bench load and installed load differ. A GFCI or AFCI parent may see leakage or arcing the bench never presented. Confirm whether the device is downstream of a protective device, and test that protective device's own line/load wiring. A bootleg neutral or a neutral-to-ground bond downstream will trip a GFCI on first real load.

Branch 4: Works alone, fails with other loads on the circuit. Suspect a shared neutral (multiwire branch circuit) with an open or wrong-phased neutral. The device floats when its neighbor draws current. Verify both ungrounded conductors land on opposite legs and the shared neutral is continuous.

Confirming diagnosis

You have the fault when you can make the symptom appear and disappear by acting on one connection or conductor, not by swapping the device again.

• Voltage-drop confirmation: meter at the device under load reads low; meter at the upstream box under the same load reads normal. The drop lives in the segment between, and that segment's connection or conductor is the fault.
• Open-connection confirmation: continuity (power off, locked out) across the suspect segment reads open or a meaningful resistance where it should read near zero.
• Neutral fault confirmation: neutral-to-ground voltage that climbs with load, or current measured on the equipment grounding conductor with a clamp, confirms a misrouted or borrowed neutral.

Re-land the one suspect connection, re-energize, and the device that "only worked on leads" now works installed. That is the proof.

Remediation

• Replace backstabbed terminations with screw terminals or listed pigtails; torque to the device's marked value.
• Repair or replace the open/high-resistance splice; use listed connectors rated for the conductor material and count.
• For shared-neutral float, confirm opposite-leg landing and a continuous neutral; install handle ties where required for the multiwire branch circuit.
• For protective-device trips, correct downstream neutral bonds or arcing terminations rather than defeating the device.
• Re-verify voltage under load and confirm the formerly weak reading now holds within tolerance.

References

• NFPA 70 (National Electrical Code), Article 110.14, Electrical Connections and termination torque.
• NFPA 70, Article 210, Branch Circuits, and Article 250 for grounding and bonding of the equipment grounding conductor.
• NFPA 70, Article 100 and 210.4, Multiwire Branch Circuits.
• NFPA 70B, Recommended Practice for Electrical Equipment Maintenance, connection inspection and thermography guidance.
• UL 486A-486B, Wire Connectors, for listed termination practice and conductor material compatibility.