Customer Says "It Only Does It Sometimes" Translation Decision Tree

Why this matters

"It only does it sometimes" is the customer's way of describing an intermittent fault, and intermittent faults are where diagnostic time goes to die. The phrase sounds like there is no pattern, but "sometimes" almost always has a hidden trigger the customer has not connected to the symptom. The whole translation job is to convert "sometimes" into a condition: sometimes when it is humid, sometimes when both pumps run, sometimes after it has been on for an hour, sometimes on Mondays. Find the condition and the intermittent fault becomes reproducible and diagnosable. Miss it and you arrive to a unit that works perfectly and leave with nothing. This article structures that translation for any trade.

Symptom presentation

The customer reports a fault that is real and recurring but not constant. They cannot make it happen on demand, and crucially, the tech usually cannot either, because the visit window is a small slice of time and the fault may not appear in it. The customer's framing is that the fault is random; the diagnostic premise is that it almost never is.

The danger is twofold. First, the tech finds nothing and either walks away (reads as incompetent) or replaces a part on a guess (often wrong). Second, the customer's confidence erodes because the problem they know is real went unaddressed. The translation from "sometimes" to "when X" is what rescues the call.

Quick checks at intake

Ask the customer to walk through the last several occurrences in detail. Not "what's wrong" but "tell me about the last few times it happened." Time of day, weather, what else was running, what they had just done. Patterns hide in the specifics customers do not volunteer.

Ask what the occurrences have in common. The customer may not have noticed that it always happens on hot afternoons, or always when the laundry runs, or always within an hour of startup. Prompting them to compare events surfaces the trigger.

Look for physical evidence and logged faults. Stains, marks, fault counters, error logs, and event histories often record intermittent events the customer-visible interface did not show. A logged fault is fully diagnosable even when the fault is currently absent.

Isolation tree

Branch A: the "sometimes" correlates with an environmental condition. Temperature, humidity, supply voltage, pressure, or time of day. A breaker that trips only on humid days is moisture intrusion; a unit that fails only on hot afternoons is a load or charge issue; a fault only at night may be a low-voltage or thermal-contraction issue. Match the condition to the affected subsystem and walk that tree.

Branch B: the "sometimes" correlates with concurrent events. It happens when two devices run together, when a large load starts, when a storm passes, when a neighbor's equipment cycles. These point at supply-side transients (voltage sag, pressure drop, interference). Capture with a logger while the trigger is active rather than testing at rest.

Branch C: the "sometimes" correlates with run time or duty cycle. The fault appears only after the system has run a while, or only on heavy use, or only on cold starts. Run-time-dependent faults point at thermal expansion opening a connection, a component overheating, or a load-sensitive part. Reproduce by running the system to the condition that triggers it.

Branch D: the "sometimes" is truly patternless after careful questioning. The leading suspect is a loose or marginal connection that opens randomly with vibration or thermal cycling. Inspect and re-torque connections at terminals, junctions, and plugs per spec. Random-with-no-pattern is the signature of an intermittent open, not a failing core component.

Branch E: the frequency is increasing. A fault happening more often each week will reproduce soon. Schedule a return at an interval where capture is likely, set the customer's expectation that the next visit is the diagnostic visit, and deploy a logger if available.

Branch F: the frequency is steady over a long time. A symptom at the same low rate for a year often traces to a chronic upstream condition (hard water, marginal supply voltage, low gas pressure) the device tolerates most of the time. The investigation moves upstream of the device.

Branch G: the "sometimes" clears with a specific action (power cycle, opening a window, rebooting a controller). The reset pattern is diagnostic. Clears on power cycle points at control electronics or firmware; clears on ventilation points at an air or pressure issue; clears on a controller reboot points at communications.

Confirming diagnosis

Confirm by reproducing the fault under the identified condition. If forcing the trigger (heat soak, dual load, humidity, run time) makes the fault appear, you have converted intermittent to reproducible and can diagnose directly. That reproduction is the strongest possible confirmation.

When you cannot reproduce on the visit, the close-out is a capture plan, not a guess: a logger left in place, a defined trigger for the customer to call on, a recording or photo of the next occurrence, and a scheduled return if warranted. Document what was tested and ruled out so the next visit is narrower.

Tell the customer that "sometimes" usually means "when something specific happens," and that finding the trigger is the diagnosis. Setting that expectation prevents the visit reading as a failure when no part is replaced on the first trip.

Next steps

Once a trigger is identified, walk the isolation tree for the affected subsystem under that condition. When the fault is captured by a logger or event history, diagnose from the captured data directly.

If no trigger emerges and no fault is captured, leave the customer with a defined monitoring plan rather than a speculative repair. A part swapped on an unconfirmed intermittent fault is a callback in waiting.

References

• IEEE Std 100 (Authoritative Dictionary of IEEE Standards Terms; intermittent-fault definitions).
• ISO 13379-1 (Condition monitoring and diagnostics of machines; intermittent symptom analysis).
• NFPA 70B (Standard for Electrical Equipment Maintenance; connection integrity and intermittent faults).
• ASHRAE Handbook (HVAC Applications; intermittent and condition-dependent fault diagnosis).