Proving It's Fixed Before the Customer Tests It
Why this matters
The most expensive way to find out a repair did not hold is to let the customer find out for you. A fix that fails after you leave is a callback, a refund argument, and a dented reputation, and it costs far more than the few minutes of verification that would have caught it. The discipline is simple: you prove the repair before you pack up, under the same conditions that produced the fault, so the system has already passed the customer's test before they ever run it. This article is the verification method that turns "should be good" into "I watched it work."
The core rule: reproduce the original symptom, then prove it gone
You cannot verify a fix you never reproduced. The whole method hangs on one rule: recreate the exact condition that caused the fault, confirm the symptom is gone under that condition, and only then call it fixed.
If the unit only failed when hot, run it until hot and confirm it holds. If it only failed at peak load, load it to peak. If it only failed on a cold start, let it cool and start it cold. A repair that passes a gentle, comfortable test but was never pushed to the failing condition is not proven; it is hoped.
Test under the failing condition, not the easy one
The single most common verification failure is testing under conditions easier than the ones that caused the fault.
- Intermittent that needed vibration, heat, or time - reproduce that. Tap, flex, heat-cycle, or run it long enough for the fault to have appeared before. A fix that survives deliberate disturbance is real; one that survives only a static check is not.
- Load-dependent fault - run it under real load, not idling. Many faults hide at no load and appear only under draw.
- Condition-dependent fault - match the temperature, the humidity, the cycle point that triggered it.
If you genuinely cannot reproduce the failing condition on site (a fault that only appears in deep winter, say), say so plainly and tell the customer what to watch for. Honest uncertainty beats a false all-clear.
Run a full cycle, not a spot check
A part that powers on is not a system that works. Run the equipment through a complete duty cycle: start, run, reach steady state, and shut down normally. Watch the whole arc.
| What to confirm | Why it matters |
|---|---|
| Starts cleanly | Catches a startup fault a running check misses |
| Reaches normal operating state | Confirms it performs, not just powers |
| Holds steady under load | Catches a fault that needs time or heat to appear |
| Readings land in spec | Confirms the numbers, not just the behavior |
| Shuts down and restarts normally | Catches a cycling or reset fault |
A spot check that catches the unit "running fine for a second" is exactly the test the customer would have passed too, right before it failed on them.
Check the numbers, not just the behavior
"It seems to work" is not verification. Read the values that prove it: the voltage, the amp draw, the pressure, the temperature, the flow, against the nameplate or a known-good baseline. A system can appear to run while a reading is marginal and headed back toward failure. The meter sees what the eye does not. A fix is proven when the behavior is right and the numbers are in spec, both.
Verify you did not break something else
A repair can fix one thing and disturb another. Before you leave, confirm the systems you touched and the ones adjacent still work. Did opening that panel loosen a neighboring connection. Did replacing that part change a setting downstream. Did the shutoff you closed get reopened. A quick check of the surrounding function catches the self-inflicted second fault that would otherwise be your next callback.
Document the proof
Write down what you verified and what it read: "ran full cycle, confirmed symptom gone under load, amp draw and pressure in spec." This does three things: it protects you if the customer later claims it was never fixed, it tells the next tech what good looked like, and it forces you to actually run the test instead of assuming. A verification you cannot describe is one you probably did not do. The repair is finished when you have watched it pass the customer's test before they ever gave it.
References
- Trade-standard commissioning and acceptance-test practice (verify under design conditions).
- Manufacturer documentation on operating specifications and post-repair checkout.
- ISO 13379-1 (Condition monitoring and diagnostics of machines; post-maintenance verification).
- See related: Baseline vs Faulty Comparison Diagnosis Method; The Fault That Moved After You Touched It Decision Tree.