Last Thing Changed: A Fault Diagnosis Principle

Why this matters

When a system worked yesterday and fails today, something changed. The single most productive question you can ask is "what changed last?" because the change and the failure are almost always connected. Chasing exotic causes while ignoring the thing somebody touched an hour ago wastes the whole call. This principle costs nothing and routinely solves a fault in minutes.

The core idea

A stable system does not usually fail on its own schedule. It fails because a condition moved: a part was serviced, a setting was adjusted, power was cycled, weather shifted, something new was added nearby, or a worn component finally crossed its limit. The most recent change is the prime suspect, and you check it first.

This is not superstition. It is probability. The universe of things that changed recently is tiny compared to the universe of things that could theoretically fail, so a recent change is a far higher-odds bet.

What counts as a "change"

Cast a wide net when you interview. Changes the customer does not think to mention:

• A recent repair or service, even by another tech, even unrelated work in the same area.
• A setting or control adjustment, a new schedule, a swapped knob position.
• A part replacement - and whether the new part is correct and installed right.
• Power events: an outage, a tripped breaker that was reset, new equipment sharing the supply.
• Environmental shifts: a heat wave, a freeze, heavy rain, dust from nearby construction.
• Added load: new equipment, more demand, a different usage pattern.

Ask "what is different now versus when it last worked?" and keep asking until you have a list.

The most dangerous change: the last repair

If the system failed shortly after someone worked on it, suspect that work first - including your own. Common after-service faults:

• A connector left loose or not fully seated.
• A wire pinched, chafed, or routed against something hot or moving.
• A valve, switch, or setting left in the wrong position.
• A "new" part that is the wrong spec, defective out of the box, or installed backward.
• A step skipped (a bleed, a purge, a calibration) that the system needs to run right.

This is not about blame. Even careful work introduces faults, and a freshly serviced system that fails is telling you exactly where to look.

When "nothing changed"

Customers often insist nothing changed, yet something always did. When the interview turns up nothing:

• Look for slow changes: a component drifting out of spec, gradual wear, building corrosion, a filter or screen clogging over weeks. The "change" is time crossing a threshold.
• Look for changes they would not connect to the symptom, like work in another room or a different appliance added to the same circuit.
• Check what you can verify rather than what you are told: positions, settings, and connections that should be correct.

A failure with truly zero change is rare. Usually "nothing changed" means "nothing I noticed."

How to use it without tunnel vision

The last-thing-changed principle points you to the highest-odds suspect, but it is a starting point, not a verdict. Check the recent change first, prove or clear it, then fall back to normal isolation if it comes up clean. Do not force the evidence to fit the change - if the measurements say otherwise, follow the measurements.

References

• Trade-standard change-correlation troubleshooting practice.
• Reliability-engineering guidance on change as a leading failure cause.
• See related: Diagnose Before You Replace, Root Cause vs Symptom Fix.