Which Test First: Cheapest vs Most Diagnostic Sequencing Decision Tree

Why this matters

Every diagnostic visit is a sequence of tests, and the order you run them in determines how fast you reach a confident answer and how much of the customer's time and your truck-roll you consume. Two competing instincts pull at every tech: run the cheapest, fastest test first to clear the easy causes, or run the single most diagnostic test first to split the fault space in half. Pick wrong and you spend an hour confirming things that did not matter while the real fault sits one test away. This article gives a repeatable rule for choosing test order across any trade, so the sequence is deliberate rather than habitual.

Symptom presentation

The situation is any multi-cause fault where more than one component could produce the same observed failure. A device that will not start could be supply, control, or load. A leak could be supply, fixture, or drain. A no-heat call could be fuel, ignition, control, or distribution. The tech sees one symptom and faces a menu of possible tests, each with a different cost in time and each ruling out a different fraction of the candidate causes.

The decision is not which component failed. It is which test to run first to learn the most per minute spent.

Quick checks before sequencing

Confirm the symptom is real and current before testing anything (a stale or misreported symptom wastes the whole sequence). Read recent service history; a repeat fault changes the priors and may make a "cheap" test pointless.

List the candidate causes out loud or on the work order. You cannot sequence tests against a fault space you have not defined. Three to six candidates is typical; if you have more than eight, you have not narrowed the symptom enough.

For each candidate, note two numbers in your head: how long the test takes, and how big a fraction of the candidate list it eliminates. Those two numbers drive the whole decision.

Isolation tree

Branch A: one test is both cheap and high-yield. Run it first, always. A power-present check at the device (60 seconds, eliminates the entire supply branch) is the canonical example. When the fastest test also splits the fault space, there is no tradeoff to resolve.

Branch B: the cheapest test eliminates only one candidate; a slower test eliminates half. Prefer the high-yield test when the slow test is under a few minutes and access is already open. Splitting the fault space in half twice gets you to one suspect in two tests; clearing candidates one at a time can take five. The exception is when the slow test requires disassembly or shutting down a system the customer needs running.

Branch C: a cheap test is a prerequisite for the diagnostic test. Run the cheap one first even though it is low-yield, because the expensive test is invalid without it. You confirm voltage is present before condemning a control board; you confirm water supply is on before chasing a no-flow fixture; you confirm the thermostat is calling before measuring at the equipment. Prerequisite tests are not optional sequencing choices.

Branch D: the most diagnostic test is also destructive, costly, or requires a part you may not need. Defer it. Run the non-destructive eliminations first so that by the time you reach the expensive test, it is the only candidate left and you are confident before you cut, drain, or order.

Branch E: tests are roughly equal in cost and yield. Sequence by reversibility and by what you can observe while there. Run the test that leaves the system in a known state and that you can read without leaving the site. Prefer a measurement over a part swap; a swap that "fixes" the fault by coincidence hides the real cause.

Branch F: the fault is intermittent and the diagnostic test only works while the fault is present. Capture first. Place a logger or recreate the trigger condition before running any elimination, because a clean reading during a no-fault window proves nothing.

Confirming the sequence

State the plan before you start: "I will check supply, then control signal, then the load, in that order, because each step rules out a branch." A spoken or written sequence keeps you from drifting into part-swapping when a test comes back ambiguous.

After each test, re-read the remaining candidate list. A test result that does not shrink the list means the test was wrong for this fault; do not run it again with a different meter setting hoping for a different answer. Re-sequence against what is left.

Stop testing when one candidate remains and a confirming test has verified it. Do not keep testing to "be thorough" once the fault space is a single suspect; that is time the customer pays for and you do not need.

Next steps

Once the sequence resolves to a single confirmed cause, move to the repair tree for that component. Record the test order and results on the work order so a second visit or a second tech inherits the elimination, not a blank slate.

If the sequence resolves to "no candidate confirmed," the symptom definition was wrong or the fault is upstream of the device. Return to symptom confirmation and widen the candidate list rather than re-running the same tests.

When a single test would have resolved the call but you reached for it last, note that in the close-out for your own calibration. Test sequencing is a skill that improves only when you review which order actually worked.

References

• NFPA 70B (Standard for Electrical Equipment Maintenance; condition-based and diagnostic test ordering).
• ISO 13379-1 (Condition monitoring and diagnostics of machines; data interpretation and diagnostic sequencing).
• ASHRAE Handbook (HVAC Applications; commissioning and systematic diagnostics).
• OSHA 29 CFR 1910 Subpart S (electrical safety-related work practices governing energized testing).