Two Readings Don't Agree (Decision Tree)

Why this matters

You take a reading, then take another that should confirm it, and the two numbers do not line up. The reflex is to pick the one that matches your hunch and move on. That reflex throws away a finding. When two honest measurements of what looks like the same thing disagree, the gap is usually not noise and not always a bad tool. Sometimes both readings are correct and they are telling you the system is doing something you have not accounted for. Learning to reconcile instead of choose is what separates a diagnosis from a guess.

Start here: are they even measuring the same thing

Before you call it a contradiction, rule out the boring explanation. Two numbers can differ for reasons that are not a disagreement at all.

• Different point. A reading at the source and a reading at the load should differ if anything sits between them. That is a drop, and a drop is data.
• Different instant vs an average. A live gauge shows this second; a logged or averaged value shows a window. A transient will show on one and be smoothed out of the other.
• Different units or scaling. Gauge vs absolute, one scale vs another, a value that is already compensated vs a raw one. Confirm both are in the same terms.
• Different state. Was the system running, loaded, and at the same temperature for both, or did the condition change between them?

If any of these differ, that is very likely your whole answer, and it is not a fault in either reading. Line the conditions up and re-measure both before going further.

The mindset: what would make both true

Assume, as a working position, that both readings are honest. Then ask what physical condition would produce both at once. This is the opposite of averaging. Averaging two disagreeing diagnostic numbers manufactures a middle value that describes nothing real, and unless averaging is a documented method for that specific measurement, do not do it.

Reconciling forces the useful question. If the source reads full and the load reads low, what is between them that eats the difference? If a live probe reads high and the system's own display reads normal, which one sits where the actual problem is? The answer to "what would make both true" is often the diagnosis itself.

When the difference is the point

A delta between two points is one of the most useful readings you can take, and you only get it by trusting both numbers rather than discarding one.

• A pressure or flow drop across a filter, coil, valve, or run tells you the restriction is between the two taps. A big drop where there should be a small one names the restricted part.
• A voltage drop across a connection, switch, or length of conductor under load points to a high-resistance joint you would never find by trusting only the source reading.
• A temperature split across a component tells you whether it is doing its job.

If the two readings straddle a component and disagree by more than that component should cause, the component (or its connections) is your suspect. The disagreement located the fault.

Rank the sources by what they are authoritative for

When the readings genuinely measure the same quantity at the same point and still disagree, stop treating them as equals. Each source is trustworthy for some things and weak for others.

• A direct measurement at the physical point usually beats a value inferred, calculated, or reported from elsewhere, though a well-placed permanent sensor can beat a hurried handheld reading in a bad spot.
• A high-resolution instrument beats a coarse one when the decision hinges on a small margin; on a gross pass/fail, either is fine.
• A source you can verify against a known reference beats one you cannot.

If they still contradict

Two readings of the same quantity, at the same point, in the same state, that still disagree and land on opposite sides of your decision line, cannot be resolved by preference. Bring in a third independent check, and if the suspicion is that an instrument itself is lying, verify the tool before you trust either number (see the related articles). Do not present one as settled fact until something independent breaks the tie.

Recap

1. Rule out different point, instant, unit, or state before calling it a disagreement.
2. Assume both are honest and ask what condition makes both true.
3. A delta across two points is a finding, not an error to erase.
4. When they measure the same thing and still fight, rank by authority, then confirm with a third independent source.

References

• Trade-standard practice on differential measurement (pressure drop, voltage drop, temperature split) as a diagnostic method
• NIST traceable calibration principles for field measurement confidence
• See related: Two Tools Give Two Different Readings; The Connected System's Data Contradicts Your Meter