Start with the symptom and consequence
Define what appears wrong: fixed value, noisy value, slow response, offset, intermittent loss, disagreement with another instrument or implausible process behavior. Also understand whether the measurement is informational, used for control, or associated with a safety-related function.
Compare independent evidence
Use other process indicators, redundant measurements, trends, known operating states and diagnostic status to determine whether the process itself changed. Agreement between two devices is useful evidence but not proof that either is correct.
Follow the chain
Check the measurement conceptually from process interface to sensor to transmitter to signal path to input conversion to software scaling. The safest practical checks and isolation methods depend on site procedures and equipment.
Use calibration and failure history
Repeated drift, moisture ingress, plugged impulse paths, sensor fouling or connector faults can reveal a recurring failure mechanism. History can justify changes in maintenance interval, installation or technology.