AMNETPRO Insights · Instrumentation
How Do You KnowWhat to Measure?
Start with the operational question, not the sensor. The right variable is the one that helps explain a physical condition, support a decision or validate an operational improvement.
Direct Answer
A useful measurement explains something the operation needs to know.
Sensors should not be selected because they are modern, inexpensive or commonly used.
The engineering process begins by defining what is unknown or difficult to understand inside the physical operation.
Once the operational question is clear, the useful physical variable usually becomes much easier to identify.
Measurement Selection
Question → Physical Cause → Variable → Value
1. Define the operational question
Start with what the organization needs to understand, detect, control or improve.
2. Identify the physical cause
Determine which asset, condition or process relationship explains the operational question.
3. Select the variable
Choose the measurement that provides useful evidence about that physical condition.
4. Validate the value
Confirm that the measurement can support a decision, control action or measurable operational improvement.
Typical Variables
Different operations require different physical evidence.
Weak Reasons to Add Sensors
More instrumentation is not the same as better engineering.
Because the sensor exists
Technology availability is not enough reason to add a measurement.
Because more data seems better
Data volume does not automatically create better operational understanding.
Because AI needs data
AI should use measurements that already have operational meaning, not create arbitrary sensor requirements.
Because another company measures it
The useful variable depends on the specific physical operation and decision context.
Existing Measurements
The information may already exist inside the operation.
Existing sensors, meters, PLCs, SCADA systems and databases may already capture the variables required for the operational question.
The correct next step may therefore be integration or monitoring rather than additional instrumentation.
Do You Need More Sensors?→Measurement & ROI
Every measurement creates cost. The information should create value.
A sensor may require hardware, installation, calibration, communications, maintenance and long-term data management.
That investment is easier to justify when the measurement supports a specific operational or financial objective.
ROI-Driven Engineering→Measurement Before AI
AI can interpret the operation only through the physical information available to it.
Operational AI does not change the fundamental instrumentation question.
The operation still needs the right physical variables measured reliably enough to support interpretation.
Operational AI Readiness→Measurement Selection FAQ
Common questions about choosing industrial variables, sensors and measurements.
How do I know what to measure in an industrial process?
Start with the operational question. Identify which physical condition or asset behavior explains that question, then select the variable that provides the most useful evidence about it.
Should I measure every available variable?
No. Measurements should be selected because they support a useful operational decision, alarm, control action or analysis.
Can existing sensors already provide the data I need?
Yes. Existing sensors, meters, PLCs and SCADA systems should be evaluated before additional instrumentation is proposed.
What if the important variable is not currently measured?
Then additional instrumentation may be justified if that variable provides information that materially improves operational understanding or decision-making.
How does ROI affect sensor selection?
The value of the information should be compared with the cost of instrumentation, installation, communications, maintenance and data management.
Does operational AI change what should be measured?
Not fundamentally. Operational AI still depends on meaningful physical variables. The correct measurement is determined by the operation, not by the AI platform.
Continue Exploring
