Compare the indicated value with a suitable reference under stated conditions.
Industrial measurement guide
Sensor Accuracy vs Precision: What Is the Difference?
Short answer: accuracy describes how close a sensor result is to a suitable reference value. Precision describes how closely repeated readings agree with one another. A sensor can repeat the same wrong value, so a tight cluster alone does not prove accuracy.
Image: xsz sensor manufacturing and application support.
Repeat the same measurement and quantify how tightly the values group.
Set an error limit, a repeatability limit, and the conditions behind each claim.
Direct answer
Close to the reference and close to each other are two different tests.
For a practical sensor check, record several readings at a stable reference point. The distance between the average and the reference shows bias. The spread among the readings shows precision under those test conditions.
Use a suitable reference. If the average is consistently high or low, the system has a systematic effect or bias that must be understood.
Repeat the measurement under defined conditions. Range or sample standard deviation can describe the observed spread.
Practical rule: never approve a sensor from one reading. One result cannot show repeatability, drift, hysteresis, or whether a stable offset is present.
Three patterns to recognize
Look at both the center and the spread.
These diagrams use a target as a memory aid. A real sensor must still be checked across its working range, temperature, mounting, direction of approach, and time in service.
Precise but not accurate
Readings agree closely, but the group is offset from the reference. A stable zero, span, installation, or calibration bias may be present.
Tight spreadBiased center
Centered on average but not precise
The mean may land near the reference by chance, while individual readings vary too much for reliable control, sorting, or inspection.
Centered meanWide spread
Accurate and precise in this test
The readings group tightly around the reference. This is strong evidence at this point and condition, but it is not proof across the full application.
Tight spreadCentered result
Language that prevents mistakes
Accuracy, precision, trueness, and repeatability answer different questions.
Metrology uses these terms carefully. In an RFQ or acceptance plan, replace vague words with a limit, method, range, statistic, and test condition.
| Term | Core question | Useful evidence | Common mistake |
|---|---|---|---|
| Accuracy | How close is the result to a suitable reference? | Reference comparison with stated conditions and uncertainty. | Treating one plus/minus value as universal. |
| Precision | How closely do repeated results agree? | Replicate readings plus range or standard deviation. | Calling a tight cluster accurate without checking its offset. |
| Trueness | Is the mean close to the reference? | Mean of repeated values compared with a reference. | Using trueness and precision as synonyms. |
| Repeatability | Does the same setup repeat over a short period? | Same method, device, location, operator, and conditions. | Assuming a bench result covers field changes. |
| Reproducibility | Do results agree after defined conditions change? | Comparison across stated changes such as operator, time, or location. | Changing conditions without recording which ones changed. |
For more detail on short-term spread, read what sensor repeatability means and how it is measured.
Why a sensor can repeat the wrong value
Bias can stay stable while every reading looks consistent.
A stable offset is a systematic effect. Because it can remain nearly constant during a short test, the sensor may look excellent when only the spread is checked. That is why calibration and repeat testing are complementary.
- Zero offset: the result is displaced by a similar amount across part of the range.
- Span error: the output changes too much or too little as the input changes.
- Nonlinearity: error varies across the range instead of following the assumed response.
- Hysteresis: the result depends on whether the input approached from the rising or falling direction.
- Installation influence: mounting stress, heat conduction, target angle, pressure lines, vibration, or EMI shifts the complete chain.
A correction may reduce a known bias, but it does not prove that span, drift, environmental effects, and uncertainty are acceptable. Recheck the required points after any adjustment.
The minimum useful calculation
Record the center, the bias, and the spread.
These values describe one dataset. They do not replace a complete uncertainty analysis or a test across the full working range.
The average of the repeated readings. Compare this center with the reference value.
A positive result reads high; a negative result reads low relative to the selected reference.
A smaller value means less observed spread for this point, method, and set of conditions.
Reading check tool
Calculate the center and spread of repeated readings.
Enter one reference value and at least two sensor readings. Optional limits let you compare the observed mean error and range with your own acceptance plan.
Calculated result
This calculator describes only the entered readings. It does not include reference uncertainty, full-range behavior, hysteresis, drift, environmental influence, or calibration traceability.
The complete measurement chain
A good sensor can still produce a bad PLC or displayed value.
Accuracy at the sensing element is only one contributor. Installation, power, signal conversion, wiring, PLC input performance, scaling, filtering, and control logic can all change the value used for a production decision.
Define the physical quantity and where it exists in the process.
Include target, probe location, bracket stress, immersion, port, or optical geometry.
Check transmitter, amplifier, linearization, analog range, and response setting.
Include voltage at the device, shielding, grounding, cable resistance, and interference.
Review input error, conversion, engineering units, filtering, and sample timing.
Verify the displayed, logged, alarmed, or controlled value that affects production.
How to read the datasheet
Do not compare two accuracy numbers until their basis is equivalent.
A supplier may state error as percent of full scale, percent of reading, an absolute unit, a typical value, a maximum value, or a composite band. Some numbers include nonlinearity, hysteresis, repeatability, and temperature effects; others do not.
Ask this before ranking models:
“Under what conditions, over which range, and including which effects?”
Use the complete industrial sensor datasheet review guide when model suffixes, outputs, distance, timing, housing, and protection also affect the decision.
Error basis
Percent of full scale, percent of reading, span, output, or absolute engineering units? Typical or maximum?
Included effects
Does the number include nonlinearity, hysteresis, repeatability, calibration tolerance, conversion, and temperature?
Reference conditions
Confirm supply, warm-up, mounting, target or media, range point, response setting, ambient temperature, and load.
Actual operating point
Calculate the stated error at the portion of range your machine uses. An oversized range can waste tolerance.
Time and environment
Review warm-up, drift, recalibration, vibration, ingress, EMI, target change, and process temperature.
Measurement-chain scope
Clarify whether the claim ends at the sensor output or includes the transmitter, PLC input, scaling, and display.
A practical verification sequence
How to test sensor accuracy and precision.
- Define the measurand and decision.State the quantity, operating range, process tolerance, and whether the limit applies to every reading, the mean, or a control action.
- Choose a suitable reference.Use reference equipment with an appropriate range, status, and uncertainty for the acceptance limit.
- Stabilize the setup.Control or record warm-up, supply, mounting, temperature, target or media, filtering, sample time, and process influences.
- Test multiple points and directions.Use the normal region plus relevant low, mid, high, rising, and falling points. One point cannot reveal span, nonlinearity, or hysteresis.
- Collect repeated readings.At each point, calculate the mean, bias, range, and a suitable measure of spread.
- Compare separate limits.Evaluate permissible error, repeatability, response, drift, and environmental requirements independently.
- Document the result and action.Keep raw readings, conditions, criteria, corrections, uncertainty information, and any adjustment, repair, replacement, or installation change.
Calibration is not automatically adjustment. It establishes a relation between reference values and indications under stated conditions. After adjustment, recheck the required points.
Buyer and OEM checklist
What to put in a sensor RFQ or acceptance plan.
A fair supplier comparison starts with one controlled requirement sheet. Do not ask only for “high accuracy.” State the process decision and the evidence needed to release the model.
Measurand and range
Quantity, normal operating region, full range, units, media or target, and process tolerance.
Permissible error
Maximum error at defined points and conditions; basis as full scale, reading, or absolute units.
Repeatability method
Number of repeats, approach direction, settling time, statistic, point, warm-up, and allowed spread.
Environment and mounting
Temperature, vibration, washdown, EMI, target or media, bracket, cable, supply, and response setting.
Calibration evidence
Reference equipment, status, test points, readings, conditions, uncertainty statement, and date.
Measurement-chain scope
Sensor output, transmitter, PLC module, scaling, filtering, display, and final value to verify.
For supplier qualification beyond the measurement claim, use the industrial sensor supplier selection guide.
Frequently asked questions
Sensor accuracy vs precision FAQ
Can a sensor be precise but not accurate?
Yes. It can produce nearly identical readings that are consistently high or low relative to a suitable reference. The tight cluster shows good precision in that test; the offset shows poor trueness and potentially unacceptable measurement error.
Can a sensor be accurate but not precise?
A group of scattered readings can average near the reference, but individual readings may still be unsuitable for control or inspection. In practical specifications, set a permissible error for the relevant result and a separate repeatability limit.
Is repeatability the same as precision?
Repeatability is precision under repeatability conditions, typically the same procedure, device, location, operator, and short time period. Precision is broader and can also be evaluated under intermediate or reproducibility conditions.
Is resolution the same as accuracy?
No. Resolution is the smallest input change that produces a perceptible output change. Fine resolution can coexist with offset, nonlinearity, noise, drift, and poor calibration.
What does plus or minus percent of full scale mean?
The error band is based on the device's full measurement span rather than the current reading. Its exact definition and included effects must be confirmed in the datasheet. At the low end of a wide range, a full-scale-based error can consume a large part of the process tolerance.
Does calibration make a sensor accurate?
Calibration establishes a relation between reference values and sensor indications under stated conditions. It can reveal error and support correction or adjustment, but it does not automatically adjust the device or guarantee untested field conditions.
How often should an industrial sensor be calibrated?
There is no universal interval. Base it on process risk, manufacturer guidance, drift history, environmental severity, quality-system requirements, and the consequence of a wrong measurement. Use evidence from previous checks to adjust the interval.
Which matters more: accuracy or precision?
It depends on the decision. Closed-loop control, compliance measurement, and product release often need both. A threshold task may prioritize stable repeatability and adequate margin at one decision point. Start with the process tolerance and failure cost.
Technical sources
- BIPM / JCGM 200:2012, International Vocabulary of Metrology: definitions for measurement accuracy, precision, trueness, bias, repeatability, reproducibility, resolution, uncertainty, and calibration.
- JCGM VIM online definitions with informative annotations: searchable terminology and notes for practical interpretation.
- NIST Technical Note 1297: guidance for evaluating and expressing uncertainty in measurement results.
Define the requirement before choosing the model
Share the range, tolerance, output, mounting, and acceptance conditions.
xsz sensor can review the sensing task, operating point, interface, environment, repeatability expectation, and sample-validation plan before a model is selected.