A Sensor Calibration Schedule: How Often, Which Sensors and What to Record
November 11, 2026 · 6 min read · Etzal Group Team
A cold store has kept perfect temperature records for two years. During a customer audit, a reference thermometer placed next to the room sensor reads 2.4 degrees warmer. The sensor has been drifting for months, and every record in that period is now in doubt. Nobody did anything wrong on any single day. There was simply no sensor calibration schedule, so nothing ever compared the sensor with the truth.
Why sensors go wrong
- Drift. Sensing elements age. Readings shift slowly in one direction.
- Contamination. Dust, moisture, product build-up, scaling on probes.
- Mechanical stress. Vibration, shock, a probe knocked out of position.
- Environment. Temperature extremes, humidity, corrosive atmosphere.
- Installation changes. A sensor moved, a cable replaced, a fitting altered.
- Electronics. Reference voltages and amplifiers change over years.
None of these raises an alarm. A drifting sensor keeps reporting confidently, and the dashboard looks healthy.
Step 1: list every sensor
You cannot schedule what you have not listed. For each sensor record:
- An identifier and location.
- What it measures and the range.
- Make, model and serial number.
- Required accuracy for its purpose, not just what the datasheet offers.
- What depends on it: an alarm, a control loop, a record for audit, a bill.
- Installation date and last calibration.
Step 2: classify by consequence
Ask what happens if this sensor reads wrong and nobody notices.
- Class A, critical. Safety, product quality or legal compliance, or billing. A wrong reading can harm someone, spoil product, breach a regulation or misstate money. Examples: cold chain temperature, pasteurisation temperature, gas detection, custody flow meters, weighing for trade.
- Class B, important. Process control and efficiency. A wrong reading wastes energy or material or reduces output. Examples: pressure and flow in process loops, energy meters used for cost allocation.
- Class C, indicative. Trends and general awareness. Examples: ambient temperature in a warehouse aisle, a vibration trend used alongside other signs.
Most of the effort belongs with class A. Treating every sensor alike spends the budget on the unimportant and still leaves the critical ones under-checked.
Step 3: set intervals
Use four inputs.
- Requirements. Regulations, standards, customer contracts and certification schemes may fix an interval. These come first.
- Manufacturer's recommendation.
- Consequence class.
- History. What past calibrations found.
A reasonable starting point where nothing else is specified:
| Class | Calibration | Field check between |
|---|---|---|
| A | Every 6 to 12 months | Monthly to quarterly |
| B | Every 12 months | Every 6 months |
| C | Every 24 months, or check only | Annually |
These are starting points, not standards. Adjust them with evidence, as described below.
Step 4: field checks between calibrations
A field check is a quick comparison that needs no laboratory.
- Side-by-side comparison. A calibrated portable reference placed next to the installed sensor, read after both have stabilised.
- Known points. An ice bath for temperature at zero degrees. A reference weight for a scale. A calibration gas for a gas detector.
- Cross-checks. Two sensors measuring the same thing should agree. A third value calculated from others, such as a mass balance or energy balance, should match.
- Plausibility rules in software. A reading that never changes, changes too fast, or sits outside physical limits is flagged automatically.
Record the reference used, both readings and the difference. If the difference exceeds the tolerance, the sensor goes for calibration now and the records since the last good check are reviewed.
Step 5: who calibrates
- Accredited laboratory. For class A sensors and wherever an auditor or regulator needs evidence. Look for accreditation to ISO/IEC 17025 by the national accreditation body, and a certificate that states traceability to national standards and the measurement uncertainty.
- In-house with reference instruments. Suitable for many class B and C sensors, provided the reference instruments are themselves calibrated by an accredited laboratory and the method is written down.
- On site by a service provider. Useful where removing a sensor stops a process.
The reference should be notably more accurate than the sensor being checked. A common rule of thumb is a ratio of four to one.
Step 6: what to record
For each calibration:
- Sensor identifier.
- Date and who performed it.
- Reference instrument and its certificate number.
- As-found readings. What the sensor read before any adjustment.
- As-left readings. What it read after.
- Tolerance and pass or fail.
- Next due date.
The as-found reading is the one people skip, and it matters most. It tells you whether the sensor was within tolerance during the period just ended. If it was not, you have a decision to make about the product or the records from that period. It is also the data that lets you adjust the interval.
Step 7: adjust intervals with evidence
After a few cycles, look at the as-found results.
- Always well within tolerance. Consider extending the interval, where requirements allow.
- Close to the limit each time. Keep the interval, or shorten it.
- Out of tolerance when found. Shorten the interval, and ask why: environment, installation or the sensor type.
A schedule that never changes is not using its own data.
When a sensor fails calibration
- Take it out of service or adjust it.
- Record the as-found error.
- Assess the impact: what decisions, products or records relied on it since the last good check?
- Decide what to do about those: re-inspect, quarantine, notify, or document that the error did not matter.
- Record the decision.
This step is the reason for field checks. The shorter the gap since the last good reading, the smaller the period in question.
Using the monitoring system itself
A connected system can do part of the work.
- Reminders. Due dates per sensor, with alerts before they pass.
- Status on the data. Readings from an overdue sensor are marked as such.
- Drift detection. Comparing redundant sensors and flagging a growing gap.
- Stuck and out-of-range detection.
- Records in one place. Certificates attached to the sensor record.
Software cannot replace the comparison with a physical reference, but it can make sure the comparison happens and show when something looks wrong in between.
Common mistakes
- No list of sensors.
- One interval for everything.
- Calibrating only when an audit is announced.
- Skipping as-found readings.
- Reference instruments that are themselves overdue.
- Certificates filed and never read.
- Replacing a sensor without checking the new one once installed.
- No action when a sensor fails, beyond adjusting it.
We include a sensor register, classes and calibration reminders in the monitoring systems we build under IoT and Control Systems. If you have a system in place and are unsure how far to trust its readings, contact us.
Summary
List every sensor and what depends on it. Classify by the consequence of an unnoticed error. Set calibration intervals from requirements, manufacturer advice, class and history, and add quick field checks in between. Use accredited laboratories for critical sensors. Record as-found and as-left readings, adjust intervals with that evidence, and when a sensor fails, assess what relied on it.
Frequently asked questions
How often should sensors be calibrated?
There is no single interval. Start from the manufacturer's recommendation and any regulatory or customer requirement, then adjust by risk and by evidence. A sensor that guards product safety or billing is checked more often than one that shows a trend. If records show a sensor staying within tolerance, the interval can be extended, and if it drifts, shortened.
What is the difference between calibration and verification?
Calibration compares a sensor against a reference of known accuracy and documents the difference, often followed by adjustment. Verification, or a field check, is a quicker confirmation that the sensor still reads within tolerance, for example comparing it with a portable reference. Verification between calibrations catches problems early at low cost.
Do IoT sensors need calibration if they are new?
New sensors come with a factory accuracy specification, and some with a calibration certificate. That describes the sensor when it left the factory. Installation, environment and time change readings. Check a new sensor against a reference once it is installed, and put it on the schedule like any other.
Related service
WhatsApp AI Chatbot
We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.
