Measurement Uncertainty in Plain English

Why every measurement has some uncertainty, why it matters near pass/fail limits, and how biomeds should think about analyzer readings without turning calibration into a math class

A biomed puts a device on an analyzer, gets a number, and compares that number with a specification.

Published September 2, 2026 · Revised September 6, 2026

Back to Biomed Basics

What This Page Explains

This page covers:

The Simple Version

Suppose the acceptable output is 95 to 105 units. An analyzer result of 100.1 sits comfortably inside that window, while 104.9 sits very close to the upper boundary. Neither display is a perfect statement of the true value: the analyzer, accessories, environment, setup, resolution, and test method all contribute some uncertainty.

Uncertainty does not mean the measurement is useless or that a technician can invent extra tolerance. It describes the reasonable doubt around the result. The closer a result is to a decision limit, the more important it is to follow the approved procedure, understand the analyzer specification, avoid premature rounding, and apply the organization's acceptance or guard-band policy consistently.

Worked Example: A Result Near the Limit

A device measures 104.9 where the upper limit is 105.0. Before calling it good, verify the correct test point, units, warm-up, zero, accessories, analyzer range, and calibration status. Record the unrounded result and check whether the service procedure or quality system defines a decision rule for results near the boundary.

Repeating a valid test can reveal instability, but repeated readings do not erase analyzer uncertainty or authorize averaging unless the procedure says to average. If the result remains borderline, document the evidence and follow the approved escalation path. “The screen was below 105” may be mathematically true while still being incomplete measurement practice.

A Measurement Is an Estimate

When you measure something, you are estimating a physical quantity.

Examples include:

You use a test instrument because your senses cannot determine those values accurately enough on their own.

But the test instrument is also a physical system.

It has limits.

Digital Does Not Mean Exact

A digital analyzer may show:

5.00 L/min

with two decimal places.

That does not mean the actual flow is exactly:

5.000000000 L/min.

The display resolution simply tells you how finely the device presents the result.

That is different from how accurately it can determine the true value.

Resolution

Resolution is the smallest change the instrument can display or distinguish.

Example:

An analyzer displays:

4.99 5.00 5.01

Its displayed resolution may be:

0.01 L/min.

But that does not automatically mean its accuracy is ±0.01 L/min.

Accuracy

Accuracy describes how close a measurement is expected to be to the true value.

An analyzer might have a specification such as:

±1% of reading

or:

±0.5% plus a fixed amount.

The exact form varies by instrument.

Precision

Precision is about how closely repeated measurements agree with each other.

You can have a system that is very precise but inaccurate.

Example: Precise but Wrong

Measure the same reference five times:

101.8 101.9 101.8 101.9 101.8

Those results are highly repeatable.

But if the true reference value is:

100.0

the measurement system may be consistently biased high.

Accurate but Less Precise

Another analyzer gives:

99.7 100.2 100.0 100.3 99.8

The readings vary more, but they cluster around the true value.

That is a different measurement characteristic.

Accuracy and Precision Are Not the Same Thing

A test instrument can be:

In practical biomed work, you usually care about whether the total measurement system is good enough to evaluate the manufacturer's specification.

What Is Measurement Uncertainty?

Measurement uncertainty expresses the reasonable doubt around a measured result.

It answers something like:

Based on this measurement method and equipment, how tightly do I actually know this value?

Instead of thinking:

The value is exactly 100.0

you think more like:

The best measured value is 100.0, with some small uncertainty around it.

This Is Not the Same as Saying the Measurement Is Bad

Every valid measurement has uncertainty.

The goal is not to eliminate uncertainty completely.

That is impossible.

The goal is to understand whether the uncertainty is small enough for the decision you are making.

Test Equipment Has Its Own Specification

Suppose you are checking an infusion pump that must be within:

±5%.

If your analyzer is only accurate to:

±10%,

then it is not a useful instrument for proving the pump meets a ±5% specification.

Your test equipment has to be good enough for the job.

Better Test Equipment Reduces Uncertainty

If your analyzer is substantially more accurate than the device tolerance, your pass/fail decision becomes easier.

This is one reason calibration programs care about:

A Useful Practical Concept: Test Accuracy Ratio

A common measurement concept is comparing the accuracy of the test instrument with the tolerance of the device being tested.

You may hear terms such as:

Test Accuracy Ratio, or TAR.

The exact quality-system requirement varies, but the general idea is simple:

Your reference should be significantly better than the thing you are trying to evaluate.

Why?

Suppose a temperature device is allowed:

±2°C.

Your reference thermometer is only accurate to:

±2°C.

If the device reads 1.5°C high, you have a problem.

Is the medical device wrong?

Is the reference wrong?

You do not have enough separation between them.

Better Reference

If the reference thermometer uncertainty is much smaller, perhaps a fraction of a degree, you have much greater confidence about whether the device itself meets its ±2°C requirement.

Reference Standards

A reference standard is something you trust more than the device under test.

Examples include:

You are effectively saying:

I trust this reference enough to use it as the basis for judging the medical device.

That Trust Has to Come From Somewhere

This leads to:

Traceability.

Traceability

Measurement traceability means that the measurement can be connected through an unbroken chain of calibrations to recognized reference standards.

A simplified chain might be:

Medical Device

↓

Biomed Analyzer

↓

Calibration Laboratory Reference

↓

Higher-Level Standard

This helps establish confidence that your analyzer's measurement has a known relationship to accepted standards.

Calibration Certificate

Your analyzer's calibration certificate may contain information such as:

That certificate contains far more technical information than a sticker that simply says:

Calibrated.

Why Two Calibrated Analyzers Can Disagree

Suppose Analyzer A reads:

199 J.

Analyzer B reads:

201 J.

Does that mean one is broken?

Not necessarily.

Both values may fall within each analyzer's expected measurement uncertainty.

Example

True defibrillator output might reasonably be around:

200 J.

One analyzer reads slightly low.

The other reads slightly high.

Both may still be functioning within specification.

Do Not Expect Every Meter to Show the Identical Last Digit

If two calibrated instruments differ by a very small amount, the difference may be normal.

Look at:

before deciding one is defective.

Repeatability

Repeatability describes how consistent a measurement is when performed repeatedly under the same conditions.

Example

Defibrillator delivers:

198 J 199 J 198 J 199 J 198 J

That is fairly repeatable.

Now suppose it delivers:

170 J 205 J 185 J 201 J 176 J

That variation is itself important even if one individual shot happens to land inside specification.

One Passing Measurement May Not Tell the Whole Story

If the manufacturer procedure requires multiple measurements:

Do them.

Repeatability can reveal intermittent or unstable performance.

Setup Creates Uncertainty Too

The analyzer itself is not the only source.

Your test setup can affect the result.

Examples include:

Pressure Example

You test a pressure device.

The analyzer is highly accurate.

But the tubing leaks.

Your measurement system is still bad.

The reference instrument cannot compensate for a poor setup.

Flow Example

You measure ventilator flow through an adapter that introduces unexpected resistance.

The analyzer may be accurate, but the setup changes the system being measured.

Temperature Example

You place the device probe and reference probe at different depths in a water bath.

One region is slightly warmer.

Now you are not actually comparing the same temperature.

Electrical Example

You measure voltage at a point with a poor ground reference.

Your meter may be perfect.

Your test point is not.

Measurement Uncertainty Is About the Whole Measurement Process

Think:

Reference Instrument

+

Test Setup

+

Environment

+

Method

+

Device Behavior

All of these influence confidence in the final value.

Environment

Temperature, humidity, altitude, and other environmental factors may affect some measurements.

Manufacturers may specify operating ranges for both:

Example

A precision analyzer may have one accuracy specification at:

23°C ±5°C

and a different behavior outside that range.

Read the specification.

Warm-Up Time

Some test equipment needs time to stabilize after power-up.

If the manufacturer requires:

15 minutes warm-up

and you begin precision measurements immediately:

You may be adding unnecessary uncertainty.

Pass/Fail Limits

Suppose manufacturer specification is:

100 ±5.

Acceptable range:

95 to 105.

Your measured result:

100.

Easy pass.

Clearly Outside

Measured:

110.

Easy fail.

There is plenty of distance between the measurement and the limit.

Near the Boundary

Measured:

104.9.

This is where uncertainty matters.

If your measurement uncertainty is meaningful relative to the remaining:

0.1

margin, you should not behave as if the last decimal place provides absolute certainty.

Guard Bands

Some calibration systems use:

Guard bands.

A guard band creates an internal acceptance limit slightly tighter than the manufacturer's absolute specification to account for measurement uncertainty.

Simple Example

Manufacturer limit:

95 to 105.

Internal acceptance might be:

95.5 to 104.5

depending on the uncertainty policy.

This gives additional confidence that accepted equipment truly falls inside the manufacturer's limit.

Do Not Invent Your Own Guard Band

Whether guard bands are used and how they are calculated should come from:

Not personal preference.

Why This Matters for Biomeds

Most of the time, your results will not sit directly on the limit.

The pump will either be:

Clearly in specification

or:

Clearly out of specification.

Those are easy decisions.

The uncertainty conversation becomes most important when the result is:

Right near the boundary.

Example: Infusion Pump

Specification:

±5%.

Programmed:

100 mL/hr.

Allowable:

95 to 105 mL/hr.

Analyzer reads:

99.5.

No meaningful concern.

Analyzer reads:

104.9.

Now review:

before declaring absolute certainty.

Longer Infusion Tests

At low flow rates, short test durations can introduce more variation.

Example:

Pump programmed:

1 mL/hr.

Testing for only a few minutes may produce a weak estimate because the delivered volume is tiny.

Longer Duration Improves Confidence

A longer test may average out:

and give a more meaningful result.

Example: Defibrillator Energy

Set:

200 J.

Manufacturer tolerance:

Defined by service manual.

Analyzer reads comfortably in range.

Pass.

If it reads exactly at the specification boundary:

Look at:

Example: NIBP

NIBP static pressure check:

Reference analyzer:

200.0 mmHg.

Monitor:

199 mmHg.

No issue if within tolerance.

If monitor reads right on the failure boundary:

Repeat the measurement under controlled conditions.

Example: Temperature

Reference:

37.00°C.

Device:

37.1°C.

Fine under a ±1°C specification.

Device:

37.95°C under a ±1°C specification.

Now reference uncertainty and stabilization become more important.

Example: Ventilator Tidal Volume

Set volume:

500 mL.

Analyzer reports:

499 mL.

No meaningful concern if tolerance is broad enough.

Analyzer reports:

Right at the manufacturer's maximum allowable error.

Now check:

Analyzer Configuration Matters

Some analyzers require settings for:

Wrong configuration can produce a systematic error.

BTPS and STPD

Respiratory measurements may use different reference conditions.

You may encounter terms such as:

These describe how gas volume is corrected for:

If the ventilator and analyzer report values using different reference conditions, they may disagree even when both are working correctly.

This Is Not Always Calibration Error

It may be a comparison error.

Make sure you are comparing like with like.

Resolution Can Mislead You

An analyzer may display:

100.000

but have accuracy only to:

±0.5.

The extra digits look impressive.

They do not make the instrument more accurate.

Significant Digits

Do not report more certainty than your measurement system supports.

Writing:

100.000000 mL

when your analyzer is accurate to only a much broader range adds meaningless precision.

Calibration Drift

Test equipment can drift over time.

That is one reason it is recalibrated periodically.

As-Found Failure of Test Equipment

Suppose your infusion analyzer returns from calibration and is found significantly out of tolerance.

Now you may need to consider:

Which medical devices were tested using that analyzer since its previous acceptable calibration?

This is a quality-system question.

Why Analyzer Records Matter

If you know:

you can investigate intelligently.

Calibration Traceability Is Not Bureaucracy for Its Own Sake

It allows you to answer:

What did we trust when we made this measurement?

Measurement Error vs Device Error

Suppose every pump tested today reads:

3% high.

Before adjusting ten pumps, ask:

Did the pumps all drift in the same direction at the same time, or is the analyzer/setup biased?

Patterns matter.

Cross-Check

If something looks suspicious:

Use another calibrated reference when appropriate.

Known Reference

A secondary check can help determine whether the first analyzer is behaving normally.

Do Not Use an Unverified Tool to “Confirm” a Calibrated Analyzer

The second instrument needs to be trustworthy enough to add meaningful evidence.

Uncertainty and Troubleshooting

Measurement uncertainty applies outside formal calibration too.

Suppose you measure a:

5 V rail.

Meter reads:

4.98 V.

You probably do not care about hundredths if the circuit operates normally over a broad voltage range.

Context Matters

The required measurement quality depends on the decision.

Checking:

Is power present?

requires less precision than:

Is this pressure sensor calibrated to within 0.5%?

Use the Right Tool for the Question

You do not need laboratory-grade precision for every troubleshooting step.

But you need enough accuracy to support the decision.

Example

Question:

Is 24 V supply completely missing?

Basic calibrated multimeter is sufficient.

Question:

Is a low-level sensor reference within a very tight tolerance?

You may need better equipment and a more controlled method.

Measurement Uncertainty Is Not an Excuse to Ignore Specs

Do not say:

Everything has uncertainty, so close enough.

That is not the lesson.

The manufacturer specification still matters.

Uncertainty helps you understand the confidence of the measurement used to evaluate that specification.

Avoid False Precision

Bad thinking:

The limit is 105 and I measured 104.99, therefore it definitely passes.

Better thinking:

The result is extremely close to the limit. I need to consider the required test method and measurement uncertainty before making the final call.

Avoid Unnecessary Panic Too

Measured:

100.1.

Limit:

95–105.

Analyzer uncertainty:

Small.

Do not overcomplicate an obvious pass.

Practical Biomed Rule

The farther the result is from the limit:

The less uncertainty usually matters to your decision.

The closer the result is:

The more attention you should pay.

What Information Should You Know About Your Test Equipment?

At minimum, you should be able to find:

You do not need every specification memorized.

Know where to find it.

Manufacturer Test Procedures Matter

If a service manual tells you:

follow it.

Those conditions are part of the measurement method.

Do Not Improve the Procedure Into a Different Procedure

Changing the setup may change:

Calibration Lab vs Biomed Shop

A calibration laboratory may perform detailed uncertainty calculations that are far beyond normal daily biomed work.

That is okay.

A biomed usually needs practical awareness, not a metrology degree.

What a Biomed Should Understand

You should understand:

  1. Your analyzer is not infinitely accurate.
  2. The test setup can introduce error.
  3. The reference should be more accurate than the device tolerance being evaluated.
  4. Measurements near the pass/fail boundary deserve more attention.
  5. Calibration and traceability establish confidence in the reference equipment.

That gets you surprisingly far.

Common Mistakes

Treating Every Displayed Digit as Exact

Resolution is not the same as accuracy.

Ignoring Test-Equipment Accuracy

The analyzer is part of the measurement.

Testing With an Improvised Setup

A good analyzer cannot rescue a bad method.

Calling Something a Fail Based on a Tiny Boundary Difference Without Reviewing the Procedure

Near-limit results need care.

Calling Something a Pass Just Because the Display Is Barely Inside the Limit

Same problem in the opposite direction.

Comparing Two Different Measurement Conditions

Make sure both values refer to the same thing.

Assuming “Calibrated” Means Zero Uncertainty

Calibration does not make an instrument perfect.

A Useful Measurement Framework

When a result matters, ask:

What is the manufacturer's allowed range?

Then:

What is my analyzer's accuracy or uncertainty?

Then:

Am I using the correct setup and test conditions?

Then:

Is the result comfortably inside the limit, clearly outside it, or right near the boundary?

If it is near the boundary:

Slow down.

Verify the method.

Repeat if required.

Review the relevant acceptance procedure.

Another Useful Question

Ask:

If my analyzer were slightly wrong within its own specification, would that change my pass/fail decision?

If the answer is:

No,

your result is probably comfortably clear.

If the answer is:

Yes,

the measurement deserves more scrutiny.

What Did You Actually Prove?

If your analyzer reads:

100.0,

you proved:

Your measurement system produced a best estimate of approximately 100.0 under the conditions of the test.

You did not prove:

The true value is exactly 100.000000 with no uncertainty.

If the result is comfortably inside specification using properly calibrated test equipment and the approved procedure:

You have strong evidence the device meets the requirement.

If the result sits directly on the boundary:

You have less room for uncertainty and should evaluate the measurement more carefully.

Final Thoughts for Biomeds

Measurement uncertainty sounds intimidating because it comes from the world of calibration laboratories and metrology.

The practical idea is much simpler:

Every measurement has some doubt around it.

Most of the time, that doubt is small enough that your decision is easy.

If the device is supposed to be:

95–105

and you measure:

100,

move on.

If you measure:

104.99,

do not let the extra decimal places create false confidence.

Look at:

The goal is not to turn every PM into a statistics exercise.

The goal is to understand when the number on the screen is strong evidence and when it is close enough to the boundary that you need to ask one more question:

What did you actually prove?

— Jake

Important Note

Measurement uncertainty, acceptance rules, guard bands, calibration traceability, test accuracy requirements, and pass/fail decision rules may be defined by the device manufacturer, test-equipment manufacturer, calibration program, laboratory, facility policy, or applicable quality standards. Follow the approved procedure for the equipment being tested and use calibrated test equipment suitable for the required measurement range and tolerance.

Related Biomed Basics

Back to Biomed Basics