Tolerance vs Accuracy

How to tell the difference between what a device measures and what the manufacturer allows

Two words show up constantly in medical equipment testing:

Published August 16, 2026 · Revised September 5, 2026

Back to Biomed Basics

What This Page Explains

This page covers:

The Simple Version

Accuracy describes how close a result is to the reference value. Tolerance describes how much variation the applicable requirement allows. If a pressure monitor reads 102 mmHg while the reference is 100 mmHg, the observed error is +2 mmHg. Whether that is acceptable is a separate question answered by the tolerance. It passes a ±3 mmHg limit and fails a ±1 mmHg limit even though the measured error has not changed.

This distinction matters at the bench because “the reading looks accurate” is not a return-to-service decision. The biomed needs the correct manufacturer or approved facility limit, a valid reference measurement, and a test setup that matches the stated conditions.

How to Use the Terms at the Bench

Keep three values separate on the worksheet: the reference or commanded value, the device's measured output or indication, and the allowed limit. The difference between the first two is the observed error. The limit tells you whether that error is acceptable. Writing those values separately prevents a common mistake: treating the analyzer reading as though it were the specification.

Also remember that the reference has uncertainty of its own. A recently calibrated analyzer is not perfect; its published accuracy, range, resolution, accessories, and setup all affect how much confidence you should place in a borderline result. When the device result sits close to a limit, repeat the test exactly as the procedure requires, check the setup, and follow the organization's uncertainty or guard-band policy rather than inventing extra tolerance.

What Is Accuracy?

Accuracy describes how close a measurement is to the correct reference value.

Example:

Pressure source:

100 mmHg.

Monitor reads:

101 mmHg.

The measurement error is:

+1 mmHg.

That is very close to the reference.

Measurement Error

A simple way to think about error is:

Measured Value - Reference Value = Error

Example:

Measured:

98 mmHg.

Reference:

100 mmHg.

Error:

-2 mmHg.

The negative sign simply means the device reads low.

Absolute Error

Sometimes you only care about the size of the error.

Reference:

100 mmHg.

Measured:

98 mmHg.

Absolute error:

2 mmHg.

Whether the device reads high or low may matter diagnostically, but the magnitude is often what determines pass/fail.

What Is Tolerance?

Tolerance defines how far a value is allowed to vary.

Example:

Target:

100 mmHg.

Tolerance:

±5 mmHg.

Acceptable range:

95 to 105 mmHg.

Anything inside that window meets the stated tolerance.

Tolerance Creates the Pass/Fail Window

Suppose the requirement is:

100 ±3 mmHg.

Then:

Lower limit:

97 mmHg.

Upper limit:

103 mmHg.

Measured:

102 mmHg.

Pass.

Measured:

104 mmHg.

Fail.

That is a simple fixed tolerance.

Accuracy Is a Performance Characteristic

A manufacturer might specify:

Pressure measurement accuracy ±3 mmHg.

That tells you how close the device is expected to measure under defined conditions.

You then use that specification as an allowable limit during testing.

Tolerance Can Apply to More Than Measurement

Tolerance is a broader concept.

You may see tolerance applied to:

Example:

Power rail:

5.0 VDC ±5%.

That is a voltage tolerance.

Fixed Tolerance

A fixed tolerance uses the same allowable error across the stated range.

Example:

±2 mmHg.

At:

50 mmHg

acceptable:

48 to 52.

At:

200 mmHg

acceptable:

198 to 202.

The allowed absolute error remains:

2 mmHg.

Percentage Tolerance

A percentage tolerance changes depending on the test value.

Example:

±5%.

At:

100 units

allowed error:

5 units.

Acceptable:

95 to 105.

At:

200 units

allowed error:

10 units.

Acceptable:

190 to 210.

The allowed error grows with the measurement.

Calculating Percentage Tolerance

Basic calculation:

Reference × Percentage = Allowed Error

Example:

Reference:

300 J.

Tolerance:

±10%.

300 × 0.10 = 30 J.

Acceptable range:

270 to 330 J.

Combined Specifications

Some specifications use more than one term.

Example:

±2% of reading ±1 unit.

Suppose reference is:

100.

Two percent:

2.

Plus:

1.

Total allowed error may be:

3 units.

Acceptable:

97 to 103.

Follow the manufacturer's exact wording and calculation method.

Percentage of Reading vs Percentage of Full Scale

These are not the same thing.

Example:

Device range:

0 to 300 mmHg.

Specification A:

±2% of reading.

At 100 mmHg:

±2 mmHg.

Specification B:

±2% of full scale.

Full scale:

300 mmHg.

Two percent:

6 mmHg.

At 100 mmHg, that would allow:

±6 mmHg.

Very different.

Read the specification carefully.

Nominal Value

A nominal value is a named or approximate design value.

Example:

24 V power supply.

That does not necessarily mean:

24.000 V.

The actual acceptable range may be:

22.8 to 25.2 V.

You need the tolerance before deciding whether the measured value passes.

Accuracy vs Precision

These terms also get confused.

Accuracy:

How close you are to the correct value.

Precision:

How consistently repeated measurements agree.

Example:

Reference:

100 mmHg.

Measurements:

110.0 110.1 110.0 110.1

Very precise.

Not accurate.

The readings are consistent, but wrong.

Accuracy vs Resolution

Resolution is the smallest increment the device displays or measures.

Example:

Device displays:

100.1 mmHg.

Resolution:

0.1 mmHg.

That does not mean the device is accurate to:

±0.1 mmHg.

Its actual accuracy specification might be:

±3 mmHg.

More decimal places do not automatically mean better accuracy.

Test Equipment Has Accuracy Too

This matters a lot.

Suppose you are testing a device that must be accurate to:

±1%.

Your analyzer is only accurate to:

±5%.

That is a problem.

The reference may not be good enough to prove the device meets its requirement.

The Reference Is Not Perfect

No measurement system is perfectly exact.

Your:

also has specifications.

That means the test result has some measurement uncertainty.

Example: Voltage Test

Device requirement:

5.00 V ±0.05 V.

Meter accuracy:

±0.10 V.

Your meter uncertainty is larger than the device's entire allowed error.

That may not be an appropriate instrument for that verification.

Better Test Equipment Gives More Confidence

If the device tolerance is tight, your test equipment should generally be substantially better than the thing you are trying to verify.

Exactly how much better depends on:

Do not invent your own ratio when a defined procedure exists.

Pass/Fail Limits

Tolerance becomes especially important at the limits.

Example:

Target:

100.

Tolerance:

±5.

Acceptable:

95 to 105.

Measured:

105.

If the limit is inclusive:

Pass.

Measured:

105.1.

Fail.

Do not casually stretch the boundary.

Borderline Results

Borderline results deserve careful handling.

Suppose limit:

105.0.

Analyzer reports:

105.0.

That may technically meet the limit.

But also consider:

If the result is right on the edge, follow the defined process rather than guessing.

Do Not Round a Failure Into a Pass

Example:

Actual result:

105.46.

Limit:

105.4.

If you round first:

105.

you may accidentally call it a pass.

Keep sufficient precision through the comparison.

Do Not Report More Precision Than You Measured

If your analyzer resolution is:

0.1 mmHg,

do not document:

100.037 mmHg.

That implies precision your instrument did not provide.

Repeatability Helps With Borderline Results

Suppose you test five times:

104.9 105.0 105.1 104.9 105.2

Now the behavior around the limit is clearer.

Repeated testing may be appropriate when permitted by the procedure.

Do Not Average Away a Failure Unless Allowed

If one measurement fails and another passes, do not automatically average them together.

Example:

99 101 108

Average:

102.7.

That does not necessarily mean pass.

Follow the manufacturer's test method.

Tolerance Can Change Across the Range

A device may have:

±2 units

in one range,

and:

±5%

in another.

Do not use one limit for every test point.

Example

Pressure specification:

0–50 mmHg:

±2 mmHg.

51–300 mmHg:

±3%.

At:

40 mmHg

use:

±2 mmHg.

At:

200 mmHg

use:

±6 mmHg.

Read the range-specific specification.

Environmental Conditions Can Affect Accuracy

Accuracy may only be guaranteed within:

A device outside those conditions may not meet the normal specification.

Warm-Up Time

Some test equipment or devices require warm-up before rated accuracy applies.

Example:

Analyzer specification valid after:

15 minutes.

If you test immediately after power-on, you may not be operating under the specified conditions.

Accessories Can Affect Accuracy

Measurement accuracy may depend on:

A bad accessory can make a good device appear inaccurate.

Calibration and Accuracy

Calibration helps establish or correct the relationship between the device and a known reference.

But calibration should not automatically be the first response to every failed accuracy test.

Ask why it failed.

Possible causes include:

Calibration should not hide another fault.

Verification vs Calibration

Verification asks:

Does the device meet the required tolerance?

Calibration adjusts the device when appropriate.

If the device already passes:

Do not calibrate just to make the number look prettier.

Example: Pressure

Reference:

100 mmHg.

Device:

102 mmHg.

Specification:

±3 mmHg.

Pass.

Do not calibrate simply because it is not exactly:

100.

It already meets specification.

Over-Adjustment Can Make Things Worse

Suppose you adjust a passing device at one test point.

You may improve that one point and worsen another.

Do not chase perfection when the device already meets the manufacturer's requirements.

Example: Infusion Pump

Set rate:

100 mL/hr.

Measured:

96 mL/hr.

Tolerance:

±5%.

Allowed:

95 to 105 mL/hr.

Pass.

The result is not exact.

It does not need to be.

Example: Defibrillator

Set:

200 J.

Measured:

188 J.

Tolerance:

±15%.

Allowed range:

170 to 230 J.

Pass.

A technician who expects exactly:

200 J

may unnecessarily adjust or repair a perfectly acceptable device.

Example: Temperature

Reference:

37.0°C.

Device:

37.4°C.

Specification:

±0.2°C.

Acceptable:

36.8 to 37.2°C.

Fail.

Even though:

37.4

looks close, it is outside the defined tolerance.

Example: Power Rail

Expected:

24 VDC.

Tolerance:

±5%.

Five percent:

1.2 V.

Acceptable:

22.8 to 25.2 V.

Measured:

23.5 V.

Pass.

Example: Battery Runtime

Required:

Minimum 90 minutes.

Measured:

88 minutes.

That is below the tolerance or minimum requirement.

Do not write:

Close enough.

If 90 is the defined minimum, 88 fails.

Typical Is Not the Same as Tolerance

Manufacturer may state:

Typical runtime: 4 hours.

That is different from:

Minimum runtime: 4 hours.

Typical values are not always intended as pass/fail limits.

Read the wording carefully.

Minimum and Maximum Limits

Sometimes there is no plus/minus tolerance.

Example:

Leak rate:

Maximum 50 mL/min.

Measured:

48.

Pass.

Measured:

55.

Fail.

Simple.

Calibration Sticker Tolerance

Do not assume the sticker on your analyzer tells you the device's tolerance.

The analyzer's calibration status tells you about the analyzer.

The medical device's service documentation tells you the medical-device acceptance limit.

Those are separate things.

Manufacturer Spec vs Facility Limit

Your facility may have approved procedures or maintenance requirements that define acceptance limits.

Use the applicable approved requirement.

Do not choose whichever number gives the result you want.

Troubleshooting With Accuracy

A failed accuracy test can help isolate a problem.

Example:

Pressure sensor reads:

+10 mmHg high across multiple test points.

That may suggest:

The pattern matters.

One-Point Accuracy Failure

If only one point fails, investigate whether:

Do not immediately assume the entire measurement system is bad.

Multi-Point Testing

Testing multiple points can show:

Example:

Reference 50 → reads 55

Reference 100 → reads 110

Reference 200 → reads 220

The error scales with the reading.

That pattern can help troubleshooting.

Offset Pattern

Reference:

50 → reads 55

100 → reads 105

200 → reads 205

Constant +5 error.

That looks different from a proportional gain error.

Specifications Help You Avoid Parts Cannon

Without objective tolerances, a tech may think:

This looks a little low.

With a specification, you can say:

It is within tolerance.

No unnecessary repair.

Specifications Also Prevent Wishful Thinking

The opposite:

It is only slightly outside.

If it fails, it fails.

Objective limits keep troubleshooting honest.

Common Mistakes

Treating Accuracy and Tolerance as the Same Word

They are related but describe different things.

Assuming More Decimal Places Means More Accuracy

That is resolution.

Ignoring Test Equipment Accuracy

The reference matters too.

Using the Wrong Percentage Basis

Reading vs full-scale can be very different.

Rounding Before Comparing

Keep enough precision.

Averaging Failed Results Without Procedure

Follow the defined test method.

Calibrating a Device That Already Passes

Within tolerance is acceptable.

Calling an Out-of-Spec Result “Close Enough”

Use the defined limit.

A Useful Troubleshooting Framework

Ask:

What is the reference value?

Then:

What tolerance applies?

Then:

What is the allowable range?

Then:

What did the device actually measure or deliver?

Then:

Is the test equipment accurate enough to support that decision?

That gives you an objective pass/fail process.

Another Useful Question

Ask:

Am I trying to prove the device is perfect, or that it meets the required specification?

Those are not the same goal.

Medical equipment does not usually need to read exactly the reference value.

It needs to perform within its approved limits.

What Did You Actually Prove?

Suppose:

Reference:

100 mmHg.

Device:

102 mmHg.

Tolerance:

±3 mmHg.

You proved:

The device measurement was within the allowed tolerance at that test point under the conditions tested.

You did not prove:

Each test supports a specific conclusion.

Final Thoughts for Biomeds

Accuracy tells you how close the device is to the reference.

Tolerance tells you how far it is allowed to be from that reference.

That distinction is simple.

But it is extremely important when deciding:

Do not chase exact numbers when the device already meets specification.

Do not accept out-of-spec results because they look close.

And do not forget that the analyzer you are using has its own accuracy limits too.

Know the reference.

Know the tolerance.

Know your test equipment.

Then let the numbers make the decision.

— Jake

Important Note

Accuracy specifications, tolerances, measurement uncertainty, acceptance limits, and test-equipment requirements vary by manufacturer, model, procedure, and facility program. Use current manufacturer documentation, approved calibration and verification procedures, and appropriately maintained test equipment when making pass/fail decisions.

Related Biomed Basics

Back to Biomed Basics