MPE and tolerance are not the same thing. Learn how maximum permissible error differs from tolerance, and how to check an instrument against both.
MPE and tolerance are both limits, but they limit different things. Maximum permissible error (MPE) is the largest measurement error an instrument is allowed to show. Tolerance is the allowed variation in the thing you are making or controlling. If you treat them as one number, you can pass an instrument that cannot judge your product, or reject good product because of an instrument that is fine.
This guide explains the difference in plain terms and shows how the two numbers work together. It also covers what the Australian standards say, and how to read a calibration certificate against both.
MPE stands for maximum permissible error. The international vocabulary of metrology (JCGM 200:2012) defines it as the extreme value of measurement error, with respect to a known reference quantity value, permitted by specifications or regulations for a given measurement, instrument or system.
Two points follow from that definition.
In Australian trade measurement the document is often legislation. The National Measurement Institute (NMI) publishes National Instrument Test Procedures (NITPs), which assess whether measuring instruments measure within maximum permissible errors and comply with their certificates of approval. The MPE is specific to the instrument type. For example, NITP 2 assesses area-measuring instruments against MPEs specified in Schedule 1 of the National Trade Measurement Regulations 2009.
An MPE can be a fixed value, a percentage, or a mix. An NMI pattern-approval specification for certain length-measuring instruments gives ±0.5% for analogue indication and ±0.5% plus 0.5 of a scale interval for digital indication. That example shows why you should read the exact wording. A digital instrument can carry an extra allowance for its resolution that an analogue one does not.
On a balance data sheet you will usually see the same idea expressed as linearity, repeatability and readability. Those figures together describe how far the display can sit from the true value.
Tolerance describes how much a characteristic of the product or process may vary and still be acceptable. A shaft diameter of 20.00 mm ±0.10 mm has a tolerance of ±0.10 mm. A fill weight of 500 g with a lower limit of 495 g has a tolerance on that side. A cold room that must hold 2 °C to 8 °C has a tolerance band of 6 °C.
Tolerance is set by design, a product specification, a customer, a pharmacopoeia or a food safety plan. It says nothing about the instrument you use to check it.
Tolerances on parts also combine. When several parts stack up in an assembly, their individual tolerances add in ways that affect the final fit. If you work with assemblies, this walkthrough of tolerance stack-up equations shows the arithmetic with a worked example. That is a design-side question. It is separate from how good your measuring instrument is.
The VIM is direct on this: the term “tolerance” should not be used to designate MPE. The 2021 ISO standard on statistical methods in process management makes the same distinction, treating tolerance and MPE as separate things rather than synonyms.
The confusion is understandable, because some fields do use “tolerance” for an allowance on a reading. The Australian Bureau of Meteorology, for instance, defines tolerance as a combined allowance calculated from instrument error and the ability to read the reference, as ±√(A² + B²). Note what that is doing. It is a derived allowance for a specific purpose, built from an instrument error figure. It is not the same as the instrument’s MPE.
In day-to-day work on a factory floor, you will hear “the scale is out of tolerance” for an instrument that has failed its calibration limit. That shorthand is common, and most people know what is meant. In a quality manual, a procedure or an audit response, use the precise terms:
| Question | MPE | Tolerance |
|---|---|---|
| What does it limit? | Error of the measuring instrument | Variation of the product or process |
| Who sets it? | Manufacturer, standard or regulation | Designer, customer or product specification |
| Where do you find it? | Data sheet, NITP, pattern approval, calibration procedure | Drawing, specification, recipe, batch record |
| What do you do with it? | Decide whether the instrument is fit to be used | Decide whether the product conforms |
| Does calibration test it? | Yes, against traceable standards | No, the calibration tests the instrument, not your product |
The two numbers meet when you ask a practical question: can this instrument tell me whether my product is inside tolerance?
The 2021 ISO standard sets out the logic. Suitability depends on comparing measurement-process uncertainty with the specification. The acceptance interval is the tolerance interval reduced by uncertainty. In other words, the instrument’s error takes a bite out of the tolerance before you measure anything.
The numbers below are illustrative, not from a client job.
You machine a part with a diameter tolerance of ±0.10 mm. You measure it with a gauge whose MPE is ±0.02 mm, and your calibration certificate reports that the gauge’s error and uncertainty stay within that.
Now change the gauge to one with an MPE of ±0.08 mm. The instrument may be perfectly in specification for its own class, and it passes calibration. But the acceptance limits shrink to ±0.02 mm, and most good parts fall in the grey zone. The instrument met its MPE and is still unsuitable for this tolerance.
That is the answer to the question we hear most often. Meeting the MPE proves the instrument is performing as designed. It does not prove the design is adequate for your requirement. A 2015 IMEKO paper describes this as metrological confirmation: equipment must conform to requirements for its intended use, which are distinct from product requirements and involve measuring range, resolution and MPE.
Many quality teams use a ratio between tolerance and instrument accuracy to screen instruments. The best known is the test uncertainty ratio (TUR). Ratios are a useful first filter, but they hide assumptions about how error is distributed and whether uncertainty has been included. We cover the limits of the approach in TUR vs TAR.
The same logic applies one level down, to the standards used to calibrate your instrument. NITP 0 states that for all reference standards, instruments or materials, the uncertainty must be no greater than one-third of the applicable MPE of the instrument being tested. If a trade instrument has an MPE of ±0.3 g at a test load, the reference used to check it needs an uncertainty of no more than 0.1 g.
NITP 0 also describes the verification and in-service inspection procedures used to assess whether approved-pattern instruments measure within applicable MPEs and comply with their certificates of approval.
A certificate does not say your instrument is suitable. It reports what the instrument did against a reference, with an uncertainty. You then compare those results with the limit that matters.
Work through these steps.
On the decision rule, NATA’s guidance on ISO/IEC 17025:2017 explains that a rule describing how measurement uncertainty is used when reporting pass or fail against specification limits is agreed with the customer at contract review. It is stated in the report unless it is inherent in the specification or standard. So ask for the rule before the job, not after the result.
The MSA Test Method 2 from the Metrology Association of Australia shows one explicit rule. Unless the user specifies otherwise, a gauge is deemed compliant if its correction and uncertainty, assessed separately, are not larger than the accuracy tolerance at all test points. Other rules add a guard band or share the risk. None is wrong, but they give different answers on the same data, so write down which one applies.
For a deeper look at the uncertainty line on a certificate, see our guide to calibration uncertainty and certificate checks.
An MPE check is only as sound as the reference behind it. The NMI states that its calibration and testing results are traceable to Australia’s primary measurement standards, and that its relevant laboratories operate under an ISO/IEC 17025-compliant quality system within NATA accreditation.
Below the national level, accredited calibration laboratories carry that traceability to your site. NATA also states that accredited facilities must not report an uncertainty smaller than their Calibration and Measurement Capability (CMC). That protects you from a certificate that claims better performance than the laboratory can demonstrate.
If you want more on how the system fits together, read Understanding NATA calibration standards in Australia and our page on traceable metrology.
We calibrate scales, balances, force gauges, thermometers, pH meters and data loggers for manufacturers, laboratories, and food and pharmaceutical sites. The most common problem we see is not an instrument that has failed. It is an instrument that passed, in a role it was never suited to.
A typical case is a bench scale bought for general use that ends up checking a tight fill weight. The scale meets its specification and carries a valid certificate. But its error and uncertainty consume most of the product tolerance, and the production team spends its day arguing about borderline results.
Our advice is simple.
Because we service multiple brands regardless of where you bought the equipment, we can look at what you already have and tell you whether it fits the job. If it does not, we can say so plainly and help you select something that does.
— Kaz
The same principle plays out differently by instrument.
Balances and scales. Repeatability, linearity and readability all contribute. A balance with fine readability can still have a coarse MPE if linearity is poor. Our article on verifying balance calibration tolerance covers four tests technicians can run. If you are seeing drift or inconsistent batches, see scale accuracy issues.
Thermometers and data loggers. Cold chain and process temperature limits are narrow bands. The instrument’s error and the logger’s resolution reduce the usable band, so a ±0.5 °C device on a 2 °C to 8 °C window leaves less margin than the headline range suggests.
pH meters. The tolerance on the product may be ±0.05 pH, so a meter and electrode with a wider error budget cannot support it, however well it is calibrated.
Trade-use instruments. Where a scale is used for trade, the MPE is set by regulation and verified under the NITPs. You do not choose it. Read NMI pattern approval to see how approval and verification fit together.
Calibration establishes, under specified conditions, the relationship between an instrument’s indication and the corresponding known value. Adjustment alters the instrument to bring it within allowable MPEs, as NITP 2 describes. Recording as-found data before any adjustment matters, because it tells you whether product measured since the last calibration might have been judged with an instrument outside its limit.
For instruments with less strict MPEs, some evaluators justify less disruptive methods such as inline verification rather than stringent laboratory testing. For tight tolerances, a full calibration with a low-uncertainty reference remains the safer choice.
We provide NATA-accredited and traceable calibration on site or in our workshop, along with preventative maintenance, repairs, trade verification and equipment hire. We are NATA accredited and ISO 9001 certified, and we offer 24/7 emergency support. Tell us the tolerance you work to and we will help you check that your instrument, and its calibration, are suited to it.
MPE stands for maximum permissible error. It is the largest error an instrument may show against a known reference, as permitted by a specification or regulation. Outside metrology the same letters mean other things, such as a medical condition or a music protocol, so check the context.
No. MPE limits the error of the measuring instrument, while tolerance limits the variation of the product or process. The metrology vocabulary says “tolerance” should not be used to designate MPE, even though people often use the words loosely on the shop floor.
Yes. If the instrument’s MPE and uncertainty take up too much of your product tolerance, you cannot reliably tell good from bad. The instrument is working as designed, but it is the wrong instrument for that requirement.
Under ISO/IEC 17025:2017, the decision rule is agreed with the customer at contract review and stated in the report, unless the specification or standard already defines it. Raise it when you request the calibration, so the pass or fail statement means what you expect.
Check the manufacturer’s data sheet, the relevant standard, or for trade instruments the NITP and the National Trade Measurement Regulations. If you cannot find it, we can help you identify the right limit for your equipment.
