Uncertainty of Measurement: The Most Undervalued Number in Your Lab
Most labs think hard about uncertainty of measurement exactly twice. Once when the assessor shows up, and once when they build CMC claims for their scope of accreditation. Between those two moments, the uncertainty value sits quietly on the certificate, and almost nobody uses it for anything. That is a problem. The number that describes how much you can trust a result should shape decisions every day. It should not just survive an audit. This article makes the case that measurement uncertainty deserves far more credit than the ISO/IEC 17025 industry gives it. It explains where the neglect comes from. It then walks through the fundamentals every lab and every lab customer should understand.
What Uncertainty of Measurement Actually Tells You
Start with the fundamentals. Every measurement result is an estimate, not the truth. Uncertainty of measurement is the parameter that describes the spread of values you could reasonably attribute to the measurand. The GUM (JCGM 100:2008) built the framework the whole industry uses. ISO/IEC 17025:2017 then turned that framework into a requirement through Clause 7.6.
Read that definition again, because it carries a message the industry keeps missing. The uncertainty value is not decoration on a certificate. It defines the limits of what the result can prove. A voltage reading of 10.0000 V with an expanded uncertainty of 0.9 mV tells one story. The same reading with an uncertainty of 90 mV tells a very different one. The first result can support a tight tolerance decision. The second one cannot. Without the uncertainty, the two results look identical, and a customer has no way to tell them apart.
So the uncertainty estimate answers the only question that matters at the end of a calibration or test. How much can I trust this number, and what decisions can it support? Treat it as paperwork, and you throw that answer away.
Why Measurement Uncertainty Gets Ignored Between Audits
The neglect follows a predictable pattern. A lab builds its uncertainty budgets during initial accreditation, often with real effort and care. The budgets produce CMC values for the scope. The scope gets published. Then the budgets go into a folder. The lab pulls them out again when the next assessment cycle demands a review.
Why does this happen in a standard that requires uncertainty evaluation for every calibration under Clause 7.6.2? Because nothing in the daily workflow forces the number to matter. Technicians report it because the certificate template has a field for it. Customers file the certificate because their own auditor wants to see one. Nobody on either side of the transaction acts on the value. As a result, the most informative number on the document becomes the least examined one.
The root cause sits one layer deeper, and it has a name: simple acceptance.
Simple Acceptance Lets Labs Report Uncertainty Without Using It
Clause 3.7 of ISO/IEC 17025:2017 defines a decision rule. It is the rule that describes how measurement uncertainty is accounted for when stating conformity with a requirement. Clause 7.8.6.1 then requires the lab to document its decision rule and consider false accept and false reject risk. The lab must also apply the rule when it states conformity.
In practice, most labs satisfy those clauses with the simplest rule available. Under simple acceptance, the acceptance limit equals the tolerance limit. If the measured value lands inside the tolerance, the item passes. If it lands outside, the item fails. The uncertainty of measurement appears on the certificate, but it plays no role in the pass or fail call. The measured value alone decides.
Here is the uncomfortable part. Simple acceptance is the most widely used decision rule in the industry. The standard accepts the practice when the lab documents it and the customer agrees to it. Precision ISO is not telling you the practice is noncompliant. However, widely accepted and technically sound are not the same thing. Picture a measured value that lands near a tolerance limit, with an uncertainty interval that straddles it. Simple acceptance can declare a pass while the true value sits out of tolerance. That is a false accept, and the customer carries the risk without ever seeing it. The lab reported the uncertainty. Nobody used it. This is precisely how a whole industry trains itself to treat measurement uncertainty as a formality.
Not sure what decision rule to use?
Precision ISO helps labs document defensible decision rules and communicate them to customers.
✓ No obligations
✓ 30 Min Free
Accreditation Bodies Are Tightening the Rules on Simple Acceptance
The industry is starting to correct itself, and the pressure is coming from the top. Guidance documents such as ILAC G8 and ASME B89.7.3.1 have long described how uncertainty should factor into conformity decisions. ANSI/NCSL Z540.3 went further and built a requirement around limiting false accept risk to 2 percent. What has changed is enforcement. Accreditation bodies now look harder at decision rules during assessments. Several have published requirements that push labs beyond bare simple acceptance.
The clearest example involves the test uncertainty ratio, or TUR. The TUR compares the tolerance you are testing against to the uncertainty of the measurement you used to test it. A TUR of 4:1 means your measurement uncertainty is four times smaller than the tolerance. That margin keeps the risk of a wrong call low. Some accreditation body policies now expect calibration labs that use simple acceptance to demonstrate how uncertainty was taken into account. A common way to satisfy that expectation is to state and meet a minimum TUR. The traditional benchmark is 4:1. When the TUR falls below the stated minimum, the lab must respond. It can flag the result, apply guard banding, or use a different decision rule.
Two things follow from this trend. First, verify your own accreditation body’s current requirements before you rely on any summary, including this one. The policies differ between bodies and change over time. Second, notice the direction of travel. The industry’s referees are saying the quiet part out loud. An uncertainty value that never touches a decision was never doing its job. Labs that build uncertainty into their decisions now will find the next assessment cycle far less eventful. Labs that wait to be told will not.
The Uncertainty Budget: The Primary Tool of the Trade
If uncertainty of measurement is going to drive decisions, the lab needs a defensible way to calculate it. The uncertainty budget is that tool. A budget is a structured table that lists every meaningful contributor to the doubt in a measurement. It converts each contributor to a common form. It then combines them into a single value you can report and act on. Every CMC on your scope came from one. Every uncertainty on a certificate should trace back to one.
The example below shows a typical uncertainty budget for a 10 V DC voltage measurement. The specific numbers are illustrative, but the structure is the structure you will see across nearly every discipline.
Example Uncertainty Budget: DC Voltage at 10 V
Illustrative uncertainty contributors for a 10 V DC measurement.
| Source of Uncertainty | Type | Input (± mV) | Distribution | Divisor | Standard u (mV) |
|---|---|---|---|---|---|
| Repeatability (10 readings) | A | 0.120 | Normal | 1 | 0.120 |
| Reference standard calibration | B | 0.500 | Normal (k = 2) | 2 | 0.250 |
| Reference standard drift | B | 0.300 | Rectangular | √3 | 0.173 |
| Resolution of unit under test | B | 0.500 | Rectangular | √3 | 0.289 |
| Temperature effects on reference | B | 0.100 | Rectangular | √3 | 0.058 |
uc 0.440 mV
k 2
U = k × uc ±0.88 mV
All contributors in this example are expressed in millivolts and have a sensitivity coefficient of 1. Values are illustrative.
What Each Field in the Uncertainty Budget Means
Each column earns its place, so walk through them once and the table reads itself. The source names the contributor in plain language, such as repeatability or reference drift. The type records how you evaluated the source. Type A comes from statistical analysis of repeated readings. Type B comes from other information such as certificates, specifications, and experience.
The distribution describes the assumed shape of each contributor. The divisor converts the raw value into a standard uncertainty based on that shape. Normal data divides by 1. A certificate value divides by its stated coverage factor, and a plain tolerance divides by the square root of 3.
The sensitivity coefficient scales each source by how strongly it drives the result. The units columns keep every conversion honest along the way. In this example every coefficient equals 1, because every source already speaks in millivolts. A budget built from an equation will carry coefficients that do real work.
The standard uncertainty is each source after the divisor and coefficient are applied, stated in the units of the result. Relative uncertainty, when you add that column, expresses each source as a fraction of the measured value. That form lets you compare budgets across ranges.
The degrees of freedom capture how much data supports each entry. They also feed the effective degrees of freedom for the whole budget. Finally, the combined standard uncertainty merges every row through the root sum of squares. The coverage factor then expands it into the certificate value at roughly 95 percent confidence.
Want this table as a working spreadsheet?
Download the Precision ISO Measurement Uncertainty Budget Template
The Budget Is a Framework for the Time of Measurement, Not a Filing Exercise
Here is where the biggest missed opportunity in the industry lives. A budget built once for a CMC describes your best case capability. It assumes near ideal conditions and a nearly ideal device. The measurement you perform on a customer’s item next Tuesday is not that measurement. The customer’s device brings its own resolution, its own stability, and its own behavior on the bench.
The budget solves this if you let it. Because the table separates every contributor, you can update the rows that change from job to job. Then you recombine in seconds. When the method allows repeated readings, take them. Put the actual repeatability of that item into the Type A row. Update the resolution row for the actual device. Then recombine, expand, and report an uncertainty of measurement that describes the job you actually performed.
Contrast that with common practice. Many labs report their scope CMC value on every certificate, regardless of the item on the bench. Sometimes that shortcut is defensible, such as when the device contributes almost nothing and conditions match the budget’s assumptions. Too often, though, labs stretch the shortcut past its breaking point. A noisy handheld meter does not perform like the reference-grade unit the CMC assumed. So a CMC value on that certificate understates the real doubt in the result. The customer then makes tolerance decisions with an uncertainty estimate that was never true for their instrument. Labs that take repeatability data at the time of measurement close that gap. The certificate they issue finally means what it says.
Why the Uncertainty Value Is So Valuable to Customers
Everything above serves the lab. Now flip the perspective, because the customer gains even more when the uncertainty estimate is real. A trustworthy uncertainty value unlocks work that a bare number never could.
Test uncertainty ratio. With a real uncertainty on the certificate, a customer can compute the TUR for every tolerance the instrument checks. That single ratio tells them whether their calibration provider was good enough for the job. Better yet, it tells them before a bad decision ships a bad product.
Interlaboratory comparisons. Two labs measure the same artifact and get different numbers. Are they in agreement or in conflict? Without uncertainties, nobody can say. With them, a normalized error calculation settles the question in one line. Proficiency testing under ISO/IEC 17025 depends on exactly this logic.
Troubleshooting measurements. When a result looks wrong, the budget tells you where to look. A result outside its expected interval points to a real change in the process, the environment, or the equipment. A result inside the interval is noise, and chasing it wastes the week. Uncertainty of measurement draws the line between a signal and a ghost.
Monitoring drift. Tracking an instrument’s readings over successive calibrations only works when each point carries its uncertainty. The interval tells you whether the movement between calibrations is genuine drift or ordinary variation. Genuine drift demands a shorter interval, and ordinary variation demands nothing.
Risk assessments and decision rules. Clause 7.8.6.1 asks labs to weigh false accept and false reject risk. Customers must do the same math when they set guard bands or accept simple acceptance from a provider. Neither calculation can start without a credible uncertainty value.
Reverse traceability studies. Does a reference standard come back from calibration out of tolerance? Then you must evaluate every measurement it touched since its last good calibration. The uncertainty attached to those measurements determines which customer results survive the review and which ones trigger recalls and notifications. A lab with honest, job-specific uncertainty estimates scopes that investigation in hours. A lab with copied CMC values scopes it in guesswork.
Look at that list again. Not one item is an audit requirement for its own sake. Every one of them protects a product, a decision, or a customer relationship. That is what the industry undervalues when it treats the uncertainty estimate as a certificate formality.
Putting It Into Practice
The fundamentals of uncertainty of measurement come down to a short chain of ideas. Every result carries doubt. The uncertainty value sizes that doubt, and the uncertainty budget is the tool that calculates it defensibly. Simple acceptance let the industry report that value without using it. Accreditation bodies are now closing that gap through decision rule scrutiny and TUR expectations. The labs that get ahead of this shift will do three things. They will keep living budgets instead of filed ones. They will take repeatability data at the time of measurement whenever the method allows it. And they will report uncertainty values that describe the actual job. Their customers, in turn, finally get a number they can use for TUR checks, comparisons, drift monitoring, and risk decisions.
You do not need to overhaul your quality system to start. Pull one budget out of the folder this week and run it against a real customer job. Then compare the result to the CMC you have been reporting. The gap you find, or do not find, will tell you exactly where your lab stands.
Stop guessing what your assessor will expect.
Precision ISO builds uncertainty budgets, decision rules, and the SOPs that connect them, across a wide range of measurement disciplines.
A note on accuracy: This article teaches; it does not give legal, regulatory, or accreditation advice, and it guarantees no audit outcome. Clause numbers reference ISO/IEC 17025:2017, and the uncertainty framework follows the GUM (JCGM 100:2008). Accreditation body policies on decision rules and TUR vary by body and change over time, so verify your body’s current published requirements and your organization’s controlled copy of the standard before relying on any summary here.
