Compare TAR’s reference accuracy with TUR’s uncertainty, then apply JCGM 106 and ISO/IEC 17025 decision rules using a practical workflow and worked example.
TUR, the Test Uncertainty Ratio, compares your tolerance against the laboratory’s measurement uncertainty, while TAR, the Test Accuracy Ratio, compares that same tolerance against a reference standard’s accuracy specification. For formal conformity decisions under ISO/IEC 17025 and JCGM 106, TUR is generally the more defensible metric because it accounts for the full uncertainty budget, not just one instrument’s rated accuracy. The calculation steps, the decision-rule interaction and a worked example follow below.
TL;DR:
- Calculate TUR using the laboratory’s uncertainty at the exact measurand point, and state whether it uses standard or expanded uncertainty and the coverage factor.
- Treat 4:1 as a practical benchmark, not an acceptance rule; a documented rule using simple acceptance or guard bands must determine outcomes near tolerance limits.
- In the example, TAR was 20:1 while TUR was 5:1; increasing expanded uncertainty from 0.4 to 0.6 units lowered TUR to 3.3:1.
- Check that the certificate’s CMC covers the actual test point, range, and parameters; interpolated values are estimates, and mismatched conditions may understate uncertainty.
- Before dividing, match units and interval types; a total tolerance band cannot be compared directly with an uncertainty stated for one side.
Getting these two ratios mixed up causes more downstream confusion in calibration programs than almost any other metrology concept. A NASA paper on measurement decision risk traces years of practical confusion back to loose or inconsistent definitions of TAR and TUR, which is exactly why we define both precisely before touching a calculator.
Test Accuracy Ratio (TAR) divides the tolerance or specification of the unit under test by the accuracy specification of the reference standard used to calibrate it. It tells you how much better the reference instrument is than the thing it’s checking, nothing more.
Test Uncertainty Ratio (TUR) divides that same tolerance by the laboratory’s measurement uncertainty, which includes the reference standard’s contribution plus environmental factors, repeatability, resolution and every other source in the uncertainty budget. This is the figure that reflects what the lab can actually demonstrate, and Australian competency guidance treats this tolerance-to-uncertainty comparison as the proper fit-for-purpose check, not the narrower accuracy comparison.
The distinction that trips people up most often:
Expanded uncertainty is the standard uncertainty multiplied by a coverage factor, commonly k=2, which approximates a 95% confidence level. A TUR built on standard uncertainty without stating that convention is not comparable to one built on expanded uncertainty at k=2, even if every other input is identical.
Both ratios are simple divisions once you have the right inputs in the right units. Most errors happen before the division, in how the inputs are gathered.
You need four pieces of information from the calibration certificate and the governing specification:
The formulas themselves are straightforward:
TAR = Tolerance ÷ Reference standard accuracy
TUR = Tolerance ÷ Laboratory measurement uncertainty
Both sides of each equation must be in the same units and represent the same type of interval. A tolerance expressed as a total band (for example, ±0.5 mm meaning a 1.0 mm band) must be matched against an uncertainty expressed the same way, not mixed with a one-sided figure.
Converting between standard and expanded uncertainty changes the ratio directly. If a certificate reports standard uncertainty and you need expanded uncertainty at roughly 95% coverage, multiply by the coverage factor, typically k=2. Dividing tolerance by the expanded (larger) figure produces a smaller, more conservative TUR than dividing by the standard uncertainty alone. Always check which convention the certificate uses before comparing it against a rule-of-thumb ratio.
Pro Tip: Write the uncertainty convention (standard or expanded, with k value) directly next to every TUR you calculate, so nobody downstream has to guess which one it is.
Common errors worth checking for on every job:
A ratio on its own doesn’t make a conformity decision. JCGM 106 describes decision rules, such as simple acceptance or guard-banded acceptance, as the mechanism that actually converts a measurement result and its uncertainty into an accept or reject outcome. TUR feeds into that decision rule; it doesn’t replace it.
The framework separates two kinds of risk:
Uncertainty widens the indeterminate zone near the tolerance limits, where neither risk can be driven to zero. A documented decision rule states how that zone is handled, whether through a straightforward acceptance interval or a guard band that shifts the acceptance limits inward to reduce consumer risk at the cost of some producer risk.
A TUR of 4:1 is often cited as a practical benchmark, but national competency guidance is explicit that a blanket 4:1 rule is not a substitute for proper uncertainty estimation and a documented decision rule. The ratio tells you how much headroom exists between tolerance and measurement capability; it doesn’t tell you what to do with results that fall near the edge.
This is why TUR plus a documented decision rule is defensible in a way that TAR alone never is. TAR reflects the reference standard’s rated accuracy, a static datasheet figure. TUR reflects what the laboratory’s full measurement process can actually resolve, which is the quantity that matters when JCGM 106’s risk-balancing logic is applied to a real accept or reject call. When false acceptance and false rejection carry different costs, guard-banding against an agreed acceptance interval is the mechanism to use, not a sharper ratio cutoff.
A workable TUR process runs the same six steps regardless of industry or measurand:
Here’s a worked example using illustrative figures. The reference standard used has an accuracy specification of 0.1 units, and the laboratory’s expanded measurement uncertainty (k=2, approximately 95% coverage) at that calibration point is 0.4 units.
The TAR of 20:1 looks comfortable, but it only describes the reference standard’s accuracy against the gauge’s tolerance. The TUR of 5:1 is the figure that reflects the laboratory’s actual demonstrated capability, including everything beyond the reference standard, and it’s the one that should inform the decision rule. Under a simple acceptance rule with no guard-banding, a 5:1 TUR comfortably supports a conformance decision for a reading well within tolerance.
Now shift the expanded uncertainty from 0.4 to 0.6 units, perhaps from a wider environmental allowance or a less stable reference. The TUR drops to 3.3:1. That’s still workable, but it narrows the margin for a reading close to the tolerance limit and may push a borderline result into the indeterminate zone that JCGM 106 describes, where a documented decision rule, not the ratio alone, determines the outcome.
For any accredited conformity decision, lean on TUR. It’s the ratio that accounts for the full measurement uncertainty budget and the one that ISO/IEC 17025 criteria expect to see supporting a conformance statement. TAR still earns its place as a fast, early screening check, useful when selecting a reference standard for a new job or doing a quick sanity check before a full uncertainty budget is built.
A few conservative exceptions are worth documenting rather than assuming:
Pro Tip: Every calibration record should show the uncertainty figure, its coverage convention, and the decision rule applied, not just the final pass or fail.
A certificate that states a pass or fail without the supporting uncertainty detail doesn’t give you enough to defend a conformity decision later. NATA’s calibration criteria require CMC uncertainty to be expressed as expanded uncertainty at roughly 95% coverage, alongside the applicable range and measurand parameters.
Run through this short checklist when auditing a certificate:
Our guide to checking lab calibration uncertainty walks through a fast method for running these checks against your own certificates.
Every uncertainty figure reported should follow the conventions set out in ISO/IEC 17025: expanded uncertainty, stated coverage factor, and the range or parameters the CMC applies to. Our explainer on NATA calibration standards covers the requirements in more depth.
When you’re assessing any calibration provider, ask for:
That level of detail is what lets you build a TUR you can defend, rather than a number you have to take on faith.
A TUR or TAR figure is only as useful as the decision rule sitting behind it. JCGM 106 frames measurement as a decision-support activity, not an end in itself, and that framing gets lost whenever a team chases a ratio without writing down what risk they’re actually willing to accept. The fix is unglamorous: state the acceptable risk level, the decision rule, and the uncertainty convention in use, then revisit all three periodically as equipment, ranges or tolerances change.
— Kaz
Working through uncertainty budgets and decision rules takes time most quality teams don’t have spare, which is where our NATA-accredited calibration services come in. We handle traceable calibration, on-site work, trade verification and hire equipment, and every certificate we issue reports uncertainty the way accreditation requires, ready to support your own TUR calculations without extra back-and-forth.
If you’re also weighing up whether an instrument’s tolerance is realistically achievable against your required TUR, our equipment range covers balances, scales, data loggers and measuring equipment with the specifications laid out up front. Book a calibration or ask about uncertainty reporting through our calibration service page and we’ll talk you through exactly what your certificate will show.
TUR, or Test Uncertainty Ratio, is the tolerance or specification limit of the item under test divided by the laboratory’s measurement uncertainty. It reflects the laboratory’s full measurement capability, including the reference standard, environment and process, which is why accreditation criteria treat it as the more reliable basis for a conformity decision than TAR alone.
In metrology, TUR describes how much smaller the laboratory’s measurement uncertainty is compared with the tolerance it’s checking against. A higher TUR means more confidence that a pass or fail result reflects the true state of the item, not measurement noise, though JCGM 106 is clear that the ratio still needs a documented decision rule to produce a defensible outcome.
Calibration is commonly grouped by where it happens and what it’s checked against: on-site calibration performed at the equipment’s location, workshop or laboratory calibration performed under controlled conditions, and trade verification for instruments used in legal-for-trade applications, with detailed guidance available on what is color calibration. Each type still relies on the same uncertainty and traceability principles described in ISO/IEC 17025.
TAR is calculated by dividing the tolerance or specification limit of the unit under test by the accuracy specification of the reference standard used to calibrate it. Both figures need to be in matching units and represent the same type of interval, whether one-sided or a total band, for the ratio to mean anything.
