Guard Band Visualizer
Your tolerance zone isn't as wide as the drawing says — your gauge took a bite from both ends. Drag the uncertainty around, move a part along the zone, and see the % chance it actually gets accepted.
Tolerance vs. instrument
g = 1 shrinks acceptance by the full uncertainty at each limit (ISO 14253-1-style conformance). g = 0 is “simple acceptance” — the drawing limits, risk and all.
4.0:1
TUR
meets the 4:1 rule
25%
of tolerance consumed
by measurement (2U / width)
30.0
acceptance width, µm
what you can actually use
Where your tolerance goes
drag the marker — or click anywhereAcceptance odds at this true value
88.5%
A part sitting exactly at +0.0120 mm gets accepted 88.5% of the time with this instrument and decision rule — the measurement noise (σ = U/2 = 2.5 µm) decides the rest.
The drawing gives, the gauge takes away
A reading of exactly the upper spec limit doesn't tell you the part is at the limit — it tells you the part is somewherewithin ±U of it, and half of that somewhere is out of spec. Every quality decision has to survive that blur. Guard banding is just being honest about it: pull the accept line inward until “measured good” reliably means “is good.” The interactive above makes the price visible — amber tolerance you can't use — and the money moment is turning the U slider up until the green vanishes: at that point the spec is unverifiable with that instrument, and no operator diligence can fix it.
I walked through the metrology behind this — where U comes from, why the CMM certificate isn't your measurement uncertainty, and what GD&T call-outs do to all of it — in GD&T vs reality: measurement uncertainty. And if the tolerance itself is the problem, remember finish and geometry set the floor on what you can hold — the surface finish calculator shows one of those floors live.
Frequently asked questions
What is guard banding?+
Shrinking the acceptance zone inside the drawing's tolerance limits by (some fraction of) your measurement uncertainty. If a feature is toleranced ±0.02 mm and your gauge is uncertain by ±0.005 mm, a reading right at the limit could belong to a good part or a bad one — you can't tell. Guard banding moves the accept/reject line inward by g·U so that a part you accept is very probably in spec. The cost is visible in the tool above: the amber bands are tolerance you paid for but no longer use.
What is TUR and where does the 4:1 rule come from?+
Test Uncertainty Ratio — the tolerance width divided by the total measurement uncertainty span (2U here). At 4:1, measurement consumes a quarter of the tolerance, which decades of gauging practice (and standards like ANSI/NCSL Z540.3) treat as the point where measurement risk stays manageable without heroic guard bands. Below 4:1, the amber bands grow fast; at 1:1 the acceptance zone is gone entirely. The old 10:1 'gagemaker's rule' is the comfortable version nobody's budget allows anymore.
How does ISO 14253-1 differ from simple acceptance?+
Simple acceptance (g = 0) uses the drawing limits as the decision rule: measure inside the spec, ship it — and eat the risk that measurement error waved a bad part through. ISO 14253-1's default conformance rule is the opposite: to *prove* conformance you shrink the acceptance zone by the full expanded uncertainty (g = 1). Slide g between 0 and 1 in the tool and you're sliding between those two worlds; many real decision rules sit in between, with g negotiated into the contract.
Who pays for the guard band?+
The producer, in the form of scrapped or reworked parts that were actually fine (producer's risk) — unless there's no guard band, in which case the customer pays through escaped nonconforming parts (consumer's risk). Guard banding doesn't destroy tolerance; it converts consumer risk into producer risk, openly. That's why the honest move is to guard-band deliberately and tell the customer, rather than silently tightening the process — and why the vendor with the better gauge can quote the same drawing cheaper.
How do I shrink U so I get my tolerance back?+
In rough order of return on investment: control temperature (20 °C is part of the definition of the millimetre in practice), fixture the part and standardize the method, average repeated measurements (uncertainty from repeatability drops with √n), calibrate against a better reference, and only then buy a better instrument. Run your gauge study, put the number in the U slider above, and the tool will tell you whether you're buying capability or just decimal places.
Fighting a tolerance your gauge can't verify?
Metrology, process capability, and honest decision rules — that's my day job.