A thickness gauge used for MSA Type 1/2/3 work has to be named by technology, range, resolution, and probe geometry on the RFQ line, otherwise the supplier cannot quote a comparable instrument and the resulting study will be a comparison of the wrong things [S1][S3].
For a usable Gauge R&R result, the buyer has to lock down the design (10 parts x 3 operators x 3 trials = 90 readings), the part-loading method (full unclamp between trials), and the acceptance thresholds (%GRR under 10% preferred, 10-30% conditional, ndc greater than or equal to 5) in writing before the gauge model is even discussed [S2][S3].
Scope of a thickness-gauge RFQ for MSA
The RFQ line must separate the gauge hardware from the study protocol, because a price for a thickness gauge without a study protocol is not a quote for a usable MSA; the supplier needs to see both before returning a number [S1].
Three technology families cover nearly every shop-floor use case and they have to be selected on the RFQ, not left to the supplier: magnetic-induction for ferrous substrates, eddy-current for non-ferrous substrates, and ultrasonic for non-conductive coatings over any substrate (paint, polymer, glass, FRP) [S3]. A coating thickness gauge is the specific term when the measured layer is a coating rather than a base material, and an ultrasonic thickness gauge is the right term when the substrate is non-conductive or the layer to be measured is a non-metallic coating [S1]. Picking the wrong family invalidates the study before the first reading is taken.
Specifying the gauge hardware
For a coating thickness gauge used in a Gauge R&R, the RFQ line must carry: measuring range (e.g. 0-2000 micrometres or 0-500 mils), resolution (1 micrometre or 0.1 mil typical for paint/coating), probe type (integrated vs separate, with cable length if separate), substrate compatibility (Fe / NFe / both), minimum sample size and curvature correction, and stated accuracy as a percentage of reading or absolute value at calibration temperature [S1][S3].
For an ultrasonic thickness gauge the equivalent set is: transducer frequency (commonly 2.25, 5, or 10 MHz), element diameter (e.g. 6 mm or 10 mm), couplant requirement, minimum measurable thickness (typically 0.6-1.0 mm in steel at 5 MHz), range (e.g. 0.8-300 mm steel), resolution (0.01 mm or 0.1 mm), and velocity calibration range covering the actual material being measured [S3]. Buyers who leave velocity calibration and transducer frequency to the supplier routinely receive a gauge that resolves to 0.1 mm when the tolerance is 0.05 mm, and the study fails on instrument resolution rather than operator method.
Specifying the study design
The AIAG MSA reference design for a crossed Gauge R&R is 10 parts x 3 appraisers x 3 trials = 90 readings, with parts selected to span the real process spread (including units near both tolerance limits), appraisers who actually run the gauge in production, and a randomised, blinded part order so an appraiser cannot reproduce a remembered reading [S2][S3].
The single most common error that invalidates a fixture- or thickness-gauge study is leaving the part clamped between trials, which reduces the study to a measurement of the indicator rather than the system; the RFQ must therefore state "full unclamp and reload between every trial" as an explicit requirement, and the same language should appear in the control plan [S3]. A worked crossed study can be set up in mfgQC as a tidy frame of one row per measurement, with columns naming the part, the operator, and the trial, and a ValueError is raised if the design is unbalanced or has fewer than 2 trials per cell [S2].
Specifying the acceptance criteria
Two numbers come out of the study and both have to pass. %GRR expresses measurement-system variation as a percentage of the reference (process) spread, and the conventional breakpoints are: under 10% acceptable without caveat, 10-30% acceptable conditional on the application, and over 30% generally rejected [S3].
The number of distinct categories (ndc) is calculated as 1.41 x (part variation / gauge variation) and answers a different question: how many non-overlapping groups the system can actually discriminate. Industry convention sets the floor at ndc greater than or equal to 5 for a usable system; an ndc of 4 or less means the gauge cannot tell good from bad parts even if %GRR looks acceptable on paper [S3]. Both numbers should appear on the RFQ acceptance line, because asking only for "%GRR less than 10%" lets a borderline gauge pass on a metric the application does not care about.
Specifying the measurement procedure
The RFQ must pin the probe location and the fixture, because a magnetic characteristic on a control plan has to specify the method, the fixture, the probe location, and the operator's clamp sequence or the R&R result will not transfer to production [S1]. Vague wording like "measure coating thickness" is the second most reliable way to invalidate a study; specific wording such as "5 readings, 10 mm apart, along the centreline of the painted face, with the probe held perpendicular within plus or minus 5 degrees" is what the supplier needs [S1][S3].
Environment also belongs on the line: the study must be run on the shop floor at working temperature, not in a metrology lab at 20 plus or minus 1 degrees C, because a gauge that passes in the lab can still fail the line where the actual readings are taken [S3]. For ultrasonic work the velocity calibration material, the couplant, and the surface preparation (e.g. file and couplant, or grit blast and couplant) should all be on the RFQ; for coating work the substrate preparation, the calibration foil set, and the zero/foil calibration sequence must be on the RFQ.
Comparison of the three common technologies
On a decision matrix of substrate compatibility, layer type, accuracy class, and lead time, the three technologies line up as follows. Magnetic-induction: ferrous substrate only, metallic and non-metallic coatings on steel, typical accuracy plus or minus 1-3% of reading, short lead time and lowest cost; eddy-current: non-ferrous substrate only, non-conductive coatings on aluminium or copper, typical accuracy plus or minus 1-5% of reading, short lead time; ultrasonic: any substrate, any coating including non-conductive layers over non-conductive substrates, typical accuracy plus or minus 0.1 mm on metal substrate, longer lead time because the transducer has to be matched to the application [S1][S3].
For routine paint-on-steel work, a dual-mode coating thickness gauge covering Fe and NFe at 0-2000 micrometres with 1 micrometre resolution is the default. For paint-on-aluminium the eddy-current side of the same instrument is used. For paint-on-FRP or for rubber-lining thickness, the ultrasonic thickness gauge is the only technology of the three that works at all, and the transducer selection (typically 5 MHz, 10 mm element) is the dominant factor in the quote [S1][S3].
Common RFQ mistakes that force a requote
Three omissions cause the majority of requote cycles on thickness-gauge RFQs. First, the RFQ does not name the technology (magnetic / eddy-current / ultrasonic / dual), so the supplier cannot price a probe family. Second, the RFQ does not state the resolution required for the tolerance to be measured, so a 0.1 mm resolution gauge is quoted for a 0.05 mm tolerance and fails acceptance. Third, the RFQ does not state the part-loading method (full unclamp between trials) and the appraiser roster (the operators who actually run the gauge), so the study runs against the wrong population and the resulting %GRR cannot be defended [S2][S3].
A fourth, less obvious omission is calibration traceability: the RFQ should require the gauge to ship with a traceable calibration certificate (accredited to ISO/IEC 17025 by the issuing lab) and a set of calibration foils or a reference block matched to the measurement range; without this, the study cannot reference its readings to a national standard and the result will be rejected at the customer audit stage [S1].
When a thickness-gauge R&R is the wrong tool
A Gauge R&R is the right tool when the characteristic is a single quantitative dimension (thickness, diameter, length) and the process is in statistical control. It is the wrong tool when the characteristic is a magnetic field rather than a dimension: a gauge R&R on flux behaves nothing like one on a bore diameter, because fields are harder to measure repeatably, the probe coupling varies with air gap, and the part-to-part spread of magnets dominates the variance budget in a way that ordinary thickness studies do not see [S1].
For a thickness-gauge programme that is being launched in 2026, the second signal to watch is whether the supplier is quoting a dual-mode instrument or a single-technology one; the dual-mode quote is usually the lower total cost when the part mix is mixed-substrate, and the single-technology quote is the lower cost when the part mix is single-substrate. Pin the substrate on the RFQ and the quote follows.
Related analysis: Shield machine selection for tunneling: 2026 spec and method map.