Industry Study Design Guide

Industry Study Design Guide

Red flags, study design tables, and instrument guidance for eight industries.

Jump to: Medical device | Manufacturing | Aerospace | Automotive | NIST / Metrology | Healthcare | Energy | Defense

---

Medical device

Standards: 21 CFR Part 820.72 | ISO 13485:2016 | AIAG MSA 4th Ed.

Red flags

Flag	Why it matters
%GR&R > 30%	Cannot distinguish conforming from non-conforming product
ndc < 5	SPC and Cpk analysis is statistically invalid
AV > EV	Operator technique dominant — training/fixture usually resolves before gage replacement
EV >> AV	Gage hardware weak link — calibration or replacement needed
Study not in production environment	FDA investigators specifically check this
Operator ranges > UCL	Erratic technique or undisclosed part-handling damage

Study design

Consideration	Guidance
Study size	10 × 3 × 3 baseline; life-sustaining devices: 10 × 3 × 5
Part selection	Must span full production tolerance range
Randomisation	Re-code parts; operators must not see each other's readings
Re-study triggers	Equipment repair, process change, new operator, calibration expiry
DHF/DMR record	Study, raw data, and acceptance decision retained per 21 CFR 820

By instrument type

• Micrometers/calipers — tighten to <=10% for critical implant dims; 30-min thermal soak
• CMM — validate fixture repeatability separately; probe qualification interval matters
• Force/torque gauges — standardise loading rate and dwell; always approach from same direction
• Hardness testers — randomise indentation; never reuse prior indents
• Surface roughness — lock traversal direction and lambda_c; use fixture for re-positioning
• Electrical test equipment — include cable/fixture impedance in study; 4-wire Kelvin for R < 1 ohm
• Balances — draught shield mandatory below 0.1 g; anti-static ioniser for plastic/powder parts

---

General manufacturing

Standards: IATF 16949:2016 | AIAG APQP 3rd Ed. | ISO 9001:2015

Red flags

Flag	Why it matters
%GR&R > 10% for SPC dimensions	PPAP OEM portals auto-reject submissions above this
ndc < 5	SPC control charts produce false signals
PPAP MSA section incomplete	Missing GR&R is a Level 3 PPAP rejection trigger
SPC not pre-established	Unstable process inflates PV, produces false-acceptable GRR

Study design

Consideration	Guidance
Production gages	<=10% GRR for SPC/Cpk dimensions
Incoming inspection	<=20% acceptable maximum
PPAP Level 3	Full raw data + MSA summary required
ndc	>=5 AIAG; >=8 for SPC-critical parameters
Re-study frequency	Annual minimum; OEMs may specify more often

By instrument type

• CMM — version-control the measurement program; validate fixture independently
• Torque tools — standardise loading rate; approach from same direction always
• Vision systems — lock illumination and threshold settings before study
• Thread gauges — attribute agreement analysis (kappa), not variable GRR
• Surface plates — flatness certificate and temperature soak time documented

---

Aerospace

Standards: AS9100D | NADCAP AC7130 | SAE ARP9013

Red flags

Flag	Why it matters
%GR&R > 10% for flight-critical	AS9100D/NADCAP mandate <=10% — AIAG 30% is not acceptable
Missing NIST traceability	NADCAP audits require documented chain to NIST or equivalent NMI
Gage changed mid-study	Frozen process control: any change after study start invalidates it
No measurement uncertainty budget	ISO 17025 requires expanded U = k*u_c documented

Study design

Consideration	Guidance
GRR threshold	<=10% mandatory for flight-critical
Traceability	Full calibration chain to NIST retained with study
Uncertainty budget	Identify u_i sources; compute u_c; state U = 2*u_c at 95% confidence
Frozen process	No gage/probe/program changes between study start and sign-off
Study size	10 × 3 × 3 minimum; NADCAP NDT often requires 5 operators

By instrument type

• CMM — temperature-compensated; document part soak time; probe qualification per trial
• Laser tracker — atmospheric correction (temperature, pressure, humidity) active and logged
• Ultrasonic thickness gauge — couplant and probe pressure standardised; velocity calibrated at study temperature
• Optical profilometer — traversal direction locked to machining lay; lambda_c fixed

---

Automotive

Standards: AIAG MSA 4th Ed. | AIAG PPAP 4th Ed. | VDA Volume 5

Red flags

Flag	Why it matters
%GR&R > 10% for SPC dimensions	OEM portals (GM, Ford, Stellantis) auto-reject PPAP
ndc < 5 (AIAG) or < 8 (VDA)	State which standard governs
No linearity and bias for PPAP Level 3	Required in addition to crossed GRR for spanning gages
Fixture effect not quantified	Embeds in EV; cannot be separated post-hoc

Study design

Consideration	Guidance
GRR threshold	<=10% PPAP SPC; <=20% with customer concurrence
Linearity and bias	Required for PPAP Level 3
ndc	>=5 AIAG; >=8 VDA Volume 5 for SPC-critical
Study timing	After production tooling installed and capable
PPAP submission	Raw data + summary + range chart + X-bar chart

By instrument type

• Air gauge — stabilise supply air and part temperature; master rings with documented temperature coefficients
• CMM — production-environment CMM preferred over metrology lab for PPAP
• Profilometer — traversal speed, force, and lambda_c locked; parameter matches drawing callout
• Functional fixtures — fixture wear is primary EV source; re-qualification in control plan

---

NIST / National metrology

Standards: NIST Handbook 44 | VIM JCGM 200:2012 | ISO/IEC 17025:2017 | ILAC P14

ISO 17025 labs express uncertainty as expanded U = k*u_c (k=2), not as %GRR.

GRR to uncertainty budget mapping

GRR component	Uncertainty equivalent	Distribution
EV (Repeatability)	Type A, within-lab	Normal
AV (Reproducibility)	Type A, between-operator	Normal
Reference standard uncertainty	Type B	Normal or rectangular
Thermal expansion coefficient x delta_T	Type B	Rectangular
Resolution / least count	Type B	Rectangular (u = res / (2*sqrt(3)))

CMC rule: U_reference <= 25% of tolerance (ILAC P14).

Red flags

Flag	Why it matters
Expanded U not reported	ISO 17025 requires U at k=2 on all calibration certificates
k factor not stated	Cross-lab comparisons are invalid without coverage factor
Reference not in accreditation scope	Non-conformity at NADCAP or A2LA audit
Environmental conditions not logged	Temperature, humidity, pressure required for every measurement

By instrument type

• Gage blocks — wringing film (~0.01 um) is systematic EV; match CTE to workpiece
• Laser interferometer — Edlen refractive index correction required; vibration isolation mandatory
• Precision balance — ioniser for electrostatic below 1 mg; matched magnetic susceptibility of weights
• Dead-weight tester — piston-cylinder area with thermal expansion and air buoyancy correction

---

Healthcare / clinical laboratory

Standards: CLIA 42 CFR Part 493 | ISO 15189:2022 | CLSI EP05-A3 | CLSI EP15-A3

In clinical labs: operator = analyst, trial = run/replicate, tolerance = TEa.

GRR to clinical precision

AIAG	Clinical equivalent
EV (within-operator)	Within-run imprecision (CV_r)
AV (between-operator)	Between-analyst imprecision
(not in AIAG)	Between-run imprecision (CV_rr)
(not in AIAG)	Between-day imprecision (CV_d)

Sigma metric = (TEa - |bias|) / CV_total

Sigma	Verdict
>= 6	World-class
>= 4	Acceptable
< 3	Immediate corrective action

Red flags

Flag	Why it matters
CV > TEa/4	Imprecision alone consumes >25% of total error budget
Sigma < 3	Unreliable for patient decision-making regardless of QC
Bias not evaluated	CLIA and ISO 15189 require both precision and bias
Matrix not matched	Must use commutable patient-like material

By instrument type

• Pipette — gravimetric verification required; standardise tip lot and aspiration speed
• Clinical analyzer — document reagent lot; calibration before study; QC passes before each run
• Glucometer — strip lot is dominant EV; compare across multiple lots for CLIA waiver
• Spectrophotometer — verify wavelength accuracy and stray light; thermostat sample temperature

---

Energy / utilities

Standards: ISO 50001:2018 | IEC 61869 | ASME PTC 19.1 | OIML R 46

Meter accuracy classes

Class	MPE	Application
0.1	+/-0.1%	Reference metering
0.2	+/-0.2%	Revenue, transmission
0.5	+/-0.5%	Revenue, distribution
1.0	+/-1.0%	Industrial sub-metering
2.0	+/-2.0%	Residential metering

GRR equivalent: M&V protocol (IPMVP Option D). Target: total uncertainty <= 2% of measured energy.

Red flags

Flag	Why it matters
Instrument transformer ratio error excluded	CT/VT errors enter meter reading directly
Flow meter not calibrated at operating conditions	Viscosity and Reynolds number affect K-factor
No reference meter	Cannot detect systematic bias

By instrument type

• Revenue meter — sealed, tamper-evident; GPS/NTP time sync for interval metering
• Current transformer — test at 5%, 20%, 100%, 120% of rated current
• Flow meter — K-factor at operating viscosity; installation effects replicated
• Power analyzer — crest factor rating verified; harmonic analysis for spectral aliasing

---

Defense / government

Standards: MIL-STD-45662A | MIL-HDBK-1828 | ANSI/NCSL Z540.3

Test Uncertainty Ratio (TUR)

TUR = Tolerance / (2 * U_measurement)

TUR	Requirement
>= 4:1	Z540.3 default — no guard band required
3:1 to 4:1	Guard band required
< 3:1	Special approval + documented risk acceptance

TUR maps to %GRR: TUR = Tolerance / (5.15 * sigma_GRR)

Guard banding: acceptance limit moves inside tolerance by k_GB * U_measurement (k_GB = 1.0 default; k_GB = 2.0 for safety-critical parameters).

Red flags

Flag	Why it matters
TUR < 4:1 without guard banding	Direct Z540.3 violation
No uncertainty on calibration certificate	MIL-STD-45662A / Z540.3 violation
Classified hardware without access log	DCSA security violation
Calibration interval exceeded	Invalidates all measurements after due date

By instrument type

• Torque wrench (mil-spec) — per MIL-DTL-28778; CW and CCW calibrated separately
• CMM — per ASME B89.4.10348; volumetric verification with ball-bar or step gauge
• Pressure gauge — deadweight calibration; hysteresis tested in both directions
• Electrical multimeter — 4-wire Kelvin for R < 1 ohm; lead resistance subtracted