PharmaCalc

Gauge Repeatability & Reproducibility (GR&R) quantifies how much of the variation you observe comes from the measurement system rather than the product. It is a prerequisite for process-capability work: if the measurement system consumes a large share of the total variation, capability indices and validation conclusions built on that data are unreliable.

Components: Repeatability (EV) — variation when the same appraiser re-measures the same item on the same equipment; Reproducibility (AV) — variation between appraisers; Part variation (PV) — real product variation; %GRR — the measurement system's share of total study variation. A crossed study (every appraiser measures every part, multiple times) separates these.

Common pharma applications: HPLC assay system qualification, tablet weight and hardness measurement, dissolution testing — anywhere a release decision rides on a number.

This calculator uses the ANOVA method: a two-way crossed analysis of variance with parts and appraisers as factors. ANOVA is preferred over the range-based (average-and-range) method because it also estimates the part-by-appraiser interaction and uses all the information in the data rather than only ranges. If the interaction term is not significant, it is pooled into repeatability, and variance components are extracted for each source.

Acceptance guidelines (established by AIAG, as reproduced in the ASQ CSSBB Handbook): %GRR < 10% — acceptable; 10–30% inclusive — marginal and should be examined, considering the application, the cost of the measuring device, the cost of repair, and other relevant factors; > 30% — inadequate. The number of distinct categories should be at least 5; fewer means the system should be reviewed for improvement. Outside the automotive industry these are guidelines, not requirements — your protocol should state the criteria it applies.

Scenario: Qualify an HPLC assay. Two analysts, 5 batches, 2 replicates each. The hand calculation below uses the classical average-and-range constants so every number can be followed on paper; the calculator itself runs the full ANOVA, which is the reportable analysis.

1. Collect ranges and averages.
   Study data (%LC)

   Analyst A ranges: 0.3, 0.2, 0.4, 0.3, 0.2 · Analyst B ranges: 0.2, 0.3, 0.3, 0.4, 0.2
   Overall R̄ = 0.28 · analyst averages 100.2% and 100.4% → X̄_diff = 0.2 · part range R_p = 1.2
2. Equipment variation.
   Calculation (K₁ = 0.8862 for 2 replicates)

   EV = 0.28 × 0.8862 = 0.2481%
3. Appraiser variation.
   Calculation (K₂ = 0.7071 for 2 appraisers; n×r = 10)

   AV² = (0.2 × 0.7071)² − 0.2481²/10 = 0.0200 − 0.00616 = 0.01384
   AV = 0.1177%
4. GRR, PV, TV.
   Calculation (K₃ = 0.4030 for 5 parts)

   GRR = √(0.2481² + 0.1177²) = 0.2746%
   PV = 1.2 × 0.4030 = 0.4836% · TV = √(0.2746² + 0.4836²) = 0.5561%
5. Assess.
   Verdict

   %GRR = 0.2746 / 0.5561 × 100 = 49.38% → INADEQUATE (> 30%)
   ndc = 1.41 × 0.4836 / 0.2746 = 2.48 → 2 — below the minimum of 5
6. Remediate — in the direction the numbers point. Here EV (0.248) is more than twice AV (0.118): repeatability — the instrument and method — dominates, not the analysts. Investigate the HPLC first (injection precision, integration parameters, column condition, sample prep variability); analyst training targets AV and would barely move this result. Repeat the study after the equipment work and do not proceed to capability analysis until the system is acceptable under your protocol's criteria.

• Crossed, balanced design: every appraiser measures every part the same number of times; randomize the run order and blind the appraisers to part identity where possible.
• Parts span the process: select parts covering the real range of process variation — a narrow part set deflates PV and inflates %GRR.
• Stable system during the study: no recalibration, maintenance, or method change mid-study.
• Guideline bands are conditional: the 10–30% band is "marginal — examine", not "acceptable"; your protocol should pre-state what is acceptable for this application, and percent-tolerance criteria may also apply when a specification exists.
• Variables data only: attribute (pass/fail) measurement systems need an attribute agreement study instead.

• ASQ Certified Six Sigma Black Belt Handbook, Ch. 16 (Measurement Systems) — GR&R methods, acceptance guidelines (adapted from AIAG), ANOVA method, ndc.
• AIAG, Measurement Systems Analysis reference manual — original source of the guideline bands and study designs.
• FDA, Process Validation: General Principles and Practices (2011) — adequacy of measurement systems for validation data.