
GUM Guide: How to Calculate Measurement Uncertainty for Calibration Technicians
What Is Measurement Uncertainty?
Every measurement has uncertainty — no matter how precise the instrument. Measurement uncertainty is a parameter that characterises the range of values that could reasonably be attributed to a measurement result.
ISO/IEC 17025 §7.6 explicitly requires calibration laboratories to report uncertainty alongside every measurement result. The international standard method for calculating uncertainty is the GUM (Guide to the Expression of Uncertainty in Measurement), published by the JCGM.
The GUM Framework — 4 Key Steps
Step 1: Define the Measurand and Measurement Model
Write a mathematical model relating the measurand (Y) to input quantities (X₁, X₂, ..., Xₙ): Y = f(X₁, X₂, ..., Xₙ)
For example, calibrating a temperature sensor against a reference thermometer:
ΔT = T_UUT − T_ref
Step 2: Identify and Evaluate Uncertainty Components (Type A and Type B)
- Type A: Evaluated by statistical analysis of repeated measurements — e.g., standard deviation of the mean: u_A = s / √n
- Type B: Evaluated by other means — calibration certificates, manufacturer data sheets, or scientific judgement. An assumed distribution (normal, rectangular, U-shaped) converts the stated value to a standard uncertainty.
Common Type B components in calibration:
- Reference standard uncertainty: u_ref = U_cert / k_cert
- Instrument resolution: u_res = resolution / (2√3) — rectangular distribution
- Temperature coefficient/drift
- Reproducibility from prior studies
Step 3: Combine Components into Combined Standard Uncertainty (u_c)
Assuming independent inputs, apply the law of propagation of uncertainty:
u_c(Y) = √[Σ(cᵢ · u(Xᵢ))²]
For linear models (sensitivity coefficients = ±1), this simplifies to:
u_c = √(u_A² + u_B1² + u_B2² + ...)
Step 4: Calculate Expanded Uncertainty (U)
U = k · u_c
For a normal distribution at ~95% confidence, use k = 2. Always state k and the coverage probability in the certificate.
Worked Example: Digital Thermometer at 100°C
- Repeatability (n=10): s = 0.05°C → u_A = 0.016°C
- Reference PRT (U_cert = 0.10°C, k=2): u_ref = 0.050°C
- UUT resolution 0.1°C: u_res = 0.029°C
- Drift ±0.02°C/yr: u_drift = 0.012°C
u_c = √(0.016² + 0.050² + 0.029² + 0.012²) ≈ 0.061°C → U = 0.12°C (k=2)
Common Mistakes
- Omitting reference standard uncertainty from the budget
- Reporting u_c instead of U — certificates must show expanded uncertainty with k
- Assuming k=2 without checking degrees of freedom — low ν_eff requires t-distribution
- No dominant component — if all components are similar in size, a major source has likely been omitted
Built-In Uncertainty Calculator in LabSync
LabSync includes a GUM Uncertainty Budget module built into every job item card. Define input components, choose distribution types, compute sensitivity coefficients, and get expanded uncertainty with k-factor — all stored against the job record with a complete audit trail. Try LabSync free for 14 days.


