Glossary term

Repeatability

Engineering definition of repeatability covering repeated readings, short-term precision, standard deviation, Type A uncertainty, repeatability limit and validation evidence.

Branch: Engineering Physics
Glossary type: metric
Content: Glossary term
Updated: Jun 26, 2026
Revision: v1.0.0 · reviewed

Definition

metric

Repeatability is the closeness of agreement among repeated measurements made under the same conditions over a short time.

In engineering measurement, repeatability describes short-term scatter when the same operator, instrument, method, item, environment and setup are held as constant as practical. It is a precision metric, not proof of accuracy. Repeatability supports Type A uncertainty, calibration evidence, Gage R&R studies, validation checks and release decisions.

Repeatability is the closeness of agreement among repeated measurements made under the same conditions over a short time. It describes short-term scatter when the operator, instrument, method, item, fixture and environment are held as constant as practical.

Repeatability is not the same as accuracy. A measurement can repeat tightly and still be biased. It is also not the same as resolution: a display with many digits can still produce poor repeatability if noise, friction, seating, temperature or operator method dominates.

Engineering Meaning

For repeated readings (x_i), the mean is:

\displaystyle \bar{x}=\frac{1}{n}\sum_{i=1}^{n}x_i

The repeatability standard deviation is commonly estimated as:

\displaystyle s_r=\sqrt{\frac{\sum_{i=1}^{n}(x_i-\bar{x})^2}{n-1}}

where (s_r) has the same unit as the measured quantity. Smaller (s_r) means tighter repeatability under the tested conditions.

Repeatability Conditions

Repeatability only has meaning when the conditions are stated. The record should define whether the item was removed and reinstalled, whether the operator reset the fixture, whether zeroing was repeated, how long the system settled, what filter or averaging was used and whether temperature was held constant.

For some instruments, repeated readings without disturbing the setup only test electronic noise. Removing and reinstalling the item may reveal seating, alignment, contact force, backlash, surface condition or cable effects. Both tests can be useful, but they answer different engineering questions.

Worked Readings

Five repeated pressure readings are:

10.01,\ 10.03,\ 9.99,\ 10.02,\ 10.00\ \text{bar}

The mean is:

\bar{x}=10.01\ \text{bar}

The sample repeatability standard deviation is:

s_r=0.0158\ \text{bar}

The range is:

R=x_{max}-x_{min}=10.03-9.99=0.04\ \text{bar}

The range is easy to understand, but (s_r) is more useful for uncertainty and statistical decisions.

Type A Uncertainty

If the reported value is the average of (n) repeated readings, the standard uncertainty of the mean is:

\displaystyle u_A=\frac{s_r}{\sqrt{n}}

For (s_r=0.0158\ \text{bar}) and (n=5):

\displaystyle u_A=\frac{0.0158}{\sqrt{5}}=0.0071\ \text{bar}

Averaging reduces random scatter in the mean, but it does not remove bias, calibration error, drift, hysteresis or installation error.

Repeatability Limit

For two results obtained under repeatability conditions, an approximate 95 percent repeatability limit is often screened as:

r\approx2.8s_r

For the example:

r\approx2.8(0.0158)=0.044\ \text{bar}

Two repeat readings that differ by much more than this value should trigger review of setup, noise, seating, friction, thermal state or data rejection rules.

Relation to Gage R and R

Repeatability is the within-condition part of measurement-system variation. Gage R&R adds reproducibility effects such as operator, fixture, part presentation or method differences. A repeatability test may look good while the full measurement system is still weak because different operators or setups disagree.

That boundary matters in production release. A good bench repeatability result does not prove that the shop-floor, field or clinical workflow is repeatable under real use.

Release Use

Repeatability becomes important when a decision is close to a limit. If the repeatability standard deviation is large compared with the acceptance margin, one pass reading may be weak evidence. The release rule may require averaging, a guard band, a better fixture, more stable temperature, a different sensor range or a full measurement-system study.

Repeatability should also be compared with the process or physical variation being interpreted. If measurement scatter is larger than the change being monitored, trends, residuals and capability indices can be dominated by the measuring process rather than the system under review.

Evidence for Validation

Useful repeatability evidence states the number of readings, time spacing, operator, instrument, fixture, environmental condition, item condition, data filtering, rejected readings, warm-up state and whether the item was removed and reinstalled between readings.

For sensors, repeatability should be checked at relevant points in the range, not only at zero. For nonlinear, hysteretic or temperature-sensitive devices, repeatability should be tested in the direction and condition that matches the engineering decision.

Limits and Common Mistakes

Common mistakes include reporting only one reading, using repeatability as proof of accuracy, averaging away a biased measurement, hiding unstable setup conditions, mixing repeatability and reproducibility, and quoting standard deviation without sample size.

Another mistake is testing repeatability under ideal conditions while the release decision depends on installed behavior. A strong repeatability statement says what was repeated, what was held constant, what was allowed to vary and how the resulting scatter enters the uncertainty budget or release rule.

REF

Disciplines