Precision in science describes the closeness of agreement among repeated measured quantity values obtained under specified conditions. It concerns the dispersion or consistency of measurements rather than their agreement with an accepted reference value. A measurement process can therefore be highly precise while systematically inaccurate. Precision is associated with random variation, repeatability and reproducibility, although these terms have distinct technical meanings. In practical measurement, a precise instrument or procedure produces results that cluster closely when the measurement is repeated.
Precision can be quantified using statistical measures such as standard deviation, variance or related dispersion measures, depending on the measurement design. It is important because low random variability allows scientists to distinguish small changes from background measurement noise, but precision alone cannot establish that the measurements are correct.
Repeatability is one important context for precision. Repeatability concerns agreement between successive measurements under conditions that are kept as similar as practical, such as the same method, instrument, operator, location and short time interval. Reproducibility can concern agreement under changed conditions, potentially involving different laboratories, operators, instruments or times, depending on the terminology adopted by the relevant field or standard. Precision can therefore be considered at several levels.
A laboratory may achieve excellent within instrument repeatability while obtaining systematically different results from another laboratory because of calibration differences. Comparing precision across laboratories requires clear definitions of the conditions being changed. Without specifying those conditions, statements such as "the method is precise" can be ambiguous.
Precision is distinct from accuracy. Suppose an instrument repeatedly reports 10.20 when a reference value is 10.00. The measurements are tightly clustered and therefore precise, but their systematic displacement indicates poor accuracy relative to the reference. Conversely, measurements of 9.7, 10.2, 10.0 and 10.1 may average close to the reference but have greater spread and therefore lower precision.
In scientific work, both properties matter because an accurate but highly variable measurement can be difficult to use for detecting small effects, while a precise but biased measurement can consistently support an incorrect conclusion. Calibration primarily addresses systematic measurement behaviour, whereas repetition and control of random influences are important for improving precision. Treating precision and accuracy as interchangeable is therefore a fundamental measurement misconception.
Sources of limited precision include electronic noise, environmental fluctuations, finite instrument resolution, sample heterogeneity, operator variability and intrinsic variability in the system being measured. Some systems are inherently stochastic, meaning that repeated observations naturally differ even when the experimental procedure is perfectly controlled. Biological measurements often contain substantial variation because organisms differ genetically and physiologically. Chemical measurements may vary because of sampling and preparation. Physical sensors may vary because of thermal or electronic noise. Precision can be improved by stabilising relevant conditions, improving instrument design, standardising procedures, increasing measurement resolution where appropriate and increasing the number of independent observations. However, improving instrument resolution does not necessarily improve total precision if another source of variation dominates.
Statistical measures allow precision to be characterised quantitatively. The standard deviation describes the spread of observations around their mean for a particular dataset. The coefficient of variation expresses standard deviation relative to the mean and can be useful when comparing variability across quantities with different scales, although it is not appropriate for every measurement situation. The standard error describes uncertainty in an estimated mean under a specified statistical model and generally decreases as the number of independent observations increases.
These measures should not be confused. A small standard deviation indicates that individual observations are closely grouped, while a small standard error indicates that the estimated mean is relatively well determined. Reporting the appropriate statistic is essential because the word precision alone does not specify the statistical property being described.
Precision has practical importance in scientific instruments and experimental procedures. Analytical chemistry methods require sufficiently precise concentration measurements to distinguish small differences between samples. Engineering measurements require consistent dimensions to verify manufacturing processes. Sensor networks require stable measurements to identify genuine environmental changes rather than instrument noise. In clinical laboratory testing, precision contributes to the reliability of repeated assays. In physics, precision measurements can detect small deviations from theoretical predictions when systematic effects are adequately controlled.
High precision can also reveal small changes over time, although interpreting such changes still requires accurate calibration and appropriate statistical analysis. Precision is therefore valuable because it improves the resolution with which scientific differences can be observed, but it does not independently determine whether those observations correspond to the correct physical quantity.
Precision can be assessed through repeated measurements, replicate samples, control experiments and interlaboratory comparisons. The experimental design must distinguish technical repeatability from genuine biological or environmental variation. For example, repeatedly measuring the same prepared sample tests instrument and procedural precision, while measuring independently collected samples also includes sampling variability. Both may be important, but they answer different questions. Quality control programmes often monitor repeated reference samples to identify changes in precision over time. Statistical control charts can reveal increases in variability before a measurement process becomes obviously unreliable. Precision studies can also identify whether a method's variability changes with concentration, temperature or another operating condition. Consequently, precision is not necessarily constant across the entire measurement range.
A common misconception is that displaying more decimal places means that a measurement is more precise. Decimal places describe numerical representation, not necessarily physical resolution or repeatability. Reporting 12.345678 when the measurement uncertainty is several units creates an unjustified appearance of precision. Significant figures should reflect the information actually supported by the measurement process. Another misconception is that averaging measurements always creates more precise information.
Averaging independent observations can improve the precision of an estimated mean, but correlated errors or systematic biases do not disappear simply through averaging. Precision should therefore be evaluated in relation to the measurement model, sampling design and intended inference rather than inferred from the number of digits shown.
