Repeatability is the closeness of agreement between results when you measure the same thing multiple times under identical conditions. Those conditions must stay fixed: same person, same instrument, same method, same location, and measurements taken within a short time frame. If a bathroom scale gives you 150.2, 150.3, and 150.1 pounds three times in a row, it has good repeatability. If it bounces between 148 and 153, it doesn’t.
The concept comes from metrology, the science of measurement, and it matters everywhere measurements matter: factories, hospitals, research labs, and quality control departments. Understanding repeatability helps you figure out how much of the variation in your results is real and how much is just noise from the measurement process itself.
The Conditions That Define Repeatability
Repeatability isn’t just “doing the same test twice.” It requires a specific, tightly controlled set of conditions. According to the International Vocabulary of Metrology and corresponding ISO standards, all of the following must remain constant:
- Same measurement procedure (identical method, step by step)
- Same operator (the same person performing the measurement)
- Same measuring instrument (used under the same settings)
- Same location (same lab, same bench, same environment)
- Short time period (measurements taken in quick succession)
The short time window is important because it minimizes the chance that external factors like temperature, humidity, or equipment drift will change between measurements. The goal is to isolate the variation that comes purely from the act of measuring, not from the world around it.
How Repeatability Differs From Reproducibility
These two terms are often confused, but they answer different questions. Repeatability asks: “If I do this again right now, under the exact same conditions, how close will my results be?” Reproducibility asks: “If someone else does this in a different lab with different equipment, will they get the same answer?”
A classic example from chemistry illustrates the distinction. A student performing five titrations back-to-back, using the same solutions and glassware, is measuring repeatability, or within-run precision. If instead those five titrations were performed by different staff on different days, in different laboratories, with different batches of chemicals, the resulting spread reflects reproducibility, or between-run precision. Reproducibility will almost always show more variation than repeatability because more things are allowed to change.
The Association for Computing Machinery summarizes it neatly: repeatability means the same team with the same setup gets consistent results, while reproducibility means a different team with a different setup can still reach the same conclusion.
How Repeatability Is Measured
Repeatability is quantified using the standard deviation of repeated measurements. You measure the same thing multiple times, then calculate how spread out the results are. A smaller standard deviation means better repeatability. In some fields, this is also expressed as a coefficient of variation, which puts the standard deviation in proportion to the average value, making it easier to compare across different scales.
For more complex analyses, a pooled repeatability standard deviation combines data from multiple groups of measurements, accounting for different operators, days, or runs. This gives a more robust picture of how much variation the measurement process itself introduces. Statistical software like Minitab is commonly used for these calculations, especially in industrial settings.
Bland-Altman plots offer another way to visualize repeatability. These scatter plots show the difference between two paired measurements on the vertical axis and their average on the horizontal axis. A well-repeatable measurement system will cluster its points tightly around zero difference. The general guideline is that 95% of data points should fall within two standard deviations of the mean difference. The plot also reveals whether measurement error gets worse at higher or lower values, which a single standard deviation number can’t show.
Repeatability in Manufacturing
In manufacturing, repeatability is formally assessed through a Gauge Repeatability and Reproducibility study, commonly called a Gauge R&R. This study breaks down the total variation in a measurement system into two components: repeatability (variation from the gauge itself, when the same operator measures the same part multiple times) and reproducibility (variation introduced when different operators use the same gauge).
The results are compared against the tolerance range for the part being measured, producing a precision-to-tolerance ratio. The automotive industry’s guidelines set clear thresholds: below 10%, the measurement system is acceptable; between 10% and 30%, it may be acceptable depending on the application and cost; above 30%, it’s considered unacceptable. A related metric, the number of distinct categories, tells you how many meaningfully different groups your measurement system can distinguish. If it can only sort parts into “high” and “low,” you’ve essentially lost all your measurement resolution. A system needs to resolve at least five distinct categories to be considered adequate.
Repeatability in Medical Testing
In clinical laboratories, repeatability directly affects whether patients receive correct diagnoses. When a blood test or imaging study is repeated on the same sample or patient, the results need to be close enough to support reliable decision-making. The stakes are highest for measurements that fall near diagnostic cutoff points, where even small variations can push a result from “normal” to “abnormal” and change the treatment plan.
Clinical testing differs from research in an important way: the unit of analysis is always a single patient, not a group average, and the consequences land on that individual. For this reason, clinicians sometimes repeat even highly accurate tests to guard against rare errors that have nothing to do with normal statistical variation, such as mislabeled samples or instrument malfunctions. Paradoxically, the most sensitive and specific tests (the ones that carry the most diagnostic weight) are often the ones most worth repeating, because an error on a high-stakes test has the largest impact.
Why Repeatability Matters in Science
Repeatability is foundational to the scientific method. If an experiment can’t produce consistent results under the same conditions, there’s no basis for trusting its conclusions. When independent researchers fail to reproduce findings, it raises questions about experimental design, data quality, statistical analysis, or the reliability of the original materials. In some cases, irreproducibility has led to allegations of data fabrication. The cold fusion episode of 1989, where Pons and Fleischmann announced results that no other lab could replicate, became a cautionary example of what happens when findings can’t withstand scrutiny.
The consequences for published science are concrete. An analysis of 423 retraction notices in PubMed that cited error (rather than misconduct) found that 16.1% of those papers were retracted specifically because their results could not be reproduced. Retraction exists as a formal mechanism to correct the scientific record when findings prove unreliable. Poor repeatability in individual measurements feeds directly into this larger problem, because if your instruments and methods aren’t producing consistent data in the first place, no amount of statistical analysis will produce trustworthy conclusions.
Common Sources of Poor Repeatability
When repeatability is worse than expected, the variation typically traces back to the instrument itself. Mechanical wear, sensor drift, electrical noise, or loose components can all introduce inconsistency between measurements. Environmental factors like temperature fluctuations or vibrations matter too, though these are partly controlled by the requirement that repeatability measurements happen in quick succession at the same location.
Operator technique also plays a role, even under “same operator” conditions. Slight differences in how a tool is positioned, how much pressure is applied, or how a reading is interpreted can introduce variability. This is why repeatability is sometimes described as the best-case scenario for measurement precision: everything that can be held constant is held constant, and whatever variation remains is the floor below which the system cannot perform.