Precision comes in two flavours that people constantly blur together, and the difference is entirely about what you hold still and what you let change. Repeatability is how tightly a measurement agrees with itself when everything stays the same; reproducibility is how tightly it agrees when the conditions deliberately change, such as a different operator, a different day, or a different setup. This guide draws the line cleanly between the two, uses proving runs and gauge repeatability and reproducibility studies to make it concrete, explains why a meter can be repeatable yet not reproducible, and why fiscal SCADA measurement has to care about both.
Repeatability vs Reproducibility in one line: Repeatability is the closeness of repeated measurements made under the same conditions, the same instrument, operator, and setup, over a short interval. Reproducibility is the closeness of measurements when conditions are deliberately changed, such as a different operator, day, or location. They are the two precision extremes: repeatability is best-case agreement with nothing changing, and reproducibility exposes the extra scatter that creeps in as conditions vary.
The whole distinction turns on one question: what did you keep constant between measurements? Repeatability describes precision under repeatability conditions, meaning the same measuring instrument, the same operator, the same method, the same location, and the same object, all within a short enough interval that nothing has drifted. It is the tightest agreement the measurement can achieve, because every source of variation you could hold still is held still. Repeatability is best-case scatter, the noise that remains even when you change nothing on purpose.
Reproducibility describes precision under reproducibility conditions, where you deliberately change one or more of those factors: a different operator takes the reading, or the same measurement is repeated on another day, in another place, or with a different unit of the same instrument model. Because more sources of variation are now free to move, the scatter is generally wider than repeatability. Reproducibility asks a harder and more realistic question: will this measurement give the same answer when the world around it is not held artificially constant?
The two are best understood as the extremes of a spectrum of precision conditions. Repeatability holds everything still and reports the irreducible core of scatter; reproducibility relaxes the conditions and reports how much extra variation those changes add. Real operations live between these poles, and knowing both numbers tells you not just how noisy a measurement is at its best, but how much worse it gets when the same measurement is made by someone else, somewhere else, or at another time, which is what actually happens in the field.
Two familiar practices make this abstract distinction tangible. In flow metering, a proving run checks a meter against a reference by taking several consecutive passes under the same conditions and asking whether they agree tightly enough; that tight agreement over back-to-back passes is a repeatability check, because the operator, the meter, the fluid, and the setup are all held constant across the short run. It is deliberately a same-conditions test, isolating whether the meter gives a stable answer when nothing is changed.
In manufacturing quality, a gauge repeatability and reproducibility study, usually shortened to gauge R and R, formally separates the two components. It has several operators each measure the same parts several times with the same gauge, then partitions the total variation into a repeatability part, the scatter when one operator repeats a measurement, and a reproducibility part, the extra scatter attributable to differences between operators. The output tells you how much of your measurement noise is the gauge itself repeating imperfectly versus how much comes from who is doing the measuring.
The value of a gauge R and R framing is that it stops people from lumping all imprecision together. A measurement process can have excellent repeatability but poor reproducibility, meaning each operator is individually consistent but they disagree with one another, which points to method, training, or setup differences rather than to the instrument. Or it can have modest repeatability that reproducibility barely worsens, meaning the instrument is the limiting factor and operators add little. Diagnosing which is which is what tells you where to spend effort to improve the measurement.
A meter can be repeatable yet not reproducible, and that combination is exactly the trap fiscal measurement has to guard against. Consecutive proving passes might agree beautifully, giving a confident repeatability result, while the same meter reads noticeably differently after a different technician re-installs it, on a colder day, or with the fluid conditions shifted. The good repeatability makes the meter look trustworthy in the moment, but the poor reproducibility means the number you can bill on this week is not the number you would get next week under changed conditions, and in custody transfer that gap is money.
This is why fiscal metering standards and practice attend to both extremes rather than one. Repeatability during proving confirms the meter is stable enough to prove against a reference at all; reproducibility across provings, seasons, operators, and re-installations confirms that the meter's calibration factor holds up in the real, changing conditions under which product is actually measured and money actually changes hands. A meter that only satisfied repeatability would give reliably wrong answers under any conditions other than the ones it was last proved in, which is not good enough when the reading determines payment.
For a cloud SCADA platform such as Merobix, aggregating fiscal and allocation measurements across many wellpads, meter runs, and delivery points, both properties surface in the data over time. Tight repeatability shows up as clean, low-scatter readings within a proving or a steady period; reproducibility shows up in whether a meter's behaviour holds consistent across provings, personnel, and seasons as those records accumulate. Having the long trend of proving results and as-found data in one place lets an operator see not just that a meter repeats well today, but whether it reproduces its calibration reliably over the changing conditions of real operation, which is the property that actually protects the accuracy of the numbers the business is settling on.
Repeatability holds everything the same, the same instrument, operator, method, and setup, over a short interval, and measures how tightly the results agree. Reproducibility deliberately changes something, typically the operator, the day, the location, or the specific unit, and measures how tightly the results still agree. Repeatability is best-case, nothing-changed precision; reproducibility is real-world, something-changed precision, which is usually wider.
Yes, and it is a common and dangerous case. A meter can give beautifully consistent readings across back-to-back passes under identical conditions, showing excellent repeatability, yet read differently after a different technician re-installs it, on another day, or under shifted process conditions. That poor reproducibility means the number is only trustworthy under the exact conditions it was last checked in, which matters greatly for fiscal measurement.
A gauge repeatability and reproducibility study has several operators each measure the same parts multiple times with the same gauge, then splits the total variation into a repeatability component, the scatter when one operator repeats a measurement, and a reproducibility component, the extra scatter caused by differences between operators. This tells you how much of your measurement noise is the instrument repeating imperfectly versus how much comes from who is doing the measuring.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.