A meter proving is not accepted on a single run; it is accepted only when several consecutive runs agree closely enough with each other. That agreement test is the repeatability criterion, and it sits at the heart of every proving. This guide explains the consecutive-runs-within-a-spread rule, why repeatability is not the same as accuracy, why a run set gets rejected and rerun, and how SCADA flags out-of-band runs automatically.
Proving Repeatability Criteria in one line: Proving repeatability criteria are the acceptance rules that decide whether a proving is valid. A required number of consecutive runs - commonly five - must produce meter factors that agree within a specified maximum spread, such as 0.05 percent, before the average factor is accepted. If the runs are too scattered, the set is rejected and the proving is repeated.
During a proving, each run produces its own meter factor. The repeatability criterion looks at a set of consecutive runs and measures the spread between them - the difference between the highest and lowest factor in the set, expressed as a percentage. A common rule is five consecutive runs whose factors fall within a spread of 0.05 percent, though the exact run count and allowed spread come from the governing standard, the contract, and the prover type. Only when a qualifying set is achieved is the average of those runs adopted as the meter factor.
The word consecutive matters. The runs must agree back-to-back; an operator cannot cherry-pick the five closest runs out of twenty attempts. If a run drops outside the band, the set is broken and the count effectively restarts from a fresh sequence of runs that do hold together. This discipline is what gives the accepted factor confidence: it demonstrates the meter, the prover, and the process conditions were stable and reproducible during the proving, not that one lucky run happened to land on a number.
It is easy to conflate the two, but repeatability and accuracy answer different questions. Repeatability asks whether the runs agree with each other - whether the proving is reproducible. Accuracy asks whether the resulting factor is close to true, which depends on the prover's traceable calibration, correct corrections, and a properly functioning meter. A proving can be beautifully repeatable and still centered on the wrong value if, for example, the prover base volume is stale or a temperature correction is misapplied.
This is why repeatability is a necessary but not sufficient condition. Meeting the spread proves the measurement was stable enough to trust the average as a real signal rather than noise; it does not by itself prove the meter is correct. The accuracy side is protected separately, through the prover's own traceability and through comparing the as-found and as-left factors and watching for unexpected drift. A run set that repeats tightly but yields a factor far from the previous proving is a flag to investigate, not simply to accept.
A run set gets rejected whenever the consecutive runs fail to hold within the allowed spread. Common causes are unstable flow, entrained gas or air, temperature swings during the runs, a sticking prover detector, or a meter beginning to fail. The correct response is not to loosen the criterion but to stabilize conditions and rerun, because a proving accepted on a scattered set carries a factor that may be off. Persistent inability to meet repeatability is itself diagnostic - it points at a real problem with the meter, prover, or process.
A cloud SCADA platform automates the tedious, error-prone part: it reads each run's factor from the proving flow computer, computes the spread of the running consecutive set, and flags in real time when a run falls out of band and when a qualifying set has been achieved. Merobix does not perform the raw proving arithmetic - that stays in the certified flow computer - but it records every run, marks rejected sets, and preserves the accepted set with its spread, so the report shows exactly which runs were counted. That makes it obvious after the fact that the accepted factor rested on a genuinely repeatable set rather than a hand-picked one.
Requiring consecutive runs stops an operator from cherry-picking the closest results out of many attempts. Back-to-back agreement demonstrates the meter, prover, and process were genuinely stable during the proving, not that one sequence happened to line up. If any run falls outside the band, the set breaks and a fresh consecutive sequence must hold together.
Repeatability measures whether the runs agree with each other, so the proving is reproducible. Accuracy measures whether the resulting factor is close to true, which depends on the prover's traceable calibration and correct corrections. A proving can repeat tightly yet still be centered on the wrong value, so repeatability is necessary but not sufficient.
The run set is rejected and the proving is repeated after stabilizing conditions, rather than loosening the criterion. Scatter usually comes from unstable flow, entrained gas, temperature swings, or a failing meter or prover. Repeated inability to meet repeatability is itself a diagnostic signal that something needs to be fixed before a valid factor can be established.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.