Water meters do not fail loudly. A mechanical register that reads a few percent low can keep producing bills, totals, and monthly trends that look ordinary for years, and nobody notices the gap unless something checks it. That gap matters more as utilities add remote reading, because a remote feed makes the numbers arrive faster and more often without making them more accurate.
Where the error starts
Most residential meters are positive displacement or velocity types. In either design, the measuring element wears, debris changes the flow path, and a meter can lose sensitivity at low flow first. Low flows are common in homes: a slow faucet, a toilet refill, a small leak. A meter that under-registers at those rates may still track high-flow use well enough that the monthly total looks reasonable. The error sits at the lower end of the curve, where the customer uses many small draws, and a bill rarely shows where on that curve the meter is operating.
Sizing matters too. An oversized meter may never run near the range where it is most accurate, while an undersized meter can wear faster under sustained high flow. Neither condition is visible on a bill, and neither is obvious from the remote read alone.
What the remote feed does and does not fix
Remote reading gives a utility interval data, which is useful for leak detection, customer alerts, and seeing when a property's use changes shape. It does not calibrate the meter. A smooth, plausible series from a slow meter is still a smooth, plausible series. Analysts who treat the feed as a truth source can end up setting leak thresholds against a drifting baseline and then wondering why the alerts look odd.
The practical check is comparison. A district meter at the source, or a bulk meter feeding a zone, gives a total against which the sum of customer meters can be read. The gap between the two, often called non-revenue water when it represents a loss, is an aggregate signal. It does not say which meter is wrong. It says the system is not balancing, and that is the reason to test a sample.
Testing is a sample decision
Bench testing a removed meter across several flow rates is the most direct way to measure its accuracy at low flow. Field testing in place is possible with portable equipment but is harder to control, since site conditions, upstream fittings, and pressure all affect the result. Either route costs labor, service interruption, and sometimes spare stock, so the utility has to decide how many meters to test and from which install vintages.
Two strategies compete. Age-based replacement is simple to schedule and does not depend on test results, but it replaces some meters that would have passed. Test-based replacement targets meters that show drift, but it requires a sampling plan that can be defended to a rate board and a record that holds up when someone asks how the decision was made. A utility that tests a small number of meters from each install cohort each year gets a trend it can act on, though a small sample will not characterize the whole population, and the utility should say so in its own documents.
What to record before the feed is trusted
Log the meter model, install date, register type, size, and the test results with the flow rates used. Keep the remote read history separate from the test record so a later comparison does not mix a field reading with a bench reading. When a new remote platform is added, record whether it changes the reading method or only the transmission path, because those two changes carry different questions about accuracy.
Badger Meter announced on Nov. 3 that it had acquired 37South, described as a "utility technology and remote monitoring company" based in Melbourne, Australia, according to Water Finance and Management's report of the announcement. The summary of that report does not describe any change to meter hardware or measurement accuracy, so it offers no basis for changing how existing meters are tested. For a utility, the relevant questions are the same as before: which meters are being read, how they were checked, and who answers when the totals stop balancing. Those answers belong in the utility's own file, not in a vendor announcement.