A number in a SCADA system can be perfectly precise and completely wrong - a stale value from a device that stopped answering an hour ago, or a reading from a failed sensor. A data quality flag is the piece of metadata that warns you which is which. It travels alongside the value and says whether you can trust it. Without it, bad data silently corrupts trends, alarms, and reports. This guide explains what a data quality flag is, the common states, and why it is essential.
Data Quality Flag in one line: A data quality flag is metadata attached to a value that indicates how trustworthy that value is - typically classifying it as good, bad, uncertain, or stale. It lets people and systems tell a valid current reading from one that is out of date, from a failed sensor, or otherwise suspect, so bad data can be excluded rather than acted on.
Quality flags usually distinguish a handful of conditions. Good means the value is current and the source is healthy - trust it. Bad means the value cannot be trusted, typically because communication to the device has failed or a sensor is faulty; the last number may still be showing, but it is no longer valid. Uncertain sits between: the value is available but its reliability is questionable, perhaps because it is out of a sensor's rated range or being manually overridden. Stale flags a value that has not updated within its expected interval, so it may be frozen even if the last reading looked fine.
The OPC standards (OPC Classic and OPC UA) formalize this with quality codes carried on every value, which is why quality has become a first-class attribute of industrial data rather than an afterthought. A value without a quality flag is really an incomplete value - you have the number but not the crucial context of whether to believe it.
The danger of unflagged bad data is that it looks exactly like good data - a frozen reading still charts a flat line, and a failed sensor still reports a number. Treated as valid, that data poisons everything downstream: trends show a phantom steady value, KPIs are computed from stale inputs, alarms fail to fire because the frozen value never crosses a limit, and baselines learn the wrong normal. The error is silent, which makes it especially dangerous.
Quality flags make the untrustworthiness explicit and actionable. Analytics can exclude bad and stale values from calculations, trend displays can gray out or gap suspect data instead of drawing it as fact, and alarms can be suppressed or specially handled when their input is bad rather than firing false or missing real events. Good data-quality handling is what separates a report you can act on from one that quietly misleads.
Quality is assessed as data is collected and carried with the value through the system. When a device stops responding, the host marks its tags bad; when a value has not refreshed within its poll interval, it is flagged stale; and manual overrides or out-of-range conditions raise uncertain. This flag then follows the value into the historian, the trend, the alarm engine, and the report.
A cloud SCADA like Merobix reads tags from PLCs, RTUs, and flow computers over Modbus, DNP3, OPC UA, and other protocols, and assigns and historizes a quality state alongside each value - marking data bad when a device goes unreachable and stale when it stops updating. Preserving quality through to the dashboard and history is part of delivering contextualized data: the platform tells you not just the number but whether to trust it, so analytics and operators can act on good data and set the rest aside.
It is metadata attached to a value indicating how trustworthy it is - commonly good, bad, uncertain, or stale. It lets people and systems distinguish a valid current reading from one that is out of date, from a failed sensor, or otherwise suspect, so bad data can be excluded instead of acted on as if it were real.
Bad means the value cannot be trusted, usually because communication to the device failed or a sensor is faulty - the displayed number is no longer valid. Stale means the value has not updated within its expected interval, so it may be frozen even if it looks reasonable. Both warn that the reading should not be treated as current truth.
Because bad data looks identical to good data - a frozen reading charts a flat line and a failed sensor still reports a number. Without quality flags, that data silently corrupts trends, KPIs, alarms, and baselines. Flags make the untrustworthiness explicit so analytics can exclude bad and stale values rather than compute on them.
This page references the protocol specifications published by the organizations below. Editions, product capabilities, and documentation change over time - confirm current requirements and specifications directly with the source.
Last reviewed: July 27, 2026. Merobix is not affiliated with, endorsed by, or sponsored by these organizations; their names are used only to identify the standards and products discussed.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.