Sometimes a system knows a reading has gone bad and decides, on its own, to serve a replacement so that the process keeps running smoothly. That replacement is a substituted value, and the honest way to deliver it is to hand it over with a quality marker that says, in effect, this number is fine to use but it is not raw field data. This guide explains when and why systems substitute values, what the Good Non-Specific and Substituted quality markers mean, why silent substitution is dangerous, and how governance keeps every substitution visible and auditable. It differs from forcing, which is an operator-driven override, because substitution is an automatic quality-handling mechanism.
Substituted Value in one line: A substituted value is a reading that a system has automatically replaced, using the last good value, a manual entry, or a calculated estimate, because the real source went bad, and then marked with a quality such as Good Non-Specific or a Substituted subcode so consumers know the number is usable but not raw field data. The marker exists so that substitution never happens silently, letting downstream systems and analysts see that the value was replaced rather than measured directly.
Substitution is an automatic response to a source that has failed. When a reading loses its communication link, a device faults, or an input is rejected by a validation check, the system faces a choice: pass along the bad value, pass along nothing, or fill the gap with a stand-in. In many designs the sensible choice for continuity is to substitute a plausible number so that displays, control loops, and calculations are not thrown by a hole in the data. The substitute might be the last good value the point held, a manually supplied figure, or a value calculated from related measurements.
The purpose of substituting is to keep the process observable and controllable through a transient fault without pretending the fault did not happen. A control loop that suddenly receives a bad or missing input can behave erratically, so holding the input at its last good value while the fault clears is often safer than letting the loop react to garbage. A display that shows a blank or a wild value alarms operators unnecessarily during a brief comms hiccup, whereas a held-over value keeps the picture stable. Substitution buys time and stability while the underlying issue is addressed.
But that stability comes with an obligation, because a substituted value is not a measurement. It is an estimate or a stand-in, and its accuracy degrades the longer the real source stays down: a last good value that was correct a minute ago may be badly wrong an hour later, and a calculated substitute inherits whatever assumptions went into it. The system's job is not just to substitute but to be honest about having done so, which is exactly where the quality marker comes in. Without a marker, a substitute is indistinguishable from a real reading, and that is where the danger lies.
The quality attached to a value is what tells a consumer how much to trust it, and substitution has its own vocabulary within that scheme. Under the OPC quality model that most SCADA systems follow, a value's quality has a major status of good, uncertain, or bad, and within those a substatus that adds detail. A substituted value is commonly delivered as good but with a substatus that flags it, and the term Good Non-Specific describes a good-quality value carrying no more specific reason, which some systems use for a value that is acceptable to use but not the ordinary live reading. Other systems use an explicit Substituted or last-usable-value substatus to name the substitution directly.
The reason the marker matters is that it changes what a correct consumer does with the number. A value marked good is safe to display and to use, so control and screens keep working, but the substatus tells an analytics engine, a historian query, or an operator that this particular good value was not measured directly and may not deserve the same weight as a raw reading. A totalizer might still include it, but a data-quality report should be able to distinguish periods of real measurement from periods of substitution, and a careful analysis might exclude substituted intervals or treat them with caution.
This is why substitution is fundamentally a quality-handling mechanism rather than just a value-handling one. The point is not merely to fill the numeric slot with something reasonable but to attach, alongside that number, an accurate statement of what the number is. A well-behaved system therefore always pairs the substitute value with the appropriate quality marker, so the two travel together through the historian, the displays, and any downstream system. The value keeps the process running; the marker keeps everyone honest about how that value was produced.
The failure mode to fear is silent substitution: a system that quietly replaces a bad reading with a stand-in and passes it along marked as an ordinary good value, with nothing to say it was substituted. When that happens, everyone downstream is deceived. An operator trusts a held-over value as a live measurement, an analyst averages substituted intervals in with real ones and never knows, and a report that looks like a record of measured data is partly a record of the system's own guesses. The data can look perfectly healthy while being partly fabricated, which is worse than an obvious gap because it hides the fault instead of showing it.
Good governance therefore insists that substitutions be visible and auditable. Every substitution should carry its quality marker through to the historian so that a later query can identify exactly which intervals were substituted, and reporting should be able to quantify the substitution rate for a tag or a dataset as a data-quality measure in its own right. Rules about when substitution is permitted, how long a last good value may be held before it should be considered stale, and which points may never be silently substituted at all belong in the system's design, not in undocumented behavior. The principle is simple: a system may substitute, but it must never hide that it did.
For a cloud SCADA platform such as Merobix, where remote sites feed a central hosted history that many people query, carrying the substituted marker all the way through is what keeps that shared history trustworthy. When a cellular link to a site drops and the platform holds a value over the gap, marking that interval as substituted rather than as a fresh measurement means a later analysis can see precisely where real data ends and a stand-in begins, and a data-quality dashboard can show how much of a tag's record over a period was substituted rather than measured. Because the substitution logic and its markers are applied centrally and consistently, the same honest handling covers every site, and no operator or report is ever left believing a held-over value was a live reading from the field.
A substituted value is produced automatically by the system as a quality-handling response to a source that went bad, filling the gap with a last good value, a manual figure, or a calculation. A forced value is an operator-driven override that deliberately replaces a live, working signal with a chosen number. Substitution reacts to a failure to keep the process running, while forcing is an intentional human intervention on a signal that is present and healthy.
Under the OPC quality model, a value's quality has a major status of good, uncertain, or bad plus a substatus that adds detail. Good Non-Specific describes a good-quality value carrying no more specific reason code. Some systems use it, or an explicit Substituted substatus, to deliver a value that is acceptable to use but is not the ordinary live field reading, so consumers know it is safe to display yet was not measured directly.
Because it replaces a bad reading with a stand-in and passes it along marked as an ordinary good value, deceiving everyone downstream. Operators trust a held-over value as live, analysts average substituted intervals in with real ones unknowingly, and a report that looks like measured data is partly the system's own guesses. It hides the fault instead of showing it, which is why governance requires substitutions to carry a marker and be auditable.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.