You cannot say a reading is abnormal without knowing what normal looks like - and that reference for normal is the baseline. It is the quiet foundation under nearly every operational analytic: deviation detection, anomaly scoring, and KPI targets all measure against a baseline. Set it well and problems stand out clearly; set it badly and you either miss real issues or drown in false ones. This guide explains what a baseline is, the main types, and how one is established.
Baseline in one line: A baseline is a reference point representing normal or expected performance, against which current data is compared to judge whether behavior is on track or deviating. It is derived from historical data or design expectations and serves as the yardstick for deviation detection, anomaly analysis, and performance targets.
A static baseline is fixed - established once from a period of known-good operation, a commissioning test, or a design specification, and then held constant. It is appropriate when 'normal' genuinely should not change, such as a design efficiency or a regulatory reference. Its weakness is that it does not adapt: legitimate long-term changes, like a well's natural decline, make a static baseline steadily less relevant and generate false deviations.
A rolling (or adaptive) baseline is recomputed continuously from recent data - often a rolling average over a trailing window - so it tracks slow, legitimate change while still flagging sudden departures. It suits assets whose normal genuinely evolves. The trade-off is that a rolling baseline can slowly absorb a real, gradual problem as if it were the new normal, so it is often paired with a static reference to catch long-term drift the rolling window would miss.
A baseline is only as good as the data and period it is built from. The core idea is to characterize normal operation from a representative stretch of history - a period known to be healthy - capturing not just the average value but the normal spread, so the natural variation is understood. A baseline usually pairs a central value with a band of acceptable variation, and deviation is measured relative to that band rather than a single line.
Context matters: a compressor's normal discharge temperature depends on its load, and a well's normal production depends on where it is in its decline. Sophisticated baselines are conditional, defining normal as a function of operating state rather than one flat number. Poor-quality data must be excluded when building a baseline, or the reference itself inherits the errors - which is why data-quality flags feed into baseline calculation.
Baselines underpin the whole condition-monitoring and analytics stack. Trend deviation measures distance from baseline; anomaly detection defines anomalies as departures from a baseline model; KPI targets are baselines the business manages against; and energy or emissions programs compare current performance to a reference to prove improvement. Nearly every 'is this normal?' question resolves to a comparison with a baseline.
Establishing and maintaining baselines requires historized data across time and assets. A cloud SCADA like Merobix stores the tags it reads from field devices over Modbus, DNP3, and OPC UA, providing the historical record from which baselines are computed and against which live readings are compared. The baseline logic and analytics are a layer on top; the platform supplies the clean, contextualized history that any credible baseline is built from.
It is a reference representing normal or expected performance, used as the yardstick to judge whether current data is on track or deviating. Derived from historical data or design expectations, it underpins deviation detection, anomaly analysis, and performance targets - you need it to say whether a reading is abnormal.
A static baseline is fixed from a known-good period or design spec and held constant, ideal when normal should not change. A rolling baseline is recomputed continuously from recent data so it tracks legitimate long-term change. Rolling adapts but can absorb a slow real problem as normal, so the two are often used together.
A setpoint is the target value a control loop actively drives a process toward. A baseline is a reference for what normal performance looks like, used to judge deviations - it is descriptive, not something a controller pursues. A setpoint says where the process should be held; a baseline says what healthy operation should look like.
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.