A single reading tells you where a process is right now, but it says nothing about whether the process is behaving as it always has. A control chart answers that second question by plotting a variable over time against limits derived from the process's own history, so ordinary variation and genuine change can be told apart. This guide explains what a statistical process control chart is, why its control limits are not the same as spec limits or alarm limits, and how pattern rules flag a process that has gone out of control.
Control Chart (SPC) in one line: A control chart is the central tool of statistical process control (SPC): a time plot of a measured variable with a center line and upper and lower control limits calculated from the process's own natural variation. Points that fall inside the limits and scatter randomly indicate a stable, in-control process, while points outside the limits or forming non-random patterns signal that something has changed. Crucially, control limits describe how the process actually behaves, not what a customer requires or when an alarm should sound, which is what separates SPC from specification and alarm limits.
The most common confusion about control charts is treating the control limits as if they were requirements. They are not. Control limits are calculated from the process's own measured variability - typically set a few standard deviations above and below the center line - so they describe the range within which the process naturally wanders when nothing unusual is happening. They say what the process does, entirely independent of what anyone wants it to do. A process can be perfectly in statistical control and still be producing product that fails specification, because control limits and spec limits answer completely different questions.
Specification limits, by contrast, are set by the customer, the contract, or the design, and they define what is acceptable in the finished product regardless of how the process behaves. Alarm limits are different again: they are set for operator action, marking the point at which someone needs to intervene, and are usually placed with safety and equipment protection in mind rather than statistics. A control chart deliberately keeps these separate because they serve separate purposes - control limits detect that a stable process has changed, spec limits judge whether output is acceptable, and alarm limits demand a response. Plotting all three on the same axes only invites the mistake of reacting to normal variation as though it were a failure.
A control chart flags trouble in two ways. The obvious one is a single point falling outside the control limits, which is unlikely to happen by chance in a stable process and therefore signals a real change. But a process can drift out of control while every individual point still sits inside the limits, so control charts also look at patterns across runs of points. The Western Electric rules are the classic set of tests for these patterns - detecting, for example, a run of consecutive points all on one side of the center line, a steady trend of points marching up or down, or several points in a row clustered far from the center even though none has crossed a limit.
The reason pattern rules matter is that they distinguish two kinds of variation. Common-cause variation is the ordinary, random scatter inherent in a stable process; reacting to it is wasted effort and often makes things worse, a mistake known as tampering. Special-cause variation is a genuine, assignable change - a new batch of feed, a worn part, a shifted setpoint - and it shows up as a point outside the limits or as a non-random pattern. The whole discipline of SPC is knowing when to leave a process alone and when to investigate, and the control chart with its rules is the instrument that makes that judgment objective rather than a matter of an operator's nerves. Because these tests can produce false signals if too many are applied at once, teams usually select a focused subset appropriate to the process.
Applying SPC to industrial data requires two things: a faithful record of how a variable has behaved, from which to compute the control limits, and a live stream to plot against them. A cloud SCADA such as Merobix supplies both, gathering process tags from the field over Modbus, DNP3, OPC UA, and MQTT and archiving them, so the natural variability that defines a process's control limits can be established from real history rather than assumed. Once the limits are set, the same live tags can be watched against them continuously.
Running SPC on operational data does demand care about what a control chart assumes. Classic charts assume points are independent samples, whereas high-frequency SCADA data is often heavily autocorrelated - consecutive readings from a slow-moving temperature are far from independent - which can make naive control limits far too tight and fill the chart with false signals. In practice this is handled by charting at a sensible interval, by aggregating raw samples into subgroups, or by charting a variable that genuinely varies sample to sample. Used with that discipline, SPC layered on top of a SCADA historian gives an operator an objective, statistically grounded way to notice that a stable process has shifted, complementing the fixed alarm limits that only fire when a value crosses a hard threshold.
Control limits are calculated from the process's own natural variation and describe how it actually behaves, so they detect when a stable process has changed. Specification limits are set by the customer or design and define what the finished product must meet. A process can be in statistical control yet still out of spec, or in spec while statistically out of control, because the two limits answer entirely different questions.
They are a classic set of pattern tests applied to a control chart to detect out-of-control conditions beyond a single point crossing a limit. They flag things like a run of consecutive points on one side of the center line, a steady upward or downward trend, and clusters of points sitting far from the center. Together they catch a process drifting out of control even while individual points remain inside the limits.
Yes, but with care, because classic control charts assume independent samples and fast SCADA data is usually autocorrelated. Consecutive readings from a slow variable are highly related, which makes naive control limits too tight and generates false alarms. The common fixes are to chart at a sensible interval, aggregate raw samples into subgroups, or apply methods designed for autocorrelated data so the limits reflect real variation.
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.