Statistical Process Control, commonly abbreviated to SPC, is a statistical method for monitoring and controlling a process by analysing measurements over time and distinguishing expected random variation from evidence that the process may have changed. Its central concept is that every process exhibits variation, and effective control requires understanding the difference between common-cause variation inherent in the process and special-cause variation associated with an identifiable change or disturbance. Control charts are among the principal SPC tools.
They plot observations sequentially against a centre line and statistically derived control limits, allowing patterns or observations inconsistent with the established process behaviour to be investigated. SPC does not simply compare every observation with product specifications. A process can remain within specification while becoming statistically unstable, and a statistically stable process can still consistently produce output outside customer requirements if its natural variation is too large or its average is incorrectly centred.
The foundations of SPC are strongly associated with Walter A. Shewhart, who developed control-chart methods at Bell Laboratories in the early twentieth century. Shewhart recognised that distinguishing random variation from assignable causes was essential to managing industrial processes. His work influenced subsequent quality-management thinkers and became a foundation for statistical quality control. SPC later expanded through broader statistical methods, process capability analysis and industrial quality programmes.
It became particularly influential in high-volume manufacturing but is also applicable to healthcare, finance, logistics, software operations and other settings where measurements are collected sequentially and process behaviour can be modelled. The underlying logic remains broadly applicable even when the specific chart type changes according to the data.
A control chart typically contains observed process measurements, a central estimate of the process level and upper and lower control limits calculated from process data or specified statistical assumptions. The limits are intended to represent the range of variation expected when the process is operating under its established conditions, not customer specification limits.
Common chart types include X-bar and R charts for subgrouped continuous measurements, individual-and-moving-range charts for individual observations, p and np charts for attribute proportions or counts, and c and u charts for defect counts under appropriate assumptions. The selection depends on whether the data are continuous or attribute-based, whether observations can be grouped rationally and what distributional assumptions are appropriate. Applying the wrong chart can produce misleading signals and unnecessary interventions.
SPC requires careful sampling and rational subgrouping because the statistical meaning of a control chart depends on how observations are collected. A rational subgroup is designed so that variation within the subgroup primarily represents short-term common-cause variation, while differences between subgroups can reveal process changes over time. Control limits should generally be estimated from representative process data when establishing a baseline, and unusual points should be investigated rather than automatically removed.
Data integrity is essential because missing observations, changes in measurement methods and biased sampling can distort the apparent process behaviour. Analysts should also consider autocorrelation, seasonality, changing demand and other characteristics that can violate simple statistical assumptions. SPC is therefore not simply the mechanical plotting of numbers; it requires understanding how the data were generated and what the process actually represents.
The distinction between control limits and specification limits is fundamental to SPC. Control limits describe statistically expected process behaviour, while specification limits represent requirements established by customers, designers, regulators or other authorities. If the process has excessive natural variation, it may frequently produce results outside specification even if it is statistically stable. Capability analysis uses indices such as Cp and Cpk under appropriate assumptions to assess how process variation and centring relate to specification limits.
A capable process is not necessarily under statistical control, because an unstable process can temporarily appear capable. Conversely, a controlled process may be incapable of meeting requirements consistently. Organisations therefore often use SPC to establish and maintain statistical control and capability analysis to evaluate whether controlled process performance is sufficient for the required specification.
SPC supports prevention because it can identify process changes before large numbers of defective outputs are produced. A sudden shift in the process average, a sustained trend or an unusual sequence of observations can prompt investigation before final-product inspection reveals widespread failures. This is particularly valuable when defects are expensive to correct or when inspection is destructive. However, SPC does not identify the cause automatically.
A control-chart signal indicates that the process behaviour warrants investigation; engineers or process owners must determine whether the cause relates to equipment, materials, environment, measurement, personnel, software, demand or another factor. Unnecessary adjustment in response to normal common-cause variation can actually increase variability, a phenomenon sometimes called tampering. Proper SPC therefore requires disciplined interpretation rather than reacting to every fluctuation.
SPC is related to Quality Control and Quality Improvement but has a different emphasis. QC can inspect outputs to determine whether they meet specifications, while SPC monitors the behaviour of the process generating those outputs. Quality Improvement can use SPC to establish baselines, identify instability and evaluate whether an intervention changed process behaviour.
Acceptance Sampling also uses statistical sampling but serves a different purpose: it determines whether a lot or batch should be accepted based on a sample, whereas SPC focuses on ongoing process behaviour. SPC can be used even when no individual product is being inspected against a pass/fail criterion, provided suitable process measurements are available. This makes it particularly useful for continuous operations, service processes and automated systems where real-time performance data can reveal changes before a formal defect is recorded.
Modern SPC increasingly uses automated data collection, statistical software, sensor networks and real-time dashboards. Manufacturing systems can automatically collect dimensional, temperature, pressure or process-speed measurements and trigger investigation when statistically significant patterns appear. Service organisations can apply related concepts to transaction times, error rates, queue durations or operational reliability, provided the underlying process and data assumptions are appropriate. Advanced analytics can supplement traditional control charts, but sophisticated algorithms do not remove the need to distinguish common and special causes.
SPC remains valuable because it provides a disciplined statistical language for understanding process variation. Its strongest application is not simply detecting bad observations but helping organisations avoid unnecessary intervention, identify meaningful changes promptly and improve the underlying process so that its natural behaviour reliably satisfies the required performance range.
