Statistical Process Control

What Is Statistical Process Control?

Statistical process control was developed by Walter Shewhart at Bell Laboratories in the 1920s and later popularized by W. Edwards Deming, who introduced the method to Japanese manufacturers after World War II. The approach is built on a straightforward premise: all processes contain variation, and not all variation requires a response.

SPC gives operations and quality control teams a systematic way to tell the difference between random noise and a genuine signal. When a process is in statistical control, its variation is predictable and bounded. When a special cause enters the system, the control chart reveals it through recognizable patterns or points that fall outside calculated boundaries.

Today, SPC is standard practice across manufacturing, pharmaceuticals, food processing, automotive production, and aerospace. It underpins ISO 9001 quality management systems and is a primary analytical tool in Six Sigma programs worldwide.

How SPC Works

The core mechanism of SPC is the control chart. A control chart is a time-series graph that plots a process measurement against three reference lines: the center line (the process mean), the upper control limit (UCL), and the lower control limit (LCL).

Control limits are typically set at three standard deviations above and below the mean. This boundary contains approximately 99.73% of all data points in a normally distributed, stable process. Any point that falls outside these limits is a signal that something has changed.

Common Cause vs. Special Cause Variation

Shewhart identified two fundamentally different types of variation:

• Common cause variation is inherent to the process itself. It comes from many small, random sources such as minor material differences, ambient temperature fluctuations, or normal operator movement. A process showing only common cause variation is considered "in control." Adjusting it based on individual data points makes it worse, not better.
• Special cause variation comes from a specific, identifiable source outside the normal process. A worn tool, a bad batch of raw material, or a shift in machine calibration can all produce special cause variation. When detected, the team must investigate and correct the root cause.

The critical discipline of SPC is acting on special causes and not reacting to common causes. Treating common variation as a problem by constantly adjusting a machine or process is called "tampering" and increases overall variability rather than reducing it.

Control Chart Signals

A point outside the control limits is the most obvious signal, but SPC practitioners also watch for non-random patterns within the limits. The Western Electric rules (also called Nelson rules) define additional signals such as:

• Eight or more consecutive points on the same side of the center line
• Six consecutive points trending steadily upward or downward
• Two out of three consecutive points in the outer one-third of the chart (between 2 and 3 standard deviations from the mean)

These patterns indicate that the process mean or variance has shifted, even if no individual point has crossed a limit.

Key SPC Tools

Different process types and data types require different control chart formats. The table below summarizes the most commonly used SPC charts.

The X-bar and R chart is the most common combination for variable data where samples of two to ten units are collected at regular intervals. The X-bar chart monitors whether the process mean is shifting, while the R chart monitors whether the process spread is increasing or decreasing. Both must be in control for the process to be considered stable.

Process Capability: Cp and Cpk

Control charts tell you whether a process is stable. Process capability indices tell you whether a stable process is good enough to meet specification.

Cp (process capability) compares the width of the specification limits to the natural width of the process (six standard deviations). A Cp above 1.33 is generally considered capable. Cp does not account for where the process mean sits relative to the specification midpoint.

Cpk (process capability index) adjusts for process centering. A Cpk of 1.0 means the process mean sits exactly one-third of the way between the center and the nearest specification limit. A Cpk of 1.33 or above is a common industrial target. If Cp and Cpk are similar, the process is well-centered. A large gap between them means the process is off-center and needs adjustment.

Capability analysis is only valid when the process is first confirmed to be in statistical control. Calculating Cpk on an unstable process produces misleading results.

SPC and Six Sigma

SPC and Six Sigma are complementary, not competing. Six Sigma is a structured improvement methodology built around the DMAIC cycle (Define, Measure, Analyze, Improve, Control). SPC is the primary tool used in the Control phase to lock in gains after a process has been improved.

Before Six Sigma, teams often improved a process and then watched it drift back to its previous state over time. By placing a control chart on the key output variable after an improvement, teams can detect any regression immediately and investigate before the process fully degrades.

The relationship also runs in the other direction. In the Measure and Analyze phases, SPC data provides baseline capability information and identifies which factors contribute most to output variation. Continuous improvement programs that combine DMAIC with ongoing SPC monitoring achieve more durable results than those that rely on either method alone.

A useful companion tool for prioritizing which defects or variation sources to address first is the Pareto chart, which ranks problems by frequency or impact so teams focus effort where it matters most.

SPC in Manufacturing Maintenance

SPC is not limited to product quality. Maintenance teams use it to track equipment condition metrics over time, turning it into an early warning system for asset degradation.

Any measurable, repeating parameter can be charted: vibration amplitude, bearing temperature, motor current draw, hydraulic pressure, or cycle time. When a machine is healthy and running normally, these readings fluctuate within a predictable band. When something changes, such as a bearing beginning to wear, a seal starting to leak, or a drive belt losing tension, the readings shift in ways that break the established pattern.

By applying control limits to sensor readings, maintenance engineers can identify the moment a process metric moves from common cause variation to special cause variation. That moment is the signal to investigate. In many cases, the deviation appears days or weeks before the equipment would have failed completely, giving teams time to schedule a repair during planned downtime rather than reacting to an unplanned outage.

This approach directly supports process reliability goals. Rather than setting a fixed inspection interval, teams respond to data. The result is fewer unnecessary interventions on healthy equipment and faster response when a real problem begins to develop.

SPC also supports overall equipment effectiveness improvement efforts. Tracking quality rate, performance rate, and availability metrics on control charts reveals whether OEE variation is stable (a process design problem) or driven by identifiable special causes (a specific equipment, material, or operator event).

How to Implement SPC

A successful SPC implementation follows a consistent sequence.

1. Select the Right Process Characteristic

Choose a measurement that is directly linked to quality, cost, or reliability. For product quality, this might be a critical dimension or a weight. For maintenance, it might be a vibration reading on a critical pump or compressor.

2. Define the Sampling Plan

Decide how often to sample, how many units to include per sample, and how to record the data. Samples should be drawn under consistent conditions (same machine, same shift, same raw material batch where possible) so the chart reflects actual process variation rather than measurement noise.

3. Collect Baseline Data

Before drawing control limits, collect at least 20 to 25 subgroups of data under normal operating conditions. This baseline establishes the process mean and standard deviation that the control limits will be calculated from.

4. Calculate and Plot Control Limits

Use the baseline data to calculate the center line, UCL, and LCL. Plot the existing data and look for any signals. If signals appear in the baseline, investigate them before moving forward. Control limits calculated from an unstable baseline are not valid.

5. Monitor and React

Once the chart is live, apply the response plan consistently. Every signal is investigated. Every investigation is documented. This documentation creates the feedback loop that enables the process to improve over time. Tools like the Shewhart cycle (Plan-Do-Check-Act) provide the iterative framework for moving from detection to root cause to permanent correction.

6. Recalculate Limits After Process Changes

If a deliberate process improvement is made, the old control limits no longer reflect the new process. Collect a new baseline after the change and recalculate. Using outdated limits will either generate false signals on the improved process or fail to detect genuine problems.

Common Mistakes in SPC

Several errors undermine SPC programs in practice.

Reacting to every data point. This is the tampering problem. Operators who adjust a machine every time a reading moves away from the center line are treating common cause variation as a special cause. The adjustment introduces additional variation and makes the output less consistent.

Using specification limits as control limits. Specification limits are defined by the customer or the engineering team. Control limits are calculated from the process data. Conflating the two leads to incorrect decisions about when the process needs attention.

Charting the wrong variable. Selecting a measurement that is weakly correlated with actual quality or failure risk produces a chart that generates noise without delivering value. The variable charted must have a direct connection to the outcome being managed.

Failing to investigate signals. A control chart that generates signals but receives no response quickly becomes an ignored chart. The discipline of investigating every signal is what creates the learning loop that makes SPC valuable over time.

Tracking defect density alongside control chart data helps teams verify that SPC interventions are actually reducing the rate of defects reaching downstream operations or customers.

The Bottom Line

Statistical process control is one of the most practical tools available for maintaining quality and reliability in manufacturing operations. By separating signal from noise, it allows teams to respond to genuine process changes without wasting resources on adjustments to normal variation.

The method applies equally to product quality and equipment health. In both cases, the logic is the same: establish what normal looks like, monitor for deviations, investigate deviations promptly, and use what you learn to improve the process over time.

For maintenance teams, integrating SPC with continuous sensor monitoring closes the gap between scheduled inspection intervals and real-time asset condition. The result is fewer unexpected failures, better planned maintenance scheduling, and stronger overall process reliability.

Frequently Asked Questions