TLDR
An out-of-control signal on a control chart means the process has changed and needs investigation. The Western Electric Rules define eight patterns that indicate special cause variation, including points beyond the control limits, runs, trends, and stratification.
This guide explains all eight rules, how to spot each pattern, what each signal typically means about the process, and what action to take when you detect one.
Beyond the Control Limits
Most people learn one rule about control charts: if a point falls outside the control limits, something went wrong. That rule is correct, but it is only the beginning. A process can show clear signs of instability while every point remains between the Upper Control Limit (UCL) and Lower Control Limit (LCL).
The Western Electric Rules (also known as the zone rules) and the related Nelson Rules provide a complete framework for reading control charts. They identify patterns in the data that are statistically unlikely to occur by chance in a stable process. When you see these patterns, the process is telling you something has changed.
Understanding the Zones
To apply the rules, divide the control chart into zones based on standard deviation units (sigma) from the center line.
Zone C: The area within 1 sigma of the center line (above and below). Approximately 68% of points should fall here in a stable process.
Zone B: The area between 1 sigma and 2 sigma from the center line. Approximately 27% of points should fall here.
Zone A: The area between 2 sigma and 3 sigma (the control limits). Approximately 4.3% of points should fall here.
Beyond 3 sigma is outside the control limits. In a stable, normally distributed process, only about 0.27% of points (roughly 3 in 1,000) should fall outside. When a point does, it is a strong signal.
The Eight Out-of-Control Rules
Rule 1: One Point Beyond 3 Sigma
A single point falls above the UCL or below the LCL.
What it means: Something sudden and significant happened. A wrong setting, a broken tool, a bad batch of material, or a measurement error. This is the most obvious and urgent signal.
Action: Investigate immediately. Identify what changed at the time of the out-of-control point. Contain any potentially nonconforming product produced since the last known good point.
Rule 2: Nine Consecutive Points on the Same Side of the Center Line
Nine or more points in a row fall above (or below) the center line without crossing it.
What it means: The process mean has shifted. Even though no individual point may exceed the control limits, the sustained run on one side indicates the process is no longer operating at its original average. Common causes include tool wear, gradual material change, or a setup adjustment that shifted the target.
Action: Investigate the process shift. Determine if the change is intentional (such as a tooling adjustment) or unintentional. If unintentional, identify and correct the cause.
Rule 3: Six Consecutive Points Steadily Increasing or Decreasing
Six or more consecutive points are trending in one direction, each point higher (or lower) than the previous one.
What it means: The process is drifting. This pattern is often associated with tool wear, temperature drift, gradual equipment degradation, or a slowly changing input variable. If the trend continues, it will eventually produce points beyond the control limits.
Action: Identify the drifting variable. Implement corrective action before the process reaches the control limits. Consider whether preventive maintenance or input monitoring can prevent future drift.
Rule 4: Fourteen Consecutive Points Alternating Up and Down
Fourteen or more consecutive points alternate: up, down, up, down, with no two consecutive points moving in the same direction.
What it means: This “sawtooth” pattern suggests systematic alternation in the process. Common causes include two alternating spindles, two raw material streams, over-adjustment (tampering), or rotation between two fixtures or operators with different setups.
Action: Investigate whether there are two distinct process streams being charted together. If the alternation is caused by over-adjustment, stop adjusting and let the process run at its natural center.
Rule 5: Two Out of Three Consecutive Points in Zone A or Beyond (Same Side)
Two of three consecutive points fall in Zone A (between 2 and 3 sigma) or beyond, on the same side of the center line.
What it means: The process variation has increased or the process mean has shifted. While any single point in Zone A is not unusual, having two out of three consecutive points this far from the center is statistically improbable in a stable process.
Action: Investigate for a shift or increase in variation. Check for changes in incoming material, equipment condition, or environmental factors.
Rule 6: Four Out of Five Consecutive Points in Zone B or Beyond (Same Side)
Four of five consecutive points fall in Zone B (between 1 and 2 sigma) or beyond, on the same side of the center line.
What it means: Similar to Rule 5 but a smaller, more sustained shift. The process mean may be drifting or has shifted slightly. This rule catches smaller shifts that Rule 1 would miss.
Action: Investigate for a subtle process change. Review recent adjustments, material lots, or operator changes.
Rule 7: Fifteen Consecutive Points in Zone C (Both Sides)
Fifteen or more consecutive points fall within Zone C, the area within 1 sigma of the center line on both sides.
What it means: This is called “stratification.” The data is hugging the center line too closely. Counterintuitively, this is a problem. It usually means the control limits are too wide because the data used to calculate them included more than one process stream (different machines, different shifts, different materials) that should have been charted separately.
Action: Re-examine how the data was collected and whether the subgroups are rational. Consider separating the data by process stream and recalculating limits for each.
Rule 8: Eight Consecutive Points on Both Sides with None in Zone C
Eight or more consecutive points fall beyond Zone C (more than 1 sigma from the center line) on both sides, with no points in Zone C.
What it means: This is called “mixture.” The data shows too much variation near the control limits and too little near the center. It typically indicates that two distinct distributions are being plotted on the same chart, such as output from two machines with different averages.
Action: Identify the two (or more) process streams. Separate them onto individual control charts with their own limits. Each stream may be in control individually even though the combined data appears unstable.
How to Respond to Out-of-Control Signals
Every out-of-control signal requires a defined response. Your control plan should include a reaction plan that specifies what to do when a signal is detected.
Step 1: Stop and assess. Do not ignore the signal or continue producing without investigation. If the process is making nonconforming product, contain it.
Step 2: Investigate the assignable cause. Determine what changed. Review the process inputs (man, machine, material, method, measurement, environment) at the time of the signal.
Step 3: Take corrective action. Address the root cause. Adjust the process, replace the worn tool, retrain the operator, or correct whatever produced the signal.
Step 4: Document. Record the signal, the investigation, the cause identified, and the corrective action taken. This documentation supports audit readiness and helps prevent recurrence.
Step 5: Verify. Confirm that subsequent data points return to a stable, in-control pattern after the corrective action.
Common Mistakes When Applying the Rules
Applying too many rules simultaneously. Using all eight rules at once increases the false alarm rate. Most SPC implementations use Rules 1 through 4 as the standard set. Add Rules 5 through 8 for critical characteristics where sensitivity matters more than false alarm risk.
Ignoring signals on the R or S chart. Operators and engineers sometimes focus only on the X-bar chart and overlook out-of-control signals on the range or standard deviation chart. Both charts matter. A change in within-subgroup variation is a different and equally important signal.
Treating the rules as absolute. The rules are probability-based guidelines, not laws of physics. A single Rule 2 violation does not guarantee a process shift. It means a shift is statistically likely and warrants investigation. Use judgment alongside the statistics.
Frequently Asked Questions
The Western Electric Rules are a set of decision rules for identifying out-of-control conditions on control charts. Originally published in the Western Electric Statistical Quality Control Handbook in 1956, they define patterns (beyond single limit violations) that indicate a process is no longer stable.
The Nelson Rules, published by Lloyd Nelson in 1984, expanded on the Western Electric Rules by adding additional patterns. The eight rules described in this article are commonly attributed to both sets. In practice, the terms are often used interchangeably in manufacturing SPC.
Most SPC programs apply Rules 1 through 4 as a standard set. Adding all eight rules increases sensitivity but also increases the false alarm rate. Apply additional rules selectively for high-risk or critical characteristics where early detection is more important than minimizing false alarms.
An out-of-control condition means that special cause variation has been detected. Something has changed in the process that is not part of the normal, expected variation. It does not necessarily mean defective parts were produced; it means the process has shifted and needs investigation.
Yes. A process can show out-of-control signals while all parts remain within specification limits. However, an unstable process is unpredictable, and continued operation without corrective action increases the risk of producing nonconforming parts.
Rule 1 (a single point beyond the 3-sigma control limit) is the most commonly detected signal. It is also the most intuitive and is the only rule that many operators are trained to recognize. However, Rule 2 (runs) and Rule 3 (trends) are equally important for catching gradual process changes.
Over-adjustment occurs when an operator adjusts the process in response to every data point, treating common cause variation as if it were a special cause. This actually increases variation and can produce a sawtooth pattern (Rule 4). The correct approach is to adjust only when a true out-of-control signal is detected.
Stay Current on Automotive Quality
PPAP updates, GD&T guidance, and practical quality resources delivered to your inbox. No spam, just what quality professionals need to know.
Subscribe
Leave a Comment