Control charts identify process problems by distinguishing common cause variation (the normal, expected noise in any process) from special cause variation (a signal that something has changed). They plot data over time with upper and lower control limits calculated from the data itself. When points fall outside the limits or show non-random patterns, the process is unstable and needs investigation. When points stay within the limits with no patterns, the variation is inherent to the process and can only be reduced by changing the system.
For Lean Six Sigma practitioners, control charts are the primary tool for answering one question: is the variation I'm seeing normal, or is something going on? Without a control chart, every dip and spike looks like a problem. With one, you know exactly when to act and when to leave the process alone.
Common Cause vs Special Cause Variation
Every process has variation. The question is whether that variation comes from the process itself (common cause) or from something outside it (special cause).
| Common Cause Variation | Special Cause Variation |
|---|---|
| Inherent in the process | Comes from an external event or change |
| Predictable within limits | Unpredictable and intermittent |
| Only reduced by changing the process system | Eliminated by finding and removing the specific cause |
| Examples: machine tolerance, ambient temperature swings, normal material variation | Examples: operator change, batch defect, machine breakdown, measurement error |
The biggest mistake in process improvement is treating common cause variation as if it were special cause. That leads to knee-jerk adjustments that make the process worse, not better. Control charts prevent that mistake by giving you a statistical basis for deciding when to act.
The 7 Warning Patterns (Western Electric Rules)
A point outside the control limits is the most obvious signal, but it's not the only one. The Western Electric rules identify seven patterns that indicate special cause variation — even when no single point is beyond the limits.
Rule 1: Any point beyond 3 sigma
A single point outside the upper or lower control limit. This is the most obvious signal — something significant changed in the process at that moment.
Rule 2: 2 of 3 consecutive points beyond 2 sigma (same side)
Two out of three consecutive points fall beyond 2 sigma on the same side of the centre line. The process is shifting before it crosses the 3-sigma line.
Rule 3: 4 of 5 consecutive points beyond 1 sigma (same side)
Four out of five consecutive points fall beyond 1 sigma on the same side. A smaller but sustained shift that's easier to catch with this rule.
Rule 4: 8 consecutive points on one side of the centre line
Eight points in a row on the same side of the centre line. The process has shifted — the average is no longer where it should be.
Rule 5: 6 consecutive points trending up or down
Six points in a consistent upward or downward trend. Something is gradually changing — tool wear, temperature drift, material degradation.
Rule 6: 14 consecutive points alternating up and down
Fourteen points in a zigzag pattern, alternating above and below the centre line. This suggests two different sources of data being mixed (e.g., two operators, two machines, two suppliers).
Rule 7: 14 consecutive points within 1 sigma (either side)
All 14 points cluster within 1 sigma of the centre line — less variation than expected. This can mean the control limits are too wide, or data is being rounded or filtered before plotting.
Key point: You don't need to memorise all seven rules to use a control chart. The free Control Limits Calculator handles the math, and the How to Read a Control Chart guide walks through interpretation step by step. But understanding that these patterns exist helps you catch problems earlier — before a point goes out of limits.
Which Control Chart Should You Use?
The chart type depends on your data. Using the wrong chart type is a common mistake that produces misleading limits.
I-MR Chart
For individual measurements (one data point per time period). Use when you can't collect subgroups — e.g., daily temperature, individual test results.
X-bar R Chart
For subgroups of 2-10 measurements. Use when you collect multiple samples per time period — e.g., 5 parts measured every hour.
p-chart
For proportion of defective items. Use when each item is pass/fail and the subgroup size varies — e.g., % of orders with errors per day.
np-chart
For count of defective items. Use when each item is pass/fail and the subgroup size is constant — e.g., number of defects in a batch of 100.
c-chart
For count of defects per unit. Use when the number of defects is counted and the inspection unit is constant — e.g., scratches per panel.
u-chart
For defects per varying unit. Use when the inspection unit size varies — e.g., defects per square metre of fabric.
Not sure which to use? The Control Limits Calculator supports all six chart types and calculates UCL and LCL from your data automatically.
A Worked Example: Order Processing Time
The Scenario
A team is tracking order processing time in minutes. They collect one measurement per day for 20 days. The average is 14.2 minutes and the standard deviation is 2.1 minutes.
Setting Up the Chart
Using the Control Limits Calculator, the control limits are:
- Centre line (average): 14.2 minutes
- Upper Control Limit (UCL): 20.5 minutes (average + 3 sigma)
- Lower Control Limit (LCL): 7.9 minutes (average − 3 sigma)
What They Found
Days 1-12: all points within limits, no patterns. The process is stable.
Day 13: a point at 21.1 minutes — beyond the UCL. Rule 1 violation.
Days 14-17: four consecutive points above 16.2 minutes (beyond 1 sigma). Rule 3 violation — 4 of 5 beyond 1 sigma on the same side.
Day 18: back within limits but still above the centre line.
The Interpretation
Something changed on or around Day 13. The initial spike (Rule 1) confirms a specific event occurred. The sustained high readings through Day 17 (Rule 3) suggest the change was not a one-off — the process average has shifted upward.
What to Do
This is special cause variation. The team should investigate what happened around Day 13 — staff change, new software, volume spike, process change. They should not simply adjust the limits or ignore the signal.
When to Use a Control Chart in DMAIC
| DMAIC Phase | How Control Charts Are Used |
|---|---|
| Define | Not typically used — the problem is being scoped, not measured yet. |
| Measure | Baseline the current process. Confirm stability before calculating capability. If the process is unstable, fix that first — capability from an unstable process is meaningless. |
| Analyse | Compare before and after charts to confirm that changes are statistically significant, not just random variation. |
| Improve | Monitor the process during pilot or implementation to detect any new issues introduced by the change. |
| Control | Establish ongoing monitoring with updated limits reflecting the improved process. Set up response plans for when signals appear. |
Common Control Chart Mistakes
- Calculating capability before confirming stability. Cp and Cpk from an unstable process tell you nothing. Run a control chart first.
- Reacting to every point as if it's special cause. If no rules are violated, leave the process alone. Over-adjustment increases variation.
- Using the wrong chart type. Attribute data on an I-MR chart, or continuous data on a p-chart, produces wrong limits. Match the chart to the data.
- Not recalculating limits after improvement. Old limits on an improved process will show false signals. Recalculate after changes are confirmed.
- Ignoring the warning patterns. Waiting for a point to go out of limits means catching problems late. The Western Electric rules exist to catch shifts early.
Free Control Chart Tools
- Control Limits Calculator — calculates UCL and LCL for X-bar R, I-MR, p, np, c and u charts. Free, no login.
- Control Chart Template — downloadable template for tracking your process data.
- How to Read a Control Chart — detailed guide to chart selection and interpretation.
- Statistical Process Control Guide — broader guide covering SPC, poka-yoke integration and variation theory.
- Statistical Tools for Lean Six Sigma — full hub of all 10 core statistical tools with calculators.
Frequently Asked Questions
Control charts identify process problems by distinguishing common cause variation (normal, expected noise) from special cause variation (a signal that something changed). They plot data over time with upper and lower control limits calculated from the data itself. When points fall outside the limits or show non-random patterns, the process is unstable and needs investigation.
The Western Electric rules are seven statistical tests that detect non-random patterns in control chart data. Rule 1: any point beyond 3 sigma. Rule 2: 2 of 3 consecutive points beyond 2 sigma. Rule 3: 4 of 5 consecutive points beyond 1 sigma. Rule 4: 8 consecutive points on one side of the centre line. Rules 5-7 detect trends, oscillation and stratification. Any rule violation indicates special cause variation.
Common cause variation is the natural, expected variation inherent in a process — it comes from the system itself and can only be reduced by changing the process. Special cause variation is unexpected and comes from something outside the normal system — a machine change, operator error, material defect or environmental shift. Control charts are the primary tool for telling them apart.
Choose based on your data type. For continuous data (measurements like time, weight, temperature), use I-MR charts for individual measurements or X-bar R charts for subgroups. For attribute data (counts or proportions), use p-charts for proportions, np-charts for counts, c-charts for defect counts per unit, or u-charts for defects per varying unit.
Yes, but it requires manual setup of the control limits and formatting. SimplicityHub provides a free Control Limits Calculator that calculates UCL and LCL for all major chart types, plus a free control chart template. For guided workflow with automatic charting, the SimplicityHub DMAIC Assistant embeds control charts directly into the Measure phase.