What is the difference between common cause and special cause variation?

Common cause variation is inherent to the process — random variation that is always present and predictable in aggregate (slight material variations, minor machine vibrations, normal operator variations). It requires fundamental process changes to address. Special cause variation comes from assignable causes — abnormal variation from identifiable sources that are unpredictable and sporadic (equipment breakdown, wrong material batch, untrained operator). Special causes should be investigated immediately for quick wins.

What is the difference between control limits and specification limits?

Control Limits (UCL/LCL) are calculated from process data using ±3σ from the mean. They represent the Voice of Process — what the process naturally produces. Specification Limits (USL/LSL) are set by customer requirements (Voice of Customer). Control limits and spec limits are COMPLETELY different! Control limits tell you if the process is stable; spec limits tell you if output meets customer requirements.

When should I use an I-MR chart?

Use an I-MR chart (Individual-Moving Range) when you have: (1) one measurement per time period (e.g., daily production output), (2) destructive or expensive testing where you can't afford multiple samples, (3) slow-moving processes like batch processes or chemical reactions, or (4) automated measurement systems that provide one reading per unit. I-MR is the most common control chart for continuous data with individual measurements.

What are the 8 control chart detection rules?

The 8 Western Electric Rules (Nelson Rules) are: (1) One point beyond 3σ, (2) Nine points in a row on same side of center line, (3) Six points in a row steadily increasing or decreasing, (4) Fourteen points alternating up and down, (5) Two out of three points beyond 2σ on same side, (6) Four out of five points beyond 1σ on same side, (7) Fifteen points in a row within 1σ of center line, (8) Eight points in a row beyond 1σ on either side. Each rule detects different types of special cause variation.

How do I calculate I-MR control limits?

For I-MR charts: (1) Calculate X̄ (average of all individuals) and MR̄ (average of all moving ranges where MR = |Xi - Xi-1|). (2) For I Chart: UCL = X̄ + 2.66×MR̄, CL = X̄, LCL = X̄ - 2.66×MR̄. (3) For MR Chart: UCL = 3.27×MR̄, CL = MR̄, LCL = 0. The constants 2.66 and 3.27 convert moving range to standard deviation estimate for ±3σ control limits.

What types of control charts are there?

For continuous data: I-MR chart (individual measurements), X̄-R chart (subgroups of 2-10), X̄-S chart (subgroups >10). For attribute data: P chart (proportion defective), NP chart (number defective with constant sample size), C chart (count of defects per unit with constant area), U chart (defects per unit with variable area). Choose based on data type, sampling method, and sample size.

How many data points do I need for a control chart?

Minimum 25-30 subgroups (or 25+ individual measurements for I-MR charts) collected when the process is believed to be stable. This provides enough data for reliable control limit estimation. Never calculate control limits from unstable data — remove special causes first, then recalculate limits using only stable baseline data.

What is tampering in Statistical Process Control?

Tampering (also called overadjustment) is adjusting the process every time a data point moves, even when it's within control limits. This increases variation instead of reducing it. W. Edwards Deming emphasized: only react to out-of-control signals (points outside limits or detection rule violations). Random variation within control limits is NORMAL common cause variation — don't tamper with it!

Control Charts: Complete Guide to SPC, Continuous and Attribute Control Charts & Detection Rules

Master Statistical Process Control (SPC) with control charts. Learn how to detect special causes, interpret patterns, calculate control limits, and maintain process stability with I-MR, X̄-R, X̄-S continuous control charts, and P, NP, C, U attribute control charts.

What are Control Charts?

Control charts (also called Shewhart charts or Statistical Process Control charts) are graphical tools used to monitor process stability and detect special cause variation over time. They plot process data chronologically with statistically calculated control limits.

Control charts answer critical questions: "Is my process stable? Are there special causes affecting my process? When did the problem start?"

Voice of Process (VOP)

Control Limits (UCL/LCL): Calculated from process data using ±3σ from the mean. They represent the Voice of Process — what the process naturally produces.

≠ Specification Limits (USL/LSL): Set by customer requirements (Voice of Customer). Control limits and spec limits are completely different!

Why Use Control Charts?

Detect special causes: Identify when process changes due to assignable causes (equipment failure, material change, operator error)
Distinguish variation types: Separate common cause (inherent) from special cause (assignable) variation
Maintain process stability: Monitor ongoing performance and prevent defects before they happen
Reduce overreaction: Avoid tampering with the process when variation is normal (common cause)
Support continuous improvement: Provide data-driven evidence of process changes and improvements

W. Edwards Deming: "94% of the problems belong to the system (common causes). Only 6% are attributable to special causes." Control charts help you focus on the 6% that you CAN fix immediately (special causes) vs the 94% that require system-level improvements (common causes).

Common Cause vs Special Cause Variation

Common Cause Variation

Inherent to the process — random variation that is always present. Predictable in aggregate.

Examples:

Slight variations in raw materials
Minor machine vibrations
Temperature/humidity fluctuations
Normal operator variations
Measurement system variation

How to address: Requires fundamental process changes (new equipment, better materials, process redesign). Don't tamper!

Special Cause Variation

Assignable causes — abnormal variation from identifiable sources. Unpredictable and sporadic.

Examples:

Equipment breakdown or malfunction
Wrong material batch received
Untrained operator on shift
Power surge or outage
Measurement device out of calibration

How to address: Investigate immediately! Find root cause and eliminate. Quick wins possible!

Process Stability Definition

A process is stable and in control when it exhibits only common cause variation (random). All points fall within control limits and no patterns indicate special causes. A stable process is predictable.

I-MR Control Chart (Individual-Moving Range)

The I-MR chart (also called Individual-Moving Range chart) is the most common control chart for continuous data when you have individual measurements (not subgroups).

When to Use I-MR Charts

One measurement per time period (e.g., daily production output)
Destructive or expensive testing (can't afford multiple samples)
Slow-moving processes (batch processes, chemical reactions)
Automated measurement systems (one reading per unit)

I-MR Chart Components

An I-MR chart consists of TWO charts plotted together:

1. Individuals Chart (I Chart)

Plots individual measurements Xi over time. Detects shifts in process mean.

Xi for i = 1..n

Center Line: X̄ = average of all Xi data

UCL = X̄ + 2.66 × MR̄

LCL = X̄ - 2.66 × MR̄

2. Moving Range Chart (MR Chart)

Plots point-to-point variation. Detects changes in process variability.

MRi = |X_i - X_i-1| for i = 2..n

Center Line: MR̄ = average of all MRi

UCL = 3.27 × MR̄

LCL = 0 (typically)

Why these constants (2.66 and 3.27)?

These constants convert moving range to an estimate of standard deviation for control limits at ±3σ. They assume normal distribution and are derived from statistical theory. 2.66 = 3/d₂ where d₂=1.128 for n=2.

How to Read I-MR Charts

Process IN CONTROL (Stable)

All points fall within UCL and LCL on BOTH charts
No non-random patterns visible
Random scatter around center line
Only common cause variation present

Process OUT OF CONTROL (Unstable)

One or more points outside control limits
Non-random patterns detected (see detection rules below)
Special causes present — investigate immediately!

I-MR Chart Calculation Example

Example: Daily production output (kg)

Data: 100, 105, 98, 103, 101, 99, 104, 102, 100, 106

Step 1: Calculate individuals statistics

X̄ = (100+105+98+103+101+99+104+102+100+106)/10 = 101.8 kg

Step 2: Calculate moving ranges

MR₁ = |105-100| = 5

MR₂ = |98-105| = 7

MR₃ = |103-98| = 5, ... (continues)

MR̄ = (5+7+5+2+2+5+2+2+6)/9 = 4.0

Step 3: Calculate control limits

I Chart:

UCL = 101.8 + 2.66 × 4.0 = 112.4 kg

CL = 101.8 kg

LCL = 101.8 - 2.66 × 4.0 = 91.2 kg

MR Chart:

UCL = 3.27 × 4.0 = 13.1

CL = 4.0

LCL = 0

Result: All points (100, 105, 98, 103, 101, 99, 104, 102, 100, 106) fall between 91.2 and 112.4 kg. All moving ranges fall below 13.1. Process is IN CONTROL!

I-MR Chart Visual Example

The I Chart and MR Chart for a diameter measurement process show all points remain within control limits (UCL and LCL) with no detection rules triggered — indicating a stable, in-control process.

I Chart (Individuals Chart): Monitors process mean. Center Line (CL) = 101.80, Upper Control Limit (UCL) = 112.44, Lower Control Limit (LCL) = 91.16. All measurements fall within ±3σ limits.

MR Chart (Moving Range Chart): Monitors process variation. Center Line (MR̄) = 4.00, Upper Control Limit (UCL) = 13.07. All moving ranges fall below UCL, indicating stable variation.

Control Chart Detection Rules (Western Electric Rules)

Beyond just "point outside control limits," there are 8 detection rules (also called Western Electric Rules or Nelson Rules) that identify non-random patterns indicating special causes.

Rule 1: One Point Beyond 3σ

Pattern: Any single point falls outside the control limits (UCL or LCL).

Indicates: Sudden process shift, measurement error, data entry error, or special cause event.

Action: Investigate immediately. Check for equipment malfunction, material change, or operator error.

Rule 2: Nine Points in a Row on Same Side of Center Line

Pattern: 9 consecutive points all above OR all below the center line.

Indicates: Process mean has shifted. New material, tool wear, calibration drift.

Action: Find cause of shift. Recalibrate equipment or adjust process settings.

Rule 3: Six Points in a Row Steadily Increasing or Decreasing

Pattern: 6 consecutive points forming an upward or downward trend.

Indicates: Gradual change — tool wear, temperature drift, operator fatigue.

Action: Identify trending cause. Schedule maintenance or process adjustment.

Rule 4: Fourteen Points in a Row Alternating Up and Down

Pattern: Points alternate direction in a sawtooth pattern (up-down-up-down).

Indicates: Overcontrol (tampering), alternating materials, or two alternating operators.

Action: Stop unnecessary adjustments. Standardize procedures.

Rule 5: Two Out of Three Points Beyond 2σ (Same Side)

Pattern: 2 of 3 consecutive points fall in zone A (beyond 2σ but within 3σ) on the same side.

Indicates: Early warning of process shift. Increased variation.

Action: Monitor closely. Check for emerging issues.

Rule 6: Four Out of Five Points Beyond 1σ (Same Side)

Pattern: 4 of 5 consecutive points fall in zone B or beyond (beyond 1σ) on the same side.

Indicates: Process mean shifting or increased variation.

Action: Investigate potential causes before out-of-control occurs.

Rule 7: Fifteen Points in a Row Within 1σ of Center Line

Pattern: 15 consecutive points fall in zone C (within 1σ of center on either side).

Indicates: Reduced variation (stratification). May indicate data fabrication or incorrect subgrouping.

Action: Verify data collection method. Check for mixing sources.

Rule 8: Eight Points in a Row Beyond 1σ (Either Side)

Pattern: 8 consecutive points, none in zone C (all beyond 1σ from center).

Indicates: Bimodal distribution. Two different sources mixed.

Action: Separate data by source. Create separate control charts.

Control Chart Zones

Zone A: Between 2σ and 3σ from center line
Zone B: Between 1σ and 2σ from center line
Zone C: Within 1σ of center line

Other Types of Control Charts

For Variable (Continuous) Data

X̄-R Chart (Average & Range)

When to use: Subgroups of 2-4 samples collected regularly

Charts: X̄ (subgroup averages) + R (subgroup ranges: Max - Min of each subgroup)

Example: Measure 4 parts every hour from production line

X̄-S Chart (Average & Standard Deviation)

When to use: Subgroups larger than 4 samples

Charts: X̄ (subgroup averages) + S (subgroup standard deviations)

Advantage: More sensitive to variation changes than range

For Attribute (Count/Proportion) Data

P Chart (Proportion Defective)

Measures: Proportion of defective items

Data: p = (# defectives) / (# inspected)

Example: % of invoices with errors

NP Chart (Number Defective)

Measures: Count of defective items (constant sample size)

Data: np = # of defectives in sample

Example: Number of defective parts in batch of 100

C Chart (Count of Defects)

Measures: Number of defects per unit (constant area of opportunity)

Data: c = # of defects per unit

Example: Scratches per car panel

U Chart (Defects per Unit)

Measures: Defects per unit (variable area of opportunity)

Data: u = (# defects) / (# units)

Example: Defects per square meter of fabric

How to Choose the Right Control Chart

Step 1: Is your data continuous (measurements) or attribute (counts/proportions)?

Step 2 (Continuous): Do you have subgroups or individual measurements?

Step 2 (Attribute): Are you counting defective units or defects per unit?

Step 3: Is your sample size constant or variable?

How to Implement Control Charts (Step-by-Step)

Step 1: Choose the Characteristic to Monitor

Select critical-to-quality (CTQ) characteristics. Focus on process outputs that impact customer satisfaction or business goals.

Step 2: Select the Appropriate Control Chart Type

Based on data type (continuous vs attribute) and sampling method (individuals vs subgroups). Use decision tree above.

Step 3: Collect Baseline Data (25-30 Subgroups Minimum)

Gather initial data when process is believed to be stable. For I-MR charts, collect 25-30 individual measurements. For X̄-R charts and X̄-S charts, collect 25-30 subgroups.

Important: Do NOT calculate control limits from unstable data!

Step 4: Calculate Control Limits

Use appropriate formulas for your chart type. Control limits are calculated at ±3σ from the center line. Do NOT confuse with specification limits!

Step 5: Plot the Data and Examine for Control

Plot baseline data on the chart. Check for out-of-control signals using the 8 detection rules.

If out of control: Investigate and remove special causes, then recalculate limits.

Step 6: Continue Monitoring and Responding

Plot new data points as they're collected. When signals appear:

Special cause detected: Investigate root cause immediately. Fix problem.
Process improved: Recalculate control limits with new stable data.
Process changed: Create new control chart if fundamental change occurs.

Step 7: Recalculate Limits Periodically

After process improvements or when significant time has passed, recalculate control limits using recent stable data (last 25-30 subgroups or individuals).

Common Mistakes to Avoid

Confusing Control Limits with Specification Limits

Mistake: Using customer specification limits (USL/LSL) as control limits.

Fix: Control limits are calculated from process data (±3σ). Spec limits come from customers. They are COMPLETELY different!

Calculating Limits from Unstable Data

Mistake: Including out-of-control points when calculating control limits.

Fix: Remove special causes first, then recalculate limits using only stable data.

Tampering (Overadjusting the Process)

Mistake: Adjusting process every time a point moves, even when within control limits.

Fix: Only react to out-of-control signals. Random variation is NORMAL — don't tamper!

Insufficient Data for Baseline

Mistake: Calculating control limits with fewer than 25-30 data points.

Fix: Collect minimum 25-30 subgroups (or 25+ individuals for I-MR) before calculating limits.

Never Updating Control Limits

Mistake: Using the same control limits forever, even after process improvements.

Fix: Recalculate limits after verified improvements or major process changes.

Ignoring the Range/MR Chart

Mistake: Only looking at the individuals/averages chart, ignoring variation chart.

Fix: ALWAYS check BOTH charts. Increased variation can precede mean shifts.

Automated Control Charts with AI

DMAIC Suite™ automates control chart creation, detection rule checking, and AI-powered root cause analysis.

Automated SPC

Auto-select correct chart type (I-MR, X̄-R, X̄-S for continuous data, P, NP, C, U for attribute data)
Automatic control limit calculation
Real-time detection of out-of-control data points
Highlight out-of-control points

AI Interpretation

AI identifies which detection rules triggered
Suggests likely root causes for patterns
Recommends corrective actions
Tracks pattern history over time