Statistical Process Control uses control charts to monitor process output over time and distinguish between normal variation (common cause — inherent to the process) and abnormal variation (special cause — something changed). When a special cause is detected, the process is investigated and corrected before it produces defects. SPC is a prevention tool, not an inspection tool.

What is the difference between control limits and specification limits?

Specification limits are set by engineering — they define what the customer will accept. Control limits are calculated from process data — they define what the process actually produces. A process can be in statistical control (all points within control limits) but still produce defects (if control limits are wider than spec limits). Conversely, a capable process has control limits well inside spec limits.

What are the Western Electric rules?

Rules for detecting special cause variation beyond the basic 'point outside control limits': (1) One point beyond 3σ. (2) Two of three consecutive points beyond 2σ on the same side. (3) Four of five consecutive points beyond 1σ on the same side. (4) Eight consecutive points on one side of the center line. Each rule detects a different type of process shift — sudden jumps, gradual drift, or increased variability.

Statistical Process Control: Detect Drift Before Defects

New to this topic?

We recommend reading these guides first to get the most out of this one:

Root Cause Analysis Toolkit A3 Problem Solving

±3σ

Control Limits

Variation Types

X̄-R

Most Common Chart

25+

Subgroups to Start

Two Types of Variation

Every process has variation. The critical question is: what kind?

Common Cause (Normal)

Inherent to the process design
Random, unpredictable in individual instances but predictable as a distribution
Always present — cannot be eliminated without changing the process
Examples: normal tool wear, ambient temperature fluctuation, material lot variation within spec
Action: Improve the process (system change), not investigate individual points

Special Cause (Abnormal)

Caused by something that changed — a specific, identifiable event
Not part of the normal process — it should not be there
Can be identified and removed
Examples: worn bearing, wrong material loaded, new operator without training, fixture out of alignment
Action: Find and eliminate the specific cause (investigation)

⚠️ The Cardinal Sin: Tampering

Tampering is reacting to common cause variation as if it were special cause — adjusting the process in response to random fluctuation. Example: the last 3 parts measured 0.502, 0.504, 0.501 against a 0.500 target. The operator adjusts the machine offset “to center it.” But those measurements are within normal variation — the adjustment itself introduces a special cause, making the process worse. The control chart prevents tampering by showing when variation is normal and when it is not.

The X̄-R Control Chart

The most common SPC chart for variable data (measurements). It plots the mean (X̄) and range (R) of small samples (subgroups) taken at regular intervals.

Step 1: Collect Data in Rational Subgroups

Take samples of 3–5 consecutive parts at regular intervals (e.g., every 30 minutes or every 25th part). “Rational” means within-subgroup variation represents common cause only. Do not mix parts from different machines, operators, or material lots in one subgroup.

Step 2: Calculate Subgroup Statistics

For each subgroup: X̄ = mean of measurements. R = max – min (range). Collect at least 25 subgroups before calculating control limits.

Step 3: Calculate Control Limits

X̄ chart: UCL = X̄̄ + A₂R̄, LCL = X̄̄ – A₂R̄. R chart: UCL = D₄R̄, LCL = D₃R̄. (A₂, D₃, D₄ are constants based on subgroup size — lookup in any SPC reference table.) These limits represent ±3σ from the process mean.

Step 4: Plot and Monitor

Plot each new subgroup’s X̄ and R on the chart. Check for special cause signals using the Western Electric rules. If a signal fires, investigate immediately — do not wait for the part to go out of spec.

📊 Worked Example: Hole Diameter on Wing Rib X̄-R Chart

Spec: 0.2500 ± 0.0010 in (LSL = 0.2490, USL = 0.2510). Subgroup size n = 5.

After 25 subgroups: X̄̄ = 0.2501, R̄ = 0.0006.

Constants for n=5: A₂ = 0.577, D₃ = 0, D₄ = 2.114.

X̄ Chart: UCL = 0.2501 + 0.577 × 0.0006 = 0.25045. LCL = 0.2501 – 0.577 × 0.0006 = 0.24975. CL = 0.2501.

R Chart: UCL = 2.114 × 0.0006 = 0.00127. LCL = 0. CL = 0.0006.

Interpretation: Control limits (0.24975–0.25045) are well within spec limits (0.2490–0.2510). The process is both in control and capable. A point at 0.25050 would signal special cause variation (investigation needed) even though it is still within spec — SPC detects drift before defects occur.

Reaction Plans

A control chart without a reaction plan is a decoration. Every monitored characteristic needs a defined response:

Signal	Response	Authority
Point outside control limits	Stop production, investigate, identify and remove special cause before resuming	Operator + team lead
Western Electric rule violation	Alert team lead, investigate trend/pattern, check for process change	Team lead + quality
Process trending toward limit	Proactive investigation — check for tool wear, material change, drift	Operator awareness

🎯 The Bottom Line

SPC is a prevention system that detects process drift before it creates defects. Control charts distinguish common cause (inherent, normal) from special cause (abnormal, investigate). The X̄-R chart is the workhorse for variable data. Rational subgroups, correct control limit calculation, and defined reaction plans make the system work. The goal: a process that is both in statistical control (stable) and capable (within spec with margin). Next: FMEA for Practitioners — anticipating failures before they occur instead of reacting after.

Keep Learning

📖

Guide

FMEA for Practitioners

🏭

Free Process Modeler