Scenario
The most common mistake process engineers make: assuming the process is fine just because no points exceed UCL/LCL on the control chart. Three days later, a batch of defective products appears, and the supervisor asks, "Why didn't the control chart provide a warning?"
The control chart did provide a warning; you just didn't know how to read it.
What are the Western Electric Rules
In 1956, Western Electric Company compiled 8 judgment rules in its "Statistical Quality Control Handbook," specifically designed to detect "non-random patterns" on control charts. Even if no points exceed the limits, these patterns indicate that the process has already started to have problems.
Plain Explanation of the Eight Rules
Rule 1: One point outside the 3σ limit
The most basic rule. Indicates a special cause has occurred; investigate immediately.
Rule 2: Nine consecutive points on the same side of the center line
Not random; the process average has shifted. Possible causes include a change in raw material batch, operator shift change, or slow machine drift.
Rule 3: Six consecutive points steadily increasing or decreasing
The process is undergoing a trend change. Common causes: tool wear, mold aging, decreasing raw material batch.
Rule 4: Fourteen consecutive points alternating up and down
Too regular, not random. Common causes: alternating production from two machines recorded together, or bias in readings from alternating inspectors.
Rule 5: Two out of three consecutive points fall between 2σ and 3σ (on the same side)
A warning signal approaching the control limit; process variation is increasing.
Rule 6: Four out of five consecutive points fall between 1σ and 3σ (on the same side)
The process has significantly shifted; although not yet out of limits, the trend is incorrect.
Rule 7: Fifteen consecutive points all within ±1σ
Sounds good? No. All points clustering in the middle may indicate data has been artificially filtered, or the measurement system resolution is insufficient.
Rule 8: Eight consecutive points on both sides of the center line, with none within ±1σ
The process may have two different distributions mixed together (bimodal distribution), e.g., data from two production lines recorded together.
Practical Priority
| Rule | Severity | Most Common Cause |
|---|---|---|
| Rule 1 | Highest | Sudden abnormality, equipment failure |
| Rule 2 | High | Process shift, material change |
| Rule 3 | High | Equipment aging, trend deterioration |
| Rule 5/6 | Medium | Gradual process shift |
| Rule 4 | Medium | Data mixing, measurement issues |
| Rule 7/8 | Medium | Data quality issues |
Golden Quote
"A control chart is not an alarm; it's a conversation—but you must first learn to understand what it's saying."