Nelson rules explained with examples
The eight Nelson rules for control charts: what pattern each one detects, what usually causes it and which rules to switch on for your process.
Updated 2026-10-24
A control chart separates common cause variation (the noise every process has) from special causes (something changed). Points outside the 3 sigma limits are the obvious signal, but patterns inside the limits can also show a special cause. Lloyd Nelson published eight tests in 1984, extending the earlier Western Electric rules.
The eight rules
| Rule | Pattern | Typical causes |
|---|---|---|
| 1 | One point beyond 3 sigma | A single large special cause: wrong setting, broken tool, measurement error, wrong material. |
| 2 | 9 points in a row on the same side of the centre line | A sustained shift of the mean: new batch, different operator, adjusted machine. |
| 3 | 6 points in a row steadily increasing or decreasing | A trend: tool wear, warm-up, temperature change, consumable running out. |
| 4 | 14 points in a row alternating up and down | Over-adjustment by the operator, or two alternating sources such as two spindles or fixtures. |
| 5 | 2 of 3 points beyond 2 sigma on the same side | A medium shift, earlier warning than rule 1. |
| 6 | 4 of 5 points beyond 1 sigma on the same side | A small shift of the mean. |
| 7 | 15 points in a row within 1 sigma | Stratification: limits calculated from mixed data, or subgroups that combine several streams. Also a gauge with too coarse resolution. |
| 8 | 8 points in a row beyond 1 sigma on either side | A mixture of two processes, for example two machines feeding one chart. |
Which rules to use
Every extra rule catches more real problems but also produces more false alarms. With all eight rules active, a perfectly stable process gives a false signal much more often than with rule 1 alone. A practical choice:
- Rule 1 always.
- Rules 2 and 3 for most processes: they catch shifts and trends early.
- Rules 5 and 6 where small shifts matter, or use an EWMA or CUSUM chart instead.
- Rules 4, 7 and 8 when setting up a chart or investigating: they point to over-adjustment, stratification and mixtures.
Rules on the range chart
The rules assume a roughly normal, symmetric distribution, which the range and standard deviation charts do not have. That is why the tool applies only rule 1 (points beyond the limits) to the R, S and MR charts.
Try it
In the process capability tool switch the rules on and off in the sidebar. Flagged points turn red and are listed below the chart with the rule that fired. Click a point to exclude it once the cause is known and removed, and add a note for the report. For pass / fail data use the attribute control charts.