- [1] A single point outside the 3SD control limits {identifies sudden changes in the measure} ●─●─●─●
- [2] Nine consecutive points on the same side of the centerline {identifies small, sustained changes, such as a small improvement in the process} ●─●─●─●
- [3] Six consecutive points, including endpoints, constantly ascending or descending {detects a small, consistent drift in a process (trend)} ●─●─●─●
- [4] Two out of three consecutive points near (outer one-third) the 3SD control limit {adds additional sensitivity to detect changes that have not yet triggered [1] or [2]} ●─●─●─●
- [5] Fifteen consecutive points close (inner one-third of the chart) to the centerline {especially useful in detecting a reduction of variation with an I chart, or for detecting improper subgrouping with an x-bar chart} ●─●─●─●
Only [1] is applied when using:
… time-ordered data where the purpose is to judge the process for control
… rational subgroups
For improvement activities, these five rules can be used to provide evidence of an improvement in a process or system. These rules also provide evidence of "losing" gains previously made in a system.
Failure to meet the guidelines for minimum subgroup size (sample too small) is shown by an "X" at the data point. ●─x─●─●
- a yellow cross ( X ) for a sample (denominator) less than 1/CL. n should be at least 1/CL. If not, ignore the subgroup or combine data, or use Poisson tables and probability limits.
- an orange cross ( X ) for a sample (denominator) less than 4/CL. It is recommended that subgroups not be smaller than 4/CL. If the subgroup sizes are not large enough, combine subgroups. If a point falls above the upper control limit when the subgroup size is greater than 1/CL but less than 4/CL, adjustment procedures are needed.
- an orange-red cross ( X ) for a sample (denominator) less than 5/CL. Some authorities recommend n at least 5/CL as the sole criterion, whether or not points are out of control.
- If the subgroup size is sufficiently large, consider using smaller subgroups with time-ordered data to get an earlier indication when special causes enter the process.
References
- Hart MK, Hart RF. Statistical process control for health care. 2000 www.amazon.com