Resources What's New Support Download Contact Sales: Arthur-Industrial@outlook.com
Home › Resources › Rational subgrouping

Rational subgrouping

How to group readings so the chart separates short-term noise from real change.

Rational subgrouping is the most important and most overlooked decision in SPC. A subgroup should contain only common cause variation, so that any difference between subgroups shows up as a signal.

Rules of thumb

  • Take the units in a subgroup close together in time, from the same machine, cavity, fixture and setup.
  • Let time pass between subgroups, so changes can appear between them.
  • Typical subgroup size: 4 or 5 consecutive parts.
  • Do not mix streams in one subgroup (two machines, four cavities). Chart them separately or use a different chart.

What goes wrong

  • Mixing streams within subgroups: the within-subgroup range includes machine-to-machine differences, limits become too wide, and the chart hugs the center line (rule 7 patterns).
  • Subgroups spread over a long time: within-subgroup variation includes drift; real shifts are hidden.
  • Subgroups of one taken too often: consecutive readings are nearly identical (autocorrelation), moving ranges are tiny, and limits are too tight.

Frequency

Sample more often when the process is new, after changes, or when it has been unstable. Reduce frequency once it has been stable and capable for a long time.

Questions and answers

We have 8 cavities. Do I need 8 charts?

Not always. Options include a chart per cavity for critical features, a group chart that plots each cavity, or a chart of cavity averages with a separate check on cavity-to-cavity differences.

The chart has almost every point inside zone C. Is that good?

It looks good but usually means the limits are too wide, often because of mixed streams within subgroups. Review the subgrouping.

← All articles