Define Control-Chart Subgroups Around the Process Variation You Need to See

A control chart depends on how measurements are grouped. Boards from one immediate process condition and boards collected across a shift do not describe the same variation. Define the subgroup around the monitoring objective before calculating limits or interpreting a signal.

NIST’s subgroup analysis example shows that different grouping levels lead to different control-chart views. The example is not a universal recipe for PCB production; it illustrates why the hierarchy of observations matters to the analysis.

State the change of interest

Decide whether the chart should detect changes between consecutive production periods, panel positions, fixtures, or another defined process level. Identify the measurement unit and the time or physical relationship among observations in one subgroup.

Keep the grouping rule stable during ordinary monitoring. A subgroup that sometimes contains consecutive boards and sometimes contains one board from each shift changes meaning from point to point. Record unavoidable departures so that a chart signal can be interpreted correctly. Preserve subgroup membership in the exported data instead of requiring an analyst to infer it from row order.

Preserve the grouping fields

Plot individual data during setup as well as subgroup summaries. Averaging can hide alternating behavior or opposite shifts in separate fixture positions. If two distinct processes are being mixed, separate charts or a model that represents the structure may be more informative.

Validate the chart construction

1. Write the operational subgroup definition.
2. Check that collected records follow it.
3. Select a chart appropriate to the data and objective.
4. Review baseline conditions and known process changes.
5. Document the monitoring and response procedure.

Do not use arbitrary spreadsheet row blocks as subgroups simply because they produce a convenient chart. Sorting by measured value before grouping destroys the original production sequence and can create an artificial impression of stable within-group behavior.

When subgroup size varies, use a method that handles the actual sizes or explain the chosen restriction. Reusing limits calculated for one size without checking applicability can change the false-alarm behavior. Keep sample-size handling visible in the analysis configuration.

A useful chart is tied to a repeatable collection process and an actionable interpretation. Its limits are not product specifications, and its subgroup definition should be part of the released monitoring method. The control-limit review explains that complementary distinction.