SOP: Control Chart Program, Rational Subgrouping, and Out-of-Control Response

This is a ready-to-use SOP for the day-to-day operation of a control chart program: choosing the chart, designing subgroups, and responding when a rule fires. Replace every <<FILL: ...>> placeholder with your own specifics, set your document numbers and dates, and route it through your normal document control, review, and approval. A worked filled specimen follows the template. Verify each cited regulation against the current source before you rely on it. This SOP assumes control limits and signal rules for a given parameter are already established per the Worksheet: Control Limit Establishment and Signal Rule Configuration; it governs what happens with the chart afterward, day to day.

Document control header

1. Purpose

This procedure defines how <<FILL: COMPANY NAME>> selects the correct control chart for a given characteristic, designs subgroups so the chart can actually see the variation it is meant to catch, and responds when a signal fires, so that special-cause variation is found and acted on and common-cause variation is not chased as though it were a problem.

2. Scope

This procedure applies to Shewhart control charts (and, where selected, EWMA/CUSUM charts) used for in-process monitoring, continued process verification, environmental and utility monitoring, and any other GxP parameter tracked over time for statistical control. It does not set the numeric control limits or select the signal rules for a given parameter; that is governed by <<FILL: SOP-ID or the control limit establishment worksheet>>. It does not cover process capability calculation, which is governed by <<FILL: SOP-ID or the capability study protocol>>.

3. Responsibilities

4. Definitions

• Rational subgroup: a set of measurements collected so that variation within the subgroup reflects only common cause, and variation between subgroups is where the special cause of interest would show up, if present.
• Signal: any confirmed rule violation per the locked rule set for the chart (a point beyond a control limit, a run, a trend, or another configured pattern).
• Special cause: an assignable, external source of variation (a raw material change, an equipment fault, an operator or shift change, an instrument drift).
• Chart annotation: a documented note on the chart itself, or in its associated record, marking an excluded point, an investigated signal, or a process change, so the chart’s history remains interpretable later.

5. Procedure

5.1 Select the chart type

1. Determine the data type: continuous (a measurement) or attribute (a count or a proportion).
2. For continuous data, determine whether subgroups of two or more are practical to collect each period. If yes, use X-bar/R (subgroup size under about 10) or X-bar/S (10 or more). If only one result is available per batch or period, as is typical for release and in-process test results in batch pharma and biologics manufacturing, use I-MR.
3. For attribute data, determine what is being counted: defective units against a fixed inspection opportunity (use np if the sample size is constant, p if it varies), or defects rather than defective units (use c if the area of opportunity is constant, u if it varies).
4. Where detecting a small, sustained shift matters more than reacting quickly to a large one, for example a slowly degrading resin or a creeping fill setpoint, add an EWMA or CUSUM chart alongside the Shewhart chart rather than replacing it; the two chart families answer different questions.
5. Record the chart-type decision and its rationale in the parameter’s monitoring record before the chart goes live.

5.2 Design the rational subgroup

This is the step most often done by convenience rather than by design, and it determines whether the chart can see anything at all.

1. Name the specific source of variation this chart is meant to catch (a shift-to-shift difference, a head-to-head difference on a multi-head filler, a lot-to-lot raw material difference).
2. Compose subgroups so that source of variation falls between subgroups, not within them. Sampling five consecutive units from one filling head tells you about that head’s own repeatability and hides head-to-head differences; sampling one unit from each of five heads does the opposite.
3. State explicitly what the subgroup composition holds constant and what it lets vary, and confirm that matches the source of variation named in step 1.
4. Where more than one source of variation matters (for example both shift and raw material lot), either run separate charts for each, or document clearly which source the shared chart is designed to catch and accept that it may be blind to the other.
5. Record the subgrouping rationale in the parameter’s monitoring record. Where a more detailed record is warranted, use the Form: Rational Subgrouping and Sampling Plan.

5.3 Operate the chart

1. Plot each new result as it becomes available, against the locked centerline and control limits.
2. Apply the locked signal rule set (see the parameter’s control limit establishment worksheet) at each new point.
3. Never plot specification limits on the chart. Control limits and specification limits answer different questions, and drawing both invites operators to adjust the process toward spec, which adds variation.
4. Review the chart on the defined cadence (<<FILL: e.g. each batch, weekly, monthly>>) even when no signal has fired, and record that the review occurred.

5.4 Respond to a signal, by severity

A signal obliges a response; the depth of the response scales with what fired and with how much margin the process has to specification, not with convenience.

Where the parameter’s practical-significance threshold (the shift size that would actually matter given the process’s margin to specification) has been defined, use it to set the depth and urgency of the investigation, not whether an investigation happens at all; a signal that fires is always assessed.

5.5 Annotate and, where needed, requalify the chart

1. Annotate the chart at the point of every confirmed special cause, with a reference to the investigation or deviation record.
2. Do not remove a confirmed special-cause point from the working chart’s visible history; exclude it only from a recalculation of the limits, and only with a documented reason, per the recalculation trigger defined in the parameter’s control limit establishment worksheet.
3. After a confirmed, corrected special cause, reassess whether the chart requires a new baseline period before it can again be relied on for monitoring, and record that decision.

6. Acceptance criteria

• Every monitored parameter has a documented chart-type decision with its rationale.
• Every monitored parameter has a documented subgrouping rationale naming the source of variation the chart is designed to catch.
• No control chart in the program has specification limits plotted on it.
• Every signal has a documented, timely response, including a documented basis for any decision not to escalate.
• Confirmed special-cause points are annotated on the chart and excluded from limit recalculation only with a documented reason.

7. Records generated

• Chart-type selection and rationale, per parameter.
• Subgrouping rationale, per parameter (or the completed rational subgrouping form).
• Periodic chart review record.
• Signal assessment and response record, including any deviation or investigation reference.
• Chart annotation history.

8. References

Shewhart, W. A., the original statistical control framework underlying all charts in this procedure. Western Electric, Statistical Quality Control Handbook (1956); Nelson, L. S., “The Shewhart Control Chart, Tests for Special Causes,” Journal of Quality Technology (1984). FDA, Process Validation: General Principles and Practices (2011), Stage 3. ICH Q9(R1), Quality Risk Management. See statistics in quality: Cpk and control charts for the underlying statistical reasoning and worked examples.

Confirm the current version of each reference before issue.

9. Revision history

10. Approvals

Filled specimen

Parameter: headspace oxygen on a lyophilized biologic fill line, one result per batch, I-MR chart, monitored under the site CPV program.

5.1, chart type: Continuous data, one result per batch (at-line analyzer, one reading per fill run). I-MR selected; subgrouped charting is not practical because the analyzer measures one composite reading per batch, not per unit. Decision recorded 3 February 2026.

5.2, subgrouping: The source of variation of interest is a slow drift in stopper seating over the tooling’s wear cycle, which shows up batch to batch, not within a batch (there is only one reading per batch). Because n = 1, “subgrouping” here means confirming that batches are plotted in true production sequence, with no reordering, so a real drift is visible as a trend rather than being scattered by an out-of-sequence data pull.

5.4, response: In week 14, four consecutive batches plotted above the centerline (a run, not yet a limit breach). The process owner opened an assessment within 2 business days per the run-rule response tier, traced the pattern to a partially worn stopper-seating tool nearing its qualified cycle count, and raised a deviation. The tooling was replaced and a wear-based preventive maintenance check was added. Six subsequent batches returned to the historical centerline, and the deviation was closed with that verification attached.

5.5, annotation: The chart was annotated at week 14 with the deviation reference. The four elevated points were retained in the visible chart history; they were not used to recalculate the baseline limits, since the corrective action removed the special cause and the pre-existing limits remained the valid reference for ongoing monitoring.

Common inspection findings this SOP prevents

• A chart type chosen with no rationale, or chosen because it was the software’s default rather than because it fits the data.
• Subgroups composed for operator convenience that bury the exact source of variation the chart was meant to catch.
• Specification limits plotted on a control chart, inviting operators to adjust toward spec.
• A signal that fired with no documented response, or a response with no timeline.
• Confirmed special-cause points quietly dropped from a limit recalculation with no documented reason.

How to adapt this SOP

1. Set your document number, owner, and effective date in the header.
2. Point the cross-references in sections 2 and 5.4 to your real control-limit-establishment, deviation, and capability procedures.
3. Set the response timelines in section 5.4 to match your process risk and batch frequency.
4. If your program uses a validated chart-software configuration, name it and reference its own validation record, so the documented rule set and the deployed rule set cannot silently diverge.
5. Confirm every regulation in section 8 against the current published version before issue.