Read a Supplier Control Chart Before Asking for Process Changes

A control-chart point can cross a limit without identifying the cause, and a stable chart can describe a process that still produces nonconforming parts. Asking a supplier to adjust the machine after every movement confuses process signals with product specifications and can add variation instead of reducing it.

A useful supplier review starts by confirming what was charted, how the data were ordered and grouped, how the control limits were established, and which signal rules apply. It then connects a signal to source data and investigation before approving an action. Product behavior remains with the existing track-chain movement and fit context. This article does not calculate capability, validate the measurement system, accept a lot or authorize a process change.

Confirm what the chart tracks

Begin with the part number, drawing or specification revision, manufacturing process and exact characteristic. Record the unit and how the characteristic is defined. “Track-chain fit” is too broad for a chart; a controlled dimension, force, count or classified outcome needs a stable operational definition.

Identify the plotted statistic. It may be an individual measurement, a subgroup average, a range or other measure of within-subgroup spread, a count, or a proportion. The points on an X-bar chart represent subgroup averages rather than every raw part measurement. An individuals chart plots observations in sequence and answers a different monitoring question. Do not infer raw-part conformity from a point that summarizes several units.

Record the chart type and the method used to select it. The data structure matters: variable or attribute result, subgroup size, event frequency and opportunity for defects can all affect the appropriate chart. Do not accept a familiar chart layout as valid merely because it has a center line and two limits.

Document sample or subgroup size, sampling frequency and the reason the observations form a rational subgroup. Parts grouped together should represent the short-term process variation the method intends to compare. If one subgroup combines output from different machines, tools, materials or shifts, that mixing may hide or create signals. Record machine, cavity or tool, material lot, shift and other stratification fields required by the control plan.

Keep the actual production order. Sorting points by value, lot average or inspection result destroys the time relationship that allows a chart to detect change. Capture timestamps and links to the original measurements so a signal can be traced to process events. ASQ’s control-chart guidance describes the use of time-ordered data, center lines, control limits and signals to distinguish consistent from special-cause variation. It does not prescribe one chart or signal set for every supplier.

Confirm the measurement system and inspection method used to generate the data. A change in gauge, operator method, resolution or calibration status can look like a process change. This review records that measurement evidence and sends deficiencies to the appropriate measurement-system process; it does not perform a gauge R&R study.

Finally, identify the baseline dataset, approval period, chart calculation method and version. The chart should show which machines, materials, shifts and product revisions the baseline represents. If the supplier changed the baseline but retained the old version label, the review cannot reliably compare current signals with earlier action records.

Distinguish control limits from specifications

The center line and upper and lower control limits describe the behavior of the process data under the selected chart method and baseline. They are statistical signal boundaries, not engineering product limits. Record the baseline observations and calculation or software version used to establish them. If limits were entered manually, ask for the controlled basis.

Specification limits come from the drawing, product specification or approved requirement. The target is the desired value or condition defined by that requirement or process plan. These three concepts—center line, control limits and specifications—may appear on one chart, but their sources and meanings remain different.

The NIST/SEMATECH statistical glossary describes control limits as calculated from process data and used to indicate possible special causes. That is why copying upper and lower specification limits into the control-limit fields does not create a valid control chart. It merely redraws the product requirements.

A point can be inside the engineering specification but beyond a control limit. The individual part may meet the stated product limit while the chart signals that the process behavior changed. The response is to investigate the process signal and evaluate affected product under the relevant plan, not automatically reject the point or adjust the machine.

The reverse can also occur. A process can remain statistically stable, with points behaving consistently inside its control limits, while its output distribution overlaps or sits beyond a specification limit. Stability means the variation is predictable under the present system; it does not mean the system is capable of meeting product requirements. Capability analysis belongs in its separate evidence path.

Review every control-limit recalculation. A stable baseline and approved rule should define when a new center line or limits may be established. Recalculating immediately after an inconvenient point can erase the signal rather than explain it. Preserve the previous chart version, excluded data, rationale, authorization and effective interval.

Minitab’s control-chart basics illustrate time order, control limits and nonrandom or out-of-limit signals in that software context. The page does not make its defaults universal or decide product disposition. The supplier must identify the actual procedure and settings used for the submitted chart.

Interpret signals as prompts to investigate

A point beyond a control limit is one possible signal. Runs on one side of the center line, shifts, trends, cycles or other nonrandom patterns may also be signals when the approved rule set defines them. Record the complete rule set, version and any software options. Do not apply a remembered run length from another chart or customer.

Multiple rules and multiple charts increase the opportunity for false alarms. If the supplier monitors several related statistics and enables many tests, the review should record that context. A highlighted point proves only that the configured rule fired on the submitted data. It does not identify worn tooling, material change, operator error or any other cause.

Return to the timestamp and source data. Confirm that the plotted value matches the original measurement or count and that the subgroup membership is correct. Check for transcription, unit conversion, rounding, duplicate entry, missing observation, changed sample size and measurement-system issues. Correcting a data error is different from changing the process.

Then correlate the signal with process events: machine or tool change, maintenance, setup, material lot, shift, operator, environment, program revision or inspection method. Record evidence for and against each special-cause hypothesis. “Signal appeared after tool replacement” is a timing relationship; it does not prove the replacement caused the change without further evidence.

ASQ’s variation guidance distinguishes common causes inherent in the system from special causes associated with unusual circumstances and warns that confusing them can lead to tampering. A control-chart signal is therefore an investigation trigger. The responsible team must establish the cause before prescribing a process adjustment.

Define the potentially affected production interval from the last known stable evidence through the investigation boundary. Link parts, lots or timestamps and apply the authorized containment plan if product risk exists. Containment and product disposition are separate from the control-chart interpretation; the chart reviewer should not release or reject product from the signal alone.

Record the investigation owner, due date, immediate data checks and evidence needed. If the source data, rule set or baseline is absent, classify the signal as unresolved. Do not accept a narrative about the cause when the submitted chart cannot be reconstructed.

Choose action based on the cause evidence

When evidence confirms a special cause, record the affected interval, cause, temporary correction and longer-term action. Examples might include restoring a verified setup, replacing a damaged tool or correcting a changed input, but the authorized response depends on the process and engineering controls. Link the change approval and any required product containment or disposition.

When the evidence shows stable common-cause variation, repeated local adjustment is not a sound response. Improvement requires a deliberate change to the system, such as an authorized method, equipment or process redesign, followed by evaluation. Moving a setting up after one point and down after the next can increase variation because the operator is reacting to ordinary noise.

Preserve the pre-action chart, action record and effective time. Define the criteria for establishing a new baseline before recalculating limits. A changed process may require a monitored period, sufficient data and approval under the customer or supplier procedure. Do not blend pre-change and post-change points without marking the intervention.

Review the post-action chart against the stated objective. Confirm that source data, subgroup design, measurement method and signal rules remain comparable. An initial absence of signals does not by itself close recurrence risk, and a stable chart does not prove product capability. Assign the closure decision to the authorized owner.

Chart design and limits	Signal and source evidence	Cause, interval and action	Status and owner
Part, characteristic, statistic, subgroup plan, baseline and rule version agree	Configured rule fires; source data and timestamp reconcile with a recorded process event	Special cause confirmed, affected interval defined and authorized action documented	Normal: investigation and action evidence align; product disposition remains separate
Chart shows limits but baseline data and signal-rule version are absent	Supplier screenshot highlights one point; raw data are unavailable	Cause narrative cannot be tested and affected interval is unknown	Missing: supplier provides source data, baseline and rule record
Submitted chart uses one limit version; action report cites another	The same point is signaled on one chart and ordinary on the rebaselined chart	Process was adjusted before the limit change was authorized	Conflict: preserve both versions and assign controlled investigation
Stable chart design and limits; no documented special-cause signal	Operator adjusts the process after ordinary point-to-point movement	No cause evidence; repeated adjustments may be tampering	Unresolved: process owner stops unauthorized changes and reviews system variation

The signal-to-investigation record should retain the chart version, source data, rule set, process events, affected interval, action authorization and post-action chart. Unknown evidence remains open; a supplier explanation should not be promoted to a confirmed cause without a traceable basis.

The practical sequence is simple even when the analysis is not: verify the chart design, keep statistical control limits separate from product specifications, investigate configured signals, and change the process only through evidence and authorization. That discipline protects the supplier from needless adjustment and gives the buyer a clearer record of what actually changed.