Definition
Statistical Process Control, or SPC, uses statistical methods to monitor process behavior and distinguish routine variation from signals requiring investigation. It helps operating teams detect changes and manage a process based on evidence rather than reacting to every fluctuation.
SPC is not simply a graph with upper and lower lines. The measurement method, chart type, data collection, limits, signal rules, and reaction plan must fit the process being monitored.
Control limits and specifications answer different questions
Specification limits describe the required product or service characteristics. Control limits describe the expected behavior of a plotted process statistic under the conditions used to establish the chart.
A measured part can be within specification while the process produces a statistical signal. Conversely, a stable process can consistently produce output that does not adequately meet the specification.
Confusing these two questions weakens both decisions. Product acceptance requires the applicable requirements, while process investigation requires an appropriate interpretation of process behavior.
Match the chart to the data
An individuals chart monitors individual measurements; subgroup charts may monitor averages and associated variation. Attribute charts address data such as nonconforming-unit counts or defect counts under their applicable assumptions.
Do not plot individual readings against limits calculated for subgroup averages. Likewise, do not treat the number of defects as interchangeable with the number of defective units when one unit can contain several defects.
Sampling should capture the variation relevant to the operating decision. Mixing distinct machines, cavities, product families, or operating conditions without a justified design can hide important differences or create misleading signals.
A hypothetical in-spec signal
Assume an appropriate prior study established an individuals chart for the precision component's bore diameter. For this illustration, the center line is 10.000 millimeters and the control limits are 9.970 and 10.030 millimeters.
The product specification is broader: 9.900 to 10.100 millimeters. These control limits are hypothetical established inputs, not values calculated from the few observations below.
| Sequential observation | Individual diameter | Within product specification? | Beyond an individuals control limit? |
|---|---|---|---|
| A | 9.998 mm | Yes | No |
| B | 10.006 mm | Yes | No |
| C | 10.004 mm | Yes | No |
| D | 10.040 mm | Yes | Yes, above 10.030 mm |
Observation D is inside the product's dimensional limits but exceeds the illustrated upper control limit. It warrants the defined investigation and reaction rather than dismissal because the part is technically in specification.
The signal does not identify the cause by itself. A setup change, tool condition, material difference, measurement issue, or another factor must be evaluated using evidence.
Use a defined reaction plan
The operating instructions should explain who responds, what product may be affected, what records to preserve, and when production or release requires additional authorization. The response should reflect the process and product risk rather than an improvised rule.
Verify the measurement and investigate relevant changes without deleting inconvenient observations. Where product could be affected, apply the appropriate containment and disposition process while the technical investigation proceeds.
Do not repeatedly adjust the machine in response to ordinary variation. Unnecessary intervention can make performance less predictable and obscure whether the original process actually changed.
Stability is not the same as capability
Process capability evaluates how a stable process relates to its specification, using suitable methods and assumptions. A capability index should not be treated as a reliable summary of future performance when the process is unstable or the measurement system is inadequate.
Sample selection, data distribution, process centering, and estimation method matter. Do not present one capability threshold as a universal acceptance rule across customers, products, and study types.
The four observations above cannot demonstrate capability or establish stable long-term behavior. They illustrate one signal against an assumed, previously established control scheme.
Preserve the meaning of the limits
Document the reference data, chart construction, signal rules, and reasons for any later revision. Control limits should not be widened merely to make a problem disappear from a dashboard.
An appropriate process change may justify a new baseline after technical review and sufficient evidence. Keep the earlier history so the organization can explain what changed rather than overwriting the record.
Being inside the limits is also not the only consideration. Depending on the adopted chart rules, sustained patterns or trends can require investigation even when individual points do not cross a limit.
Frequently asked questions
- Does SPC replace inspection? No. It addresses process behavior, while product acceptance and other verification activities follow the applicable control plan and requirements.
- Can software choose the chart automatically? Software can assist, but a qualified person should confirm the data type, assumptions, sampling, and reaction rules.
- Does one signal prove defective product? No. It indicates a reason to investigate process behavior; product conformity and disposition require their own evidence.
Use supplier data to ask better operating questions
VPP helps procurement teams distinguish a credible process-control discussion from a favorable-looking supplier metric. The focus is whether the evidence supports stable execution, timely reactions, and the supply performance the business requires.