Rolling out statistical process control (SPC)
Rolling out statistical process control (SPC) puts control charts on key process variables so operators can tell genuine, assignable shifts apart from normal random variation and act only on the former. It replaces tampering — adjusting a process in response to ordinary noise — with disciplined response to real signals, stabilising quality and reducing variation.
What it is
SPC monitors a process variable over time on a control chart bounded by statistically derived limits that reflect the process's natural variation. A point outside the limits, or a non-random pattern within them, signals an assignable cause worth investigating. The rollout selects the variables to chart, establishes their control limits from data, trains operators to read the charts, and defines what to do when a signal appears.
Why it is done
Operators naturally adjust a process whenever a reading looks off, but reacting to normal random variation actually increases variation — a phenomenon called tampering. SPC distinguishes the noise to leave alone from the signals that demand action, so processes are adjusted only when something real has changed. The result is more stable output, less scrap, and the data foundation for capability analysis and continuous improvement.
How it is done
The critical variables that drive quality are selected, and a measurement system is checked for adequate precision before any charting, because a noisy gauge ruins the chart. Control limits are calculated from data collected while the process runs normally, not from the specification. Operators are trained to recognise out-of-control signals and patterns, and a documented response plan tells them what to investigate and adjust. Charts are reviewed, limits revised as the process improves, and capability tracked over time.
- Select critical variables
- Verify measurement system
- Collect baseline data
- Calculate control limits
- Train operators on signals
- Define response plan
What to watch for
Setting control limits from the specification instead of the process data destroys the chart's ability to detect real change. Charting without an agreed reaction plan leaves operators with pretty graphs and no idea what to do when a point falls out, so nothing improves.
Related practices
Reducing changeover time with SMED
Optimising clean-in-place (CIP)
Running a process capability study
Related topics
SPC (Statistical Process Control) · Six Sigma · Root Cause Analysis (RCA)
Common in: Pharmaceuticals · Chemicals · Food Processing · Steel & Metals · Paper & Packaging · Dairy