ORDREL / FREE TOOL
SPC Studio: control charts and Cpk
Control charts and capability, from your data.
How this works
What you’ll get · Paste your data, get the right control chart and see whether the process is stable and capable — in plain words.
Use the tools and save work for free. Data exports and reports require Plus or Teams.Compare plans →
How to use control charts and process capability
Statistical process control (SPC) separates the normal variation of a process from the signals that something changed. A control chart shows it over time; capability compares the process with what the customer allows.
Pick the right control chart
Measured values, one at a time (a weight, a time): the individuals and moving range chart (I-MR). Small samples taken together: X̄-R, or X̄-S from 10 values per sample. Defective items out of a sample: the p-chart, or np when the sample size never changes. Defects counted: the c-chart on the same amount of product each time, the u-chart when the amount varies.
Read the control chart
The centre line is the average. The control limits sit about 3 standard deviations either side and come from the process itself, not from the spec. A point beyond a limit, 9 in a row on one side, or 6 in a row going up or down are signals: something specific happened. Everything else is common-cause variation.
Act on signals, not on noise
When a point signals, find out what happened while it is fresh and write it down. Adjusting the process after every point that is merely high or low adds variation instead of removing it. Common-cause variation only goes down when the system changes: methods, equipment, materials.
Process capability: Cp, Cpk, Pp and Ppk
Cp compares the width of the spec with the spread of the process; Cpk also counts how far the average is from the nearer spec limit. Pp and Ppk do the same with the overall spread. Cpk 1.33 means the nearer limit is 4 standard deviations away — the usual minimum. Cpk below 1.0 means part of the output falls outside the spec.
Stable first, then capable
Capability only predicts the future when the process is in control. Remove the special causes first. Use at least 20–25 points (or 100 values) before you trust Cpk, and check that the data look normal: if not, trust the values seen outside the spec more than the expected ones.
Keep it in control
Write who plots the point, how often, and what happens at a signal — the reaction plan. After a deliberate change, start a new stage so the chart gets new limits. Keep control limits and spec limits apart: one describes the process, the other the customer.
Before you trust the numbers
Points in time order, measured the same way each time. At least 20 points for solid limits. No unexplained signals in the current stage. Spec limits from the customer or the drawing. Enough data for a narrow Cpk interval.