Learn · updated 2026-10-28
Control Limits vs Specification Limits
Control limits are calculated from your process data and show whether it is stable; specification limits are set by the customer or the drawing and show whether a part is acceptable.
They look alike on a chart, two lines above and below the data, and that is where the confusion starts. They answer different questions and you act on them differently.
| Control limits | Specification limits | |
|---|---|---|
| Set by | Your process data (calculated) | The customer or the drawing (given) |
| Question | Is the process stable? Did something change? | Is this part acceptable? |
| Typical position | Mean ± 3σ of the plotted statistic | Wherever the design needs them |
| When a point is outside | Look for a special cause; something changed | The part is nonconforming; contain it |
| Also called | Voice of the process (UCL, LCL) | Voice of the customer (USL, LSL, tolerance) |
Both on One Chart
LimitLine's built-in example: NIST's ten flow-rate readings plus five more, with spec limits 45 to 57. The control limits come out of the data: center 51.34, UCL 58.71, LCL 43.97.
Notice the control limits are wider than the spec here (43.97 to 58.71 against 45 to 57). That is point 13's doing: one special cause inflates the moving ranges and so the limits. Find its cause, leave it out and recalculate, and the limits come in to 46.43 to 55.02, inside the spec, with Cpk 1.33. Had they stayed wider than the spec, a stable process would still make some parts out of spec: a capability problem, fixed by changing the process rather than by chasing points.
The Four Cases
| Inside the spec | Outside the spec | |
|---|---|---|
| In control | Good: keep it there, keep charting | Stable but not capable: change the process (spread or centering) |
| Out of control | Something changed: find it before it ships bad parts | Contain the parts and find the special cause |
The Mistake to Avoid: Spec Limits on an X̄ Chart
An X̄ chart plots averages of subgroups. Averages vary less than single parts, so a subgroup average can sit comfortably inside the spec while individual parts in it are out. Compare spec limits with individual values (an I-MR chart, a histogram or the capability study), never with averages. For the same reason, don't tighten control limits to the spec "to be safe": the limits describe the process, they are not a target.
Sources
- NIST/SEMATECH e-Handbook of Statistical Methods, §6.3.1 What are control charts?
- NIST/SEMATECH e-Handbook, §6.1.6 What is Process Capability?
- AIAG (2005), Statistical Process Control (SPC) Reference Manual, 2nd edition, Automotive Industry Action Group
The LimitLine Newsletter
About one email a month: new guides, tool updates and one practical SPC tip. Unsubscribe in one click.