Calculate upper and lower control limits (UCL/LCL) for statistical process control charts using the ±k sigma method.
Last updated: March 2026
Center line value
Process variation
Multiplier (typically 3)
Control limits define the boundaries of normal process variation. The Upper Control Limit (UCL) and Lower Control Limit (LCL) mark ±k standard deviations from the process mean. These limits help identify when a process is "out of control"—experiencing special cause variation rather than normal random fluctuation.
Formula: UCL = x̄ + k × σ and LCL = x̄ − k × σ, where x̄ is the process mean, σ is standard deviation, and k is the sigma multiplier (typically 3). When process measurements fall outside these limits, it signals a problem requiring investigation.
Control limits are different from specification limits. Spec limits are customer requirements or tolerances. A process can be "in control" (within control limits) but still produce out-of-spec products if the process isn't capable enough. Control limits are based on actual process performance; specs are based on customer needs.
For a normal distribution, approximately 99.73% of data falls within ±3σ. So points outside these limits are rare and highly suspicious, signaling special causes. Some industries use ±2σ (95.45% coverage) for tighter control or ±1σ for early warnings.
Beverage Bottling: Bottle Weight Control
3σ is industry standard (covers 99.73% of normal data). Signals are rare and reliable. 2σ (95.45%) is stricter for sensitive processes. 1σ is too loose. Choose based on risk tolerance and false-alarm cost.
A point outside control limits indicates special cause variation—assignable causes like equipment failure, material change, operator error. Unlike random variation, these causes must be identified and fixed.
Recalculate when process parameters change (equipment upgrade, new procedure, material supplier). Typically quarterly or after major improvements. Don't recalculate after every few measurements—use for trend monitoring.
±3σ assumes approximate normality. For highly skewed data, use Box-Cox transformation or robust methods. Percentile-based limits work for any distribution but are less sensitive to small shifts.
Specifications = customer requirements (what should be produced). Control limits = process capability (what is actually produced). You might be in control but out of spec if process isn't capable enough.
8+ consecutive points on same side of center line, trends, clustering, or cyclical patterns all signal problems. Modern SPC rules (Western Electric) look beyond just individual points.
Related Tools
Evaluate classification models.
Classification metrics.
Diagnostic test accuracy.
Diagnostic test performance.
Updated diagnostic probability.
Risk ratio comparison.