a Control Chart: Definition, Meaning, and Application
The control chart is the primary tool of statistical process control (SPC) and one of the most powerful methods for maintaining process quality and detecting problems before they produce defects. This guide explains what control charts are, how they work, the different types, how to interpret signals, and how organisations use them to achieve process stability and reduce variation.
Control chart Definition
A control chart is a time-series graph with statistically calculated upper and lower control limits that distinguishes between normal process variation (common cause) and signals of unusual events (special cause).
- Plots process data over time against statistically derived control limits
- Distinguishes common cause variation from special cause variation
- Control limits are set at ±3 standard deviations from the process mean
- Developed by Walter Shewhart at Bell Labs in the 1920s
- Foundation of statistical process control (SPC)
Explanation of Control chart
The control chart was invented by Walter Shewhart at Bell Laboratories in 1924. Shewhart's insight was that variation in any process has two fundamentally different origins: common cause variation, which is inherent, random, and predictable; and special cause variation, which signals that something unusual has affected the process. These two types of variation require different responses — treating common cause variation as if it were special cause leads to unnecessary tampering that increases variation rather than reducing it.
A control chart plots process measurements in time sequence and adds three horizontal reference lines: the centreline (the process average), the upper control limit (UCL), and the lower control limit (LCL), typically set at three standard deviations above and below the mean. When all data points fall within the control limits without any non-random patterns, the process is said to be 'in statistical control' — behaving predictably. Any point outside the control limits, or any of several non-random pattern rules, signals a special cause that should be investigated immediately.
There are many types of control charts, each suited to different data types. X-bar and R charts monitor the mean and range of measurements from subgroups of continuous data. Individuals (I) and moving range (MR) charts monitor individual readings and are used when subgrouping is not practical. P-charts and np-charts monitor the proportion and number of defective items in a sample. C-charts and u-charts monitor the count and rate of defects per unit. Selecting the appropriate chart type for the data and process is essential for valid analysis.
How to Create and Use a Control Chart
-
1Select the process characteristic
Choose a measurable process output or key quality characteristic that is important to monitor for stability and performance.
-
2Determine subgroup size and frequency
Decide how many measurements to take per subgroup and how frequently to sample, based on process speed and the cost of sampling.
-
3Collect baseline data
Gather at least 20-25 subgroups of data while the process is operating normally to calculate reliable control limits.
-
4Calculate control limits
Compute the process average (centreline) and control limits at ±3 standard deviations using the appropriate formulas for the chosen chart type.
-
5Plot ongoing data and interpret
Plot new data points as they are collected, checking for points outside control limits and non-random patterns indicating special cause variation.
-
6Investigate and act on signals
Investigate any signal immediately to identify and remove the special cause; never adjust the process in response to common cause variation.
Control chart Formula
UCL = X̄ + 3σ CL = X̄ LCL = X̄ - 3σ
The control limits are positioned at three standard deviations above (UCL) and below (LCL) the process mean (X̄). This places approximately 99.73% of all common-cause variation within the limits, so a point outside the limits has less than 0.27% probability of occurring by chance — providing a strong signal of a special cause.
Control Charts for Tablet Weight in Pharmaceutical Manufacturing
A pharmaceutical manufacturer monitors the weight of tablets during production using an X-bar and R chart. Every 30 minutes, a subgroup of five tablets is weighed and the mean weight and range are plotted. The specification for tablet weight is 500 mg ± 15 mg, and the control limits, calculated from 25 baseline subgroups, are UCL = 510 mg and LCL = 490 mg for the X-bar chart.
During a Monday morning production run, the operator observed that the X-bar chart showed a run of eight consecutive points below the centreline — a non-random pattern (Nelson Rule 2) indicating a process shift. Investigation revealed that a new batch of active pharmaceutical ingredient from a different supplier had slightly lower bulk density, causing the filling mechanism to deliver slightly less material per tablet. The supplier issue was escalated, the affected tablets were quarantined for testing, and the filling parameters were adjusted to compensate. The control chart detected this systemic shift before any tablets fell outside specification — preventing a potential batch rejection and regulatory event.
Importance of Control chart in Quality Management
Control charts are one of the few tools that can distinguish between the two types of mistakes organisations make in process management: ignoring real problems (failing to detect special causes) and creating problems by over-adjusting stable processes (tampering). Without control charts, operators who see a process measurement drift toward a specification limit naturally adjust the process — even when the drift is just normal common-cause variation. This tampering typically increases variation and makes processes worse, not better.
At the organisational level, widespread use of control charts creates a fact-based quality culture. When process owners can see the statistical evidence for process stability (or instability), quality discussions are grounded in data rather than opinion. This evidence base is essential for prioritising improvement investments, evaluating the impact of process changes, and communicating process performance to customers and regulators with confidence.
- Distinguishes common cause from special cause variation
- Enables real-time process monitoring and early problem detection
- Prevents unnecessary and harmful process tampering
- Provides statistical evidence for process capability assessment
- Creates a visual, accessible record of process behaviour over time
- Supports regulatory compliance and customer quality audits
Manufacturing process monitoring, pharmaceutical production, food and beverage quality control, healthcare clinical process monitoring, financial transaction error tracking, call centre service quality, and any repetitive process where variation needs to be understood and managed.
Control chart in ASQ Certifications
Professionals working in quality, process improvement, operations, and organisational excellence often encounter this concept in real-world applications. Many ASQ certifications cover related principles,
tools, and methods as part of the Body of Knowledge.
Frequently Asked Questions
The concept of Control chart is rigorously covered in the following ASQ certifications: Certified Quality Engineer, Six Sigma Green Belt.
Control limits are calculated from actual process data and define the range of common-cause variation. Specification limits are set by the customer or designer and define the acceptable range for the product. A process can be in statistical control (within control limits) but still produce out-of-specification product if the process is not centred or capable. These two sets of limits should never be confused.
A minimum of 20 to 25 subgroups is the conventional guideline for calculating initial control limits. Fewer points produce unreliable limits that are likely to be revised significantly as more data is collected. For I-MR charts, at least 25 individual observations are recommended.
The Western Electric (WECO) rules are a set of criteria for detecting non-random patterns in control charts beyond points outside control limits. They include: 2 of 3 points in Zone A (beyond 2σ), 4 of 5 points in Zone B (beyond 1σ), 8 consecutive points on one side of the centreline, and 6 consecutive points trending in one direction. These rules increase sensitivity to process shifts.
An X-bar chart plots the average of a subgroup of multiple measurements taken close together in time. An individuals (I) chart plots single measurements. X-bar charts are more sensitive to process shifts because averaging reduces the effect of within-subgroup variation. Individuals charts are used when only one measurement per time period is practical or when subgrouping is not meaningful.