Types Of Control Charts: How To Choose The Right SPC Chart

Control charts are the backbone of Statistical Process Control (SPC), but picking the wrong one can lead you to conclusions that don’t reflect what’s actually happening on the floor. Understanding the types of control charts available is the first step toward monitoring process behavior accurately and making decisions based on real signals, not noise.

The challenge most practitioners face isn’t whether to use a control chart. It’s figuring out which one fits their data. Variable data and attribute data require different charts. Sample sizes matter. And choosing incorrectly can mask variation or trigger false alarms, both of which cost time and money. At Lean Six Sigma Experts, we’ve spent over a decade helping organizations implement data-driven process improvements, and chart selection is one of the most common sticking points we see in SPC deployments.

This guide breaks down the major categories of control charts, explains when each one applies, and gives you a practical framework for selecting the right chart based on your data type and sample size. Whether you’re monitoring dimensions on a production line or tracking defect rates across shifts, you’ll walk away knowing exactly which chart belongs where, and why it matters.

Why control charts matter in SPC

Statistical Process Control is built on one central idea: processes vary, and not all variation carries the same meaning. A control chart gives you a way to visualize that variation over time, plotting data points against statistically calculated limits. When you track a process without one, you’re essentially guessing whether a shift in output is meaningful or just background noise. That guessing leads to two costly mistakes: reacting to variation that doesn’t require action, and ignoring variation that does.

The difference between common cause and special cause variation

Every process produces variation, but there are two distinct types, and how you respond depends entirely on which one you’re dealing with. Common cause variation is the natural, predictable noise built into any system. It stays within your control limits, and trying to eliminate it by tweaking the process usually makes performance worse, not better. Special cause variation is unpredictable and comes from outside the normal process, such as a worn tool, a new operator, or a raw material inconsistency.

Treating common cause variation as if it were a special cause is one of the most expensive mistakes in process management, and control charts are the primary tool that prevents it.

A control chart separates these two types by applying upper and lower control limits calculated from your actual process data, typically set at three standard deviations from the mean. Points within those limits signal common cause variation. Points outside them, or specific patterns within them, signal something worth investigating. Without this distinction, you’re either over-correcting or under-responding, and both paths erode process stability over time.

What control charts reveal that other tools miss

Histograms and summary statistics describe the distribution of your data, but they strip out the time dimension entirely. A histogram showing a normal distribution can hide a process that has been slowly drifting for weeks. A control chart preserves the sequence of your data, which means you can see trends, cycles, and sudden shifts that static summaries bury completely.

The different types of control charts extend this capability further by tailoring the analysis to your specific data type. You would not use the same chart for individual measurements that you would use for defect counts per unit, because the underlying statistics are different. Applying the correct chart type means your control limits accurately reflect your process behavior, giving you signals you can actually trust rather than false alarms that drain your team’s time and credibility.

Recognizing the role control charts play in continuous improvement cycles is also part of understanding why they matter. In a DMAIC project, control charts belong in both the Measure phase, where you establish baseline performance, and the Control phase, where you verify that improvements hold over time. They serve as the ongoing evidence that a process remains in a state of statistical control after changes are implemented. That is not a one-time exercise. It is a monitoring discipline that prevents hard-won gains from quietly sliding back to where they started.

How to choose the right control chart

Choosing the right chart starts with a single question: what kind of data are you working with? The answer determines which of the major types of control charts actually applies to your situation. Get this wrong, and your control limits will be calculated on assumptions that don’t match your data, which means every signal the chart produces is unreliable from the start.

Start with your data type

Your data falls into one of two categories: variable data or attribute data. Variable data is continuous, meaning it can take any value within a range, such as temperature, diameter, or cycle time. Attribute data counts things, such as the number of defective parts in a batch or the number of defects per unit. Each category has its own family of charts built around the statistical properties of that data type.

Applying a variable chart to attribute data, or vice versa, produces control limits that don’t reflect your actual process behavior, which makes the chart useless for detecting real problems.

A practical way to test your data type: ask whether the measurement can fall between two whole numbers. If it can, you are working with variable data. If it can only be a whole number count, you are working with attribute data.

Consider your sample size

Once you know your data type, sample size becomes the next deciding factor. For variable data, individual measurements require a different chart than subgrouped measurements taken in sets of two or more. For attribute data, whether your sample size stays constant across collection periods or fluctuates significantly affects which chart applies.

Your situation also depends on whether you are tracking defects per unit or simply classifying each unit as pass or fail. These are distinct cases that call for different charts even within the same attribute family. Sorting through both dimensions before you start charting saves you from rebuilding your monitoring system later when the results stop making sense.

A quick reference covering both dimensions:

Variable control charts for continuous data

Variable data is any measurement that flows along a continuous scale, such as weight, voltage, wall thickness, or cycle time. These are the types of control charts most practitioners encounter first, and for good reason: continuous measurements carry more information per data point than count-based data, which means your chart detects process shifts faster and with fewer samples collected over time.

The I-MR chart for individual measurements

When you collect one measurement per time period, the Individuals and Moving Range (I-MR) chart is your standard choice. The I chart tracks each individual value, while the MR chart monitors variation between consecutive readings by calculating the range between back-to-back data points. Together they give you a picture of both where your process is running and how stable it is from one observation to the next.

Use the I-MR chart when your process naturally produces one result per cycle, such as a batch chemical reaction, a daily production yield figure, or a single lab test per unit.

The I-MR chart performs well for low-volume or slow-cycle processes where waiting to collect subgroups is impractical. One important limitation: the I chart is more sensitive to non-normal data distributions than subgroup-based charts, because subgrouping tends to normalize averages through the central limit theorem. If your data is heavily skewed, check the distribution before trusting the control limits the chart produces.

X-bar and R or S charts for subgrouped data

When your process allows you to collect multiple measurements within a short, consistent window, grouping those readings into rational subgroups and plotting them with an X-bar chart produces more statistically stable estimates of process location. The X-bar chart tracks the average of each subgroup, while a companion range or standard deviation chart runs alongside it to monitor within-subgroup variation separately.

Your choice between the R chart and the S chart depends on subgroup size. For subgroups of two to eight observations, the Range (R) chart is straightforward to calculate and easy for operators to interpret on the floor. Once your subgroup size exceeds eight, the S chart, which tracks the standard deviation of each subgroup, becomes more sensitive to real shifts in process spread and is the statistically stronger option at that point.

Attribute control charts for count data

Attribute data does not give you a precise measurement on a continuous scale. Instead, it counts outcomes: how many units failed inspection, how many defects appeared on a surface, how many errors occurred in a batch. The types of control charts built for attribute data reflect those different counting scenarios, and selecting the right one depends on whether you are classifying entire units or counting individual defects, and whether your sample size holds constant across collection periods.

p-chart and np-chart for pass/fail data

When your process output is classified as either conforming or nonconforming, you are working with proportion or count of defective units, and two charts handle this case. The p-chart tracks the proportion of defective units in each sample, which makes it the correct choice when your sample size changes from one collection period to the next. If you pull 80 units one day and 120 the next, the p-chart adjusts its control limits accordingly.

Use the np-chart only when your sample size stays the same every time you collect data. If that consistency breaks down, switch to the p-chart immediately.

The np-chart tracks the actual count of defective units rather than the proportion, which makes it slightly more intuitive for operators to interpret on the floor. Both charts require a large enough sample size to produce meaningful control limits, and a common guideline is to collect enough units so that you expect at least one or two defectives per subgroup on average.

c-chart and u-chart for defect counts

Some products can have multiple defects on a single unit without being classified as defective overall. A painted panel might have three scratches. A circuit board might have two solder bridges. In these cases, you are counting defects rather than defective units, and a different pair of charts applies. The c-chart tracks the total number of defects per sample and works only when the inspection area or opportunity for defects stays constant across every sample.

When your inspection area or sample size varies, the u-chart calculates defects per unit, normalizing the count so that comparisons across different sample sizes remain valid. Choosing between these two comes down to a single check: confirm whether your opportunity for defects is fixed or fluctuating before you commit to either chart.

How to interpret control chart signals

A control chart only delivers value if you know how to read what it’s showing you. Plotting the data is the easy part. The harder skill is distinguishing between a signal that demands action and normal process variation that you should leave alone. Getting this right is where the different types of control charts pay off, because each chart type generates signals based on the statistical properties of its specific data, and you interpret them using the same set of rules regardless of which chart you are running.

Recognizing out-of-control signals

The most obvious signal is a single point falling outside your upper or lower control limits. That point tells you something changed in your process, and it warrants immediate investigation before more data collects and obscures the source. Beyond that single-point rule, several pattern-based rules help you catch shifts and trends before they push a point beyond the limits entirely.

A process can be statistically out of control even when every point sits within the control limits, which is why pattern rules matter as much as limit violations.

Common patterns to watch for include:

• Eight consecutive points on the same side of the center line, which signals a process shift
• Six points in a row trending steadily up or down, which signals a drift in process mean
• Two out of three consecutive points in the outer third of the chart, near but inside a control limit
• Fourteen points alternating up and down, which often signals artificial stratification in your data collection

Deciding when to act and when to wait

Once you identify a signal, your response depends on whether you can pinpoint an assignable cause. If you can trace a point or pattern back to a specific event, such as a shift change, a material lot switch, or a machine adjustment, then you have a special cause you can address directly. If no assignable cause surfaces after a thorough investigation, the signal may indicate that your process has shifted structurally, and you need to recalculate your control limits using recent data to reflect the current process baseline accurately.

Reacting to every minor fluctuation without applying these rules wastes resources and undermines confidence in your SPC system over time. Applying them consistently, on the other hand, builds the kind of disciplined monitoring that keeps your process gains intact and your team focused on problems that actually matter.

Key takeaways

Choosing the right chart from the many types of control charts available comes down to two factors you assess before you plot a single point: your data type and your sample size. Variable data calls for I-MR or X-bar charts depending on whether you collect individual measurements or subgroups. Attribute data splits into pass/fail scenarios handled by p and np charts, and defect-count scenarios handled by c and u charts.

Reading your chart correctly matters just as much as selecting it. Out-of-control signals include both points beyond the control limits and specific patterns within them, and knowing the difference between common cause and special cause variation determines whether you act or hold steady. Apply these rules consistently and your SPC system becomes a reliable early-warning tool rather than a chart you fill in after problems have already cost you.

Contact Lean Six Sigma Experts to put the right charts to work in your process.