SPC Control Charts (I-MR, Xbar-R, Xbar-S) Explained – Complete Guide

A practical guide to control charts — from common vs. special cause variation to I-MR, X̄-R and X̄-S charts, control limit formulas, and when to use each one.

---

Control charts are one of the core tools used in Six Sigma, quality management, and process improvement methodologies. They are especially common in engineering and manufacturing, but their usefulness is not limited to these fields. In fact, control charts can be applied anywhere data is collected over time.

For example, you might use a control chart when you are:

• Measuring pressure values at regular intervals
• Tracking cycle time on a production line
• Monitoring patient recovery times in a clinic

Although these examples come from different areas, they all have one important thing in common: they involve collecting data over time.

Common Cause vs. Special Cause Variation

A control chart helps answer a simple but very important question:

Is the variation in this data normal, or is something going wrong?

In any process, some amount of variation is expected. No process produces exactly the same result every single time. However, not all variation is the same. Control charts help separate two types:

• Common cause variation is the normal, expected statistical variation that naturally exists within a process.
• Special cause variation is unusual variation that suggests something unexpected may be affecting the process.

In other words, a control chart helps you understand whether the process is behaving normally, or whether there may be a problem that needs investigation.

What Does a Control Chart Look Like?

At its core, a control chart is simply a graph. Despite the statistical concepts behind it, the visual structure itself is quite straightforward. Instead of looking at isolated numbers in a spreadsheet, the control chart transforms the data into a visual story of how the process is behaving.

A standard control chart contains two main axes:

• The Y-axis, which represents the scale for measured values
• The X-axis, which represents time, sequence order, or another logical progression of the data

The most important rule is that the data points must be plotted in the order they occurred. The sequence matters because control charts are designed to show how a process behaves over time. If the data were randomly arranged, the chart would lose its meaning.

Plotting the Data

Each measurement is plotted on the chart as an individual data point. Depending on the type of control chart being used, that data point may represent different things — for example, an individual measurement (such as a single pressure reading or cycle time) or a subgroup average (mean). We will explain these different chart types and calculations later in the article.

After the points are plotted, they are connected with lines to make the behavior of the process easier to visualize: trends, sudden spikes or drops, repeating patterns, or stable behavior over time.

Control Limits vs. Specification Limits

One of the most important concepts to understand is that control limits are not the same thing as specification limits. These two are often confused, but they serve completely different purposes.

Specification Limits

Specification limits are defined by:

• Customer requirements
• Engineering drawings
• Product design standards
• Regulatory requirements

These limits define what is considered acceptable for the final product or process output. For example, a shaft diameter may be required to stay between 19.95 mm and 20.05 mm.

Specification limits are fixed targets. They usually do not change unless the product or customer requirements change. Most importantly: specification limits do not tell you whether the process itself is stable.

Control Limits

Control limits are completely different because they are calculated directly from the actual process data. They reflect how the process is currently behaving over time. Unlike specification limits:

• Control limits are statistically calculated
• They are based on variation within the process
• They can change as new data is collected
• They help evaluate process stability

In other words:

• Specification limits tell you what the process should produce
• Control limits tell you how the process is actually behaving

A process can meet specifications while still being unstable and unpredictable — and conversely, a process can be stable while not meeting specification limits.

The I-Chart (Individual Chart)

One of the most common control charts is the I-Chart, which stands for Individual Chart. In an I-Chart:

• Every single measurement is plotted individually
• Each data point represents one observation
• The chart is used when measurements are collected one at a time rather than in groups

This makes I-Charts especially useful when subgrouping data is not practical or when measurements occur individually over time. For example, an I-Chart could be used to monitor:

• Daily production cycle times
• Individual pressure readings
• Customer waiting times
• Single temperature measurements

Because every observation is shown separately, I-Charts provide a very detailed view of process behaviour over time.

Understanding Sigma and Control Limits

This is the point where a little statistics enters the picture, but the core idea is actually quite simple. Control charts are heavily based on the concept of the normal distribution, sometimes called the bell curve because of its shape. In a normal distribution:

• About 68% of all values fall within ±1 standard deviation from the average
• About 95% fall within ±2 standard deviations
• About 99.7% fall within ±3 standard deviations

The term standard deviation is often represented by the Greek letter sigma (σ). In I-Charts, the control limits are typically placed at the ±3σ level around the process average.

That means:

• The Upper Control Limit (UCL) is usually located at: UCL = μ + 3σ
• The Lower Control Limit (LCL) is usually located at: LCL = μ − 3σ

where:

• μ represents the process average (mean)
• σ represents the standard deviation of the process

Since 99.7% of normally distributed data should naturally fall within the ±3σ range, any point outside those limits is statistically very unlikely to occur by random chance alone.

In practical terms:

If a data point falls outside the control limits, something unusual may be happening in the process.

This is what control charts are designed to detect. The chart acts as an early warning system that tells you when the process behaviour may no longer be normal.

The MR Chart (Moving Range Chart)

So far, the I-Chart has focused on the actual measurement values themselves. But there is another important question we need to answer:

How much are consecutive measurements changing from one point to the next?

This is where the MR Chart, or Moving Range Chart, comes in. Instead of plotting the original data values, the MR Chart focuses on the differences between consecutive measurements.

For example, imagine your individual measurements are:

Observation	Value
1	10
2	12
3	11
4	15

The moving ranges would be calculated as the absolute differences between consecutive points:

• |12 − 10| = 2
• |11 − 12| = 1
• |15 − 11| = 4

These moving range values are then plotted on a separate chart:

MRᵢ = |Xᵢ − Xᵢ₋₁|

where:

• MRᵢ = moving range for observation i
• Xᵢ = current observation
• Xᵢ₋₁ = previous observation

The MR Chart therefore shows how much variation exists between consecutive measurements over time.

Why the MR Chart Matters

The I-Chart shows whether the actual process values are stable. The MR Chart shows whether the amount of variation itself is stable. Together, these charts provide a much clearer picture of process behaviour. This is why you will almost always see the I-Chart and the MR-Chart displayed together side by side as a combined monitoring system — usually called an I-MR Chart.

Calculating Control Limits for the MR Chart

The control limits for the MR Chart are calculated differently from the I-Chart. First, the average moving range is calculated:

M̄R = Σ MRᵢ / n

where:

• M̄R = average moving range
• MRᵢ = each moving range value
• n = number of moving ranges

The Upper Control Limit (UCL) is calculated using a statistical constant called D₄:

UCL_MR = D₄ × M̄R

The value of D₄ depends on the subgroup size. For a standard moving range chart, the subgroup size is usually 2, because each moving range compares two consecutive observations. The D₄ constant is obtained from standard statistical control chart constant tables.

The Lower Control Limit (LCL) is calculated using another constant called D₃:

LCL_MR = D₃ × M̄R

For very small subgroup sizes — especially subgroup size 2, which is most common in MR Charts — the value of D₃ is typically zero. That means:

LCL_MR = 0

As a result, the MR Chart often has:

• An Upper Control Limit
• A center line representing the average moving range
• And a Lower Control Limit equal to zero

This makes sense because variation cannot be negative.

The X̄ Chart (Xbar Chart)

So far, we have discussed I-Charts and MR-Charts, which are typically used when measurements are collected one at a time. But in many real-world processes, measurements are often collected in small groups called subgroups.

Imagine we are measuring the weight of a plastic injection-moulded part. Instead of measuring only one part at a time, we decide to measure the first 5 parts produced each day. That means:

• Each day becomes one subgroup
• Each subgroup contains 5 measurements
• The subgroup size is therefore n = 5

For every subgroup of 5 parts, we calculate the average value. For example:

Day	Measurements	Subgroup Average
1	10.1, 10.2, 10.0, 10.3, 10.1	10.14
2	10.0, 10.1, 10.2, 10.1, 10.0	10.08

When we plot all of these subgroup averages on a chart, the resulting chart is called an X̄ Chart (often written “Xbar Chart”). The name comes from X̄, which represents the subgroup mean. The X̄ Chart monitors how the process average changes over time.

The center line of the X̄ Chart is calculated using the average of all subgroup averages. This point is important:

The center line is NOT the average of all individual measurements directly. It is the average of the subgroup means.

If we have k subgroups:

X̿ = Σ X̄ᵢ / k

where:

• X̿ = overall average of subgroup means
• X̄ᵢ = mean of subgroup i
• k = number of subgroups

This overall average becomes the center line of the X̄ Chart.

Control Limits for the X̄ Chart

Now we need to determine the control limits. One common approach estimates the process standard deviation using the subgroup ranges.

Step 1: Calculate the range of each subgroup.

Within each subgroup of 5 measurements:

• Identify the maximum value
• Identify the minimum value
• Subtract the minimum from the maximum

R = X_max − X_min

This gives the range for that subgroup.

Step 2: Calculate the average range.

After calculating all subgroup ranges, we calculate their average:

R̄ = Σ Rᵢ / k

where:

• R̄ = average range
• Rᵢ = range of subgroup i
• k = number of subgroups

Step 3: Estimate sigma from the average range.

The process standard deviation can then be estimated using the constant d₂:

σ̂ = R̄ / d₂

The value of d₂ depends on the subgroup size. For subgroup size 5, d₂ = 2.326.

Step 4: Calculate the X̄ control limits.

The Upper and Lower Control Limits are calculated as:

UCL = X̿ + 3 · σ̂ / √n
LCL = X̿ − 3 · σ̂ / √n

where:

• X̿ = average of subgroup means
• σ̂ = estimated process standard deviation
• n = subgroup size

The term σ̂ / √n is called the standard error of the mean. These limits define the expected variation of subgroup averages over time. Once calculated, they are drawn directly onto the X̄ Chart.

The R Chart (Range Chart)

Now let’s go back to the subgroup ranges. Remember that for every subgroup we already calculated:

R = X_max − X_min

When these range values are plotted over time, the resulting chart is called an R Chart, or Range Chart. The purpose of the R Chart is to monitor variation within the subgroups themselves.

While the X̄ Chart monitors changes in the process average, the R Chart monitors changes in process spread or consistency.

Control Limits for the R Chart

The center line of the R Chart is the average range:

CL = R̄

The Upper Control Limit is calculated using the constant D₄:

UCL_R = D₄ × R̄

The D₄ constant comes from standard control chart tables and depends on the subgroup size. For subgroup size 5, D₄ ≈ 2.114; multiplying it by your R̄ gives the chart’s upper limit.

The Lower Control Limit uses another constant called D₃:

LCL_R = D₃ × R̄

For subgroup size 5, D₃ = 0, which means LCL_R = 0. In fact, for subgroup sizes up to 7, the value of D₃ remains zero. This means the lower control limit on many R Charts is simply zero, because variation itself cannot be negative.

The S Chart (Standard Deviation Chart)

For every subgroup, we can also calculate its own standard deviation. If a subgroup contains n measurements:

s = √[ Σ (xᵢ − x̄)² / (n − 1) ]

where:

• s = subgroup standard deviation
• xᵢ = each measurement in the subgroup
• x̄ = subgroup mean
• n = subgroup size

This value tells us how spread out the measurements are within that subgroup. After calculating the standard deviation for every subgroup, we plot those values on a chart. The resulting chart is called an S Chart, where “S” stands for standard deviation.

The S Chart monitors variation within the process, similar to the R Chart:

• The R Chart monitors variation using ranges
• The S Chart monitors variation using subgroup standard deviations

In many situations, especially with larger subgroup sizes, the S Chart provides a more reliable estimate of process variation.

Center Line of the S Chart

The center line of the S Chart is simply the average of all subgroup standard deviations:

S̄ = Σ sᵢ / k

where:

• S̄ = average subgroup standard deviation
• sᵢ = standard deviation of subgroup i
• k = number of subgroups

For example, if the average subgroup standard deviation is S̄ = 0.858, that value becomes the center line of the S Chart.

Control Limits for the S Chart

The control limits for the S Chart are calculated using statistical constants called B₃ and B₄. These constants depend on the subgroup size.

The Upper Control Limit:

UCL_S = B₄ × S̄

For a typical small subgroup size, B₄ might be around 2.089. With S̄ = 0.858, the Upper Control Limit is obtained by multiplying these two values together.

The Lower Control Limit:

LCL_S = B₃ × S̄

For subgroup sizes up to 6, B₃ = 0, which means LCL_S = 0. As with the R Chart, this makes practical sense because variation cannot be negative.

R Chart vs. S Chart: A Simple Rule of Thumb

When working with subgrouped data, the subgroup size plays a major role in choosing between an R Chart and an S Chart. A commonly used guideline is:

• If the subgroup size is 10 or less, an R Chart is usually sufficient
• If the subgroup size is greater than 10, an S Chart is generally preferred

Why?

The R Chart estimates variation using only two values: the maximum and the minimum. This works well for smaller subgroup sizes because the range still represents the subgroup variation reasonably accurately. However, as subgroup sizes become larger, relying only on the maximum and minimum values can introduce more statistical noise and reduce accuracy.

The S Chart becomes more reliable in larger subgroups because it uses all observations when calculating variation through standard deviation.

Estimating Sigma from Subgroup Standard Deviations

Earlier, the process sigma σ was estimated using the average range:

σ̂ = R̄ / d₂

But there is another method. Instead of using the average range, we can estimate sigma using the average subgroup standard deviation. To do this, we use a correction factor called c₄. The formula becomes:

σ̂ = S̄ / c₄

where:

• S̄ = average subgroup standard deviation
• c₄ = statistical correction factor based on subgroup size

Applying this formula to the example above gives σ̂ ≈ 0.9127. This result is slightly different from the sigma value obtained using the average range method.

Why Do the Control Limits Sometimes Differ Slightly?

Once sigma is estimated, the remaining control limit calculations stay essentially the same. However, because the estimated sigma value changes slightly depending on the method used, the final control limits may also shift slightly. This explains why an X̄–R Chart and an X̄–S Chart may appear to tell the exact same process story while still showing slightly different control limits.

The difference comes entirely from how the process standard deviation is estimated. In practice, the difference is usually small enough that most people barely notice it.

Summary

Control charts turn raw measurements into a clear picture of how a process behaves over time. By separating common cause from special cause variation, they let you react to real signals instead of normal noise. Choose your chart based on how data is collected:

• I–MR Chart — when measurements come one at a time
• X̄–R Chart — when data is grouped in small subgroups (≤ 10)
• X̄–S Chart — when subgroup sizes are larger (> 10)

With SIGMADESK, you don’t have to calculate d₂, D₃, D₄, B₃, B₄ or c₄ by hand — just feed your data in and the platform handles the rest, with built-in rule checking and capability analysis on top.