Individual and Moving Range Charts: When Subgroups Are Not Practical

When the I-MR Chart Is the Right Choice

The individual moving range (IMR) chart is the control chart you reach for when subgroups are not practical. If you can only collect one measurement per time period—batch chemistry, destructive testing, slow production, expensive assays—the I-MR chart is the statistically correct way to monitor your process. Trying to force this data onto an X-bar & R chart with a subgroup of n=1 produces undefined range values and meaningless control limits.

This guide covers when to use an I-MR chart, walks through a complete worked example with control limit calculations, and explains the assumptions you need to check before trusting the signals.

Naming Conventions The I-MR chart goes by several names in practice: I-MR, IMR, IX-MR, XmR, Individual-Moving Range, and Individuals chart. They all refer to the same paired chart: an Individuals (I or X) chart for the measured values and a Moving Range (MR or mR) chart for consecutive differences. The AIAG SPC Reference Manual uses “Individuals and Moving Range.”

Five Scenarios Where You Need an Individual Moving Range Chart

The common thread across all I-MR applications is one observation per sampling event. Here are the specific situations quality engineers encounter:

Batch processes: Chemical reactions, heat treatment batches, coating runs. Each batch yields one result (pH, hardness, thickness average). You cannot split a single batch into subgroups.
Destructive testing: Tensile strength, burst pressure, weld pull tests. The test destroys the part, so you measure one unit per time period and cannot go back for more.
Slow production rates: Low-volume, high-mix manufacturing where you produce 5–10 parts per shift. Waiting to collect a subgroup of 5 means you might chart one point per day—too slow to catch shifts.
Expensive measurement: CMM inspection that takes 30 minutes per part, or laboratory assays with significant per-sample cost. Measuring 5 parts per subgroup is economically impractical.
Continuous process readings: Temperature, pressure, or flow rate readings taken at regular intervals from a continuous process. Each reading is a single snapshot, not a subgroup average.

Worked Example: Pharmaceutical Tablet Weight

A pharmaceutical manufacturer measures the weight of one tablet every 30 minutes during a compression run. The target weight is 250 mg with a specification of 250 ± 7.5 mg (USL = 257.5 mg, LSL = 242.5 mg). Here are 20 consecutive measurements:

Raw Data (mg) 249.2, 250.1, 248.8, 251.3, 249.7, 250.5, 248.5, 249.9, 251.0, 250.3, 249.1, 250.8, 251.5, 249.4, 250.2, 248.7, 250.6, 251.1, 249.8, 250.4

Step 1: Calculate the Individual Chart Center Line and Limits

The center line (CL) of the Individuals chart is the mean of all measurements:

$$\bar{X} = \frac{\sum X_i}{n} = \frac{5000.9}{20} = 250.045 \text{ mg}$$

To calculate control limits, we first need the average moving range.

Step 2: Calculate Moving Ranges

The moving range is the absolute difference between consecutive measurements. For 20 individual values, there are 19 moving ranges:

MR₁ = |250.1 − 249.2| = 0.9, MR₂ = |248.8 − 250.1| = 1.3, MR₃ = |251.3 − 248.8| = 2.5, …

The average moving range across all 19 values:

$$\overline{MR} = \frac{\sum MR_i}{n-1} = \frac{24.6}{19} = 1.295 \text{ mg}$$

Step 3: Calculate I Chart Control Limits

I Chart Control Limit Formulas $$UCL_X = \bar{X} + \frac{3 \times \overline{MR}}{d_2}$$ $$LCL_X = \bar{X} - \frac{3 \times \overline{MR}}{d_2}$$

where $d_2 = 1.128$ for a moving range of span 2 (consecutive pairs). This constant converts the average moving range into an estimate of the process standard deviation: $\hat{\sigma} = \overline{MR} / d_2$.

$$\hat{\sigma} = \frac{1.295}{1.128} = 1.148 \text{ mg}$$ $$UCL_X = 250.045 + 3(1.148) = 250.045 + 3.443 = 253.488 \text{ mg}$$ $$LCL_X = 250.045 - 3(1.148) = 250.045 - 3.443 = 246.602 \text{ mg}$$

Step 4: Calculate MR Chart Control Limits

MR Chart Control Limit Formulas $$UCL_{MR} = D_4 \times \overline{MR}$$ $$LCL_{MR} = D_3 \times \overline{MR}$$

For span 2: $D_4 = 3.267$ and $D_3 = 0$. The MR chart has no lower control limit.

$$UCL_{MR} = 3.267 \times 1.295 = 4.230 \text{ mg}$$ $$LCL_{MR} = 0$$

Step 5: Interpret the Chart

All 20 points fall within the control limits (246.516–253.374 mg), and no Western Electric or Nelson rule violations appear. The process is in statistical control.

Note the relationship between control limits and specification limits: the control limits (246.6–253.5) sit well within the spec limits (242.5–257.5), indicating the process has margin. A quick capability estimate: Cpk = min[(257.5 − 250.045) / (3 × 1.148), (250.045 − 242.5) / (3 × 1.148)] = min[2.16, 2.19] = 2.16. This exceeds the 1.33 minimum—the process is both stable and capable.

The Normality Assumption and Why It Matters More for I-MR

Subgroup charts (X-bar & R) benefit from the Central Limit Theorem: even if individual measurements are not normally distributed, subgroup averages tend toward normality as subgroup size increases. The I-MR chart has no such buffer. Each plotted point is a single raw measurement, so the 3-sigma control limits assume the individual values are approximately normal.

When individual data is heavily skewed—cycle times, particle counts, chemical concentrations near detection limits—the I-MR chart produces asymmetric risk:

On the long-tail side: chronic false alarms (points above UCL that are actually normal for the skewed distribution)
On the short-tail side: missed shifts (real process changes masked by the compressed side of the distribution)

What to Do About Non-Normal Data Before applying an I-MR chart to data you suspect is non-normal, plot a histogram. If the distribution is visibly skewed, consider a Box-Cox or Johnson transformation to normalize the data before charting. Alternatively, some practitioners use the median and median moving range for a more robust (but less sensitive) individual chart. The NIST/SEMATECH e-Handbook of Statistical Methods covers these alternatives in detail.

Common Mistakes with I-MR Charts

Mistake	Why It Happens	Consequence
Using I-MR when subgroups are available	Convenience—plotting individuals is simpler than organizing subgroups	Reduced sensitivity to mean shifts. An X-bar chart with n=5 detects a 1-sigma shift about 3× faster than an I chart.
Ignoring autocorrelation	Continuous process data (temperature, flow) sampled frequently often shows serial correlation	Autocorrelated data narrows the true variation, making control limits too tight and triggering false alarms. Consider increasing the sampling interval or using an EWMA chart.
Using a moving range span > 2	Trying to “smooth” the MR chart by averaging over 3+ consecutive differences	Larger spans mask short-term variation and delay shift detection. The standard span of 2 (consecutive pairs) is recommended by the AIAG SPC Reference Manual for most applications.
Confusing control limits with spec limits	New users see that UCL = 253.4 and USL = 257.5 and assume the chart monitors specifications	Control limits describe what the process actually does; spec limits describe what it should do. They serve different purposes. For the relationship between the two, see our guide on Cpk vs. Ppk capability indices.

I-MR vs. X-bar & R: A Sensitivity Comparison

The practical question is often: “I collect 5 parts per hour—should I plot them individually or as subgroups?” The answer depends on what you are trying to detect.

An X-bar chart with n=5 has control limits that are narrower by a factor of √5 (about 2.24×) compared to the Individuals chart. This means the X-bar chart detects a 1-sigma mean shift with approximately 84% probability on the first sample after the shift, while an I chart detects the same shift with only about 16% probability. For shift detection, subgroups win decisively.

However, the I-MR chart preserves the original data resolution. If you need to see individual part-to-part variation (not just the subgroup average), the I chart shows it. Some practitioners run both: an X-bar & R chart for process monitoring and an I-MR chart as a diagnostic tool when investigating specific out-of-control signals.

For guidance on choosing between chart types across all data scenarios, see our complete decision guide.

Key Takeaways

Use the I-MR chart when you have one observation per sampling event—batch processes, destructive tests, slow production, expensive measurements
Control limits use d₂ = 1.128 and D₄ = 3.267 for the standard span-2 moving range
Check normality before interpreting—the I-MR chart is more sensitive to non-normal data than subgroup charts
Do not use I-MR when rational subgroups are available; you sacrifice significant shift detection sensitivity
Watch for autocorrelation in continuous process data sampled at high frequency