One monitors average level and the other within-subgroup spread; either can change independently.

Can an R-chart lower limit be zero?

Yes. For small subgroup sizes, conventional lower limits are truncated at zero because ranges cannot be negative.

Can a chart guarantee that the next item meets a requirement?

No. It monitors statistical behavior and does not provide that individual-item guarantee.

AI Glossary / Statistics and probability / Intermediate

X-bar-R chart

Definition

An X-bar-R chart is a paired statistical process-control display that monitors subgroup averages and subgroup ranges over time. The X-bar chart tracks the process center, while the R chart tracks within-subgroup variability. Together they help detect patterns inconsistent with a stable baseline process when the subgroup design and chart assumptions are appropriate.

What it means

Each plotted point summarizes a small group of measurements collected under comparable conditions. For example, five consecutive items might form one subgroup. Their average describes the level during that sampling occasion, and their range describes the spread within it.

The two charts answer different questions. A stable average can coexist with increasing variability, and a stable range can coexist with a shifted average. Examining both prevents a single summary from concealing an important change. Control limits describe expected process behavior under a baseline; they are not customer specifications or acceptable-performance targets.

How it works

For equal-sized subgroups, calculate each mean and range. Average the subgroup means to obtain the grand mean and average the ranges to obtain R̄. The usual X-bar limits are grand mean ± A₂R̄. The R-chart limits are D₃R̄ and D₄R̄, with constants determined by subgroup size.

Establish a credible baseline, examine variability stability and then interpret the mean chart. Classical constants assume an appropriate normal-process framework and rational subgrouping. Small subgroups are the usual setting for range charts. Keep the monitoring limits fixed for a defined phase; investigate signals rather than continually absorbing them into a newly recalculated baseline.

Three practical examples

1. Manufacturing: A team measures five consecutive parts at regular intervals. The paired charts help distinguish changes in typical dimensions from changes in short-term variation.

2. AI service operations: Engineers monitor a stable benchmark workload in repeated small batches. They first check whether subgroup construction and response-time behavior support this chart, rather than applying normal-process limits blindly to mixed traffic.

3. Administrative processing: A team records comparable transaction-processing times within each sampling period. The range chart prompts investigation when variation increases even though the period averages remain near their historical level.

Think of it this way

Think of monitoring both the center and the width of a lane of moving traffic. The center can drift while the lane stays equally wide, or the lane can widen without its center moving. One view cannot replace the other. A control chart uses a statistical baseline to decide which changes deserve investigation.

Worked example

Consider an original baseline-summary example with subgroups of five measurements. Assume a reviewed stable baseline has grand mean 50 units and average range R̄ = 4 units. For subgroup size five, use the conventional tabulated constants A₂ = 0.577, D₃ = 0 and D₄ = 2.115.

The X-bar lower and upper limits are 50 − 0.577 × 4 = 47.692 and 50 + 0.577 × 4 = 52.308. Its center line is 50. The R-chart limits are zero and 2.115 × 4 = 8.46, with center line four.

A subsequent subgroup has mean 53 and range five. Its mean exceeds the upper X-bar limit, while its range lies within the R-chart limits. That is a signal concerning location under this rule, not proof of a specific cause. The example uses supplied baseline summaries; it does not establish baseline stability from only one subgroup or from the arithmetic itself.

Illustrative case study

In this fictional case, Ash Components monitors five parts from the same machine at fixed intervals. One subgroup average exceeds its established limit, although the range remains ordinary. Staff check the time-stamped production record and find a recent setup adjustment.

The team verifies the measurements, investigates the adjustment and documents the response. It does not automatically delete the subgroup or widen the limits to make the chart appear stable. After a justified process change, it gathers appropriate new baseline evidence before defining a new monitoring phase. Product acceptance remains governed by separate engineering specifications.

Common misconceptions and mistakes

Control limits are not specification limits. A statistically stable process can consistently produce unacceptable output, while an unstable process may temporarily produce items within specification. Process capability and stability are related but distinct questions.

Do not combine unrelated machines, products or workloads into arbitrary subgroups and expect meaningful limits. Unequal subgroup sizes require suitable adjustments. Strong serial dependence or extreme skew can undermine ordinary chart interpretation. A point inside the limits does not prove that no change occurred, and a point outside them does not identify the cause. Additional run rules also affect false-alarm behavior.

Business use

An X-bar-R chart can make operational monitoring more disciplined by separating common baseline variation from signals requiring investigation. Its usefulness depends on a practical response plan: who reviews a signal, which records they inspect and how any corrective action is evaluated.

Preserve timestamps, subgroup membership, measurement methods and baseline decisions. Report changes in product mix or operating conditions that could alter the chart's meaning. For AI systems, ensure that benchmark batches remain comparable; changing prompt lengths or model configurations can create signals that reflect a changed measurement task rather than deterioration in a fixed service.

Check your understanding

In the example, which chart signals for mean 53 and range five?

Answer: The X-bar chart; the R-chart point remains within its limits.

Would replacing the control limits with customer tolerances produce the same analysis?

Answer: No. Tolerances express requirements, while control limits describe the chosen baseline's statistical behavior.

Why preserve subgroup membership?

Answer: It defines the within-subgroup variation used to estimate the limits.

Related terms

Control chart
Rational subgroups
Range
Specification limits
rational subgroup
process stability
common-cause variation
special-cause variation
specification limit
process capability

Learn next

Next, construct both charts from a time-ordered dataset with clearly defined subgroups. Examine how grouping choices affect average ranges. Then compare X-bar-S charts, individual-observation charts and methods intended for small persistent shifts or dependent measurements.

Sources

NIST/SEMATECH e-Handbook, “Shewhart X-bar and R and S Control Charts,” gives the conventional formulas and subgroup-size constants: https://www.itl.nist.gov/div898/handbook/pmc/section3/pmc321.htm

The baseline summaries, signal calculation and Ash Components case are original teaching illustrations. They assume a suitable stable reference process; the example does not independently demonstrate that assumption.

Previous: X-11 decomposition Next: X-bar-S chart