X-bar R vs X-bar S Chart: Picking the Right Variation Estimator for Your Subgroup Size

You set up an X-bar chart for a five-piece subgroup and reach for the range chart out of habit. A colleague running a fifteen-piece automated subgroup pairs the same X-bar with a standard-deviation chart instead. Both are right. The X-bar R vs X-bar S chart choice comes down to one input — how many pieces are in your subgroup — and the reasoning behind the rule is worth knowing before your next setup.

This post lays out what actually differs between the two pairings, why the range stops being a good estimator as subgroups grow, and a clean subgroup-size verdict. Both charts watch the same two things: the X-bar tracks whether the process average is stable, and the bottom chart tracks whether the process spread is stable.

Both charts pair with X-bar — only the bottom chart changes

An X-bar and R chart plots subgroup means on top and subgroup ranges (largest minus smallest value) on the bottom. An X-bar and S chart plots the same means on top and subgroup standard deviations on the bottom. The top chart's job never changes; the question is which statistic best estimates within-subgroup variation.

The control-limit formulas mirror each other, swapping the range-based constants for the standard-deviation-based ones:

Key Formulas $$UCL_{\bar{X}} = \bar{\bar{X}} + A_2\bar{R}, \quad LCL_{\bar{X}} = \bar{\bar{X}} - A_2\bar{R} \qquad\text{(with R chart)}$$ $$UCL_R = D_4\bar{R}, \quad LCL_R = D_3\bar{R}$$ $$UCL_{\bar{X}} = \bar{\bar{X}} + A_3\bar{s}, \quad LCL_{\bar{X}} = \bar{\bar{X}} - A_3\bar{s} \qquad\text{(with S chart)}$$ $$UCL_s = B_4\bar{s}, \quad LCL_s = B_3\bar{s}$$

What the range throws away

The range uses only two numbers from each subgroup: the largest and the smallest. Every value in between is discarded. The standard deviation uses all of them. With small subgroups that loss barely matters. In a subgroup of four, the range already touches most of the data, so it is nearly as good an estimator as the standard deviation — with the bonus of being computable by hand at the workstation.

As the subgroup grows, the gap widens. A subgroup of fifteen has thirteen interior values the range never looks at, and a single extreme reading swings the range more than it should. The standard deviation keeps using every point, so it gives a steadier, more efficient estimate of sigma. That declining efficiency of the range — not any change in the math — is the whole reason the chart choice flips with subgroup size.

Tip Both statistics back into the same process sigma through their own unbiasing constant: $$\hat{\sigma} = \frac{\bar{R}}{d_2} = \frac{\bar{s}}{c_4}$$ The estimates agree closely for small n; they diverge in reliability, not in target, as n grows.

The subgroup-size rule — and why sources disagree

The practitioner convention, and the one the AIAG SPC Reference Manual follows, is to use the R chart for subgroups of roughly 8 or fewer and switch to the S chart at about 9 or more. You will see the threshold quoted as 8, 9, or 10 across textbooks and vendor help pages, and the disagreement is real but minor — the efficiency of the range erodes gradually, so there is no single cliff. Any cutover in the 8-to-10 band is defensible.

In practice the deciding factor is often how you collect data. Manual gauging tends to produce small subgroups, where the R chart is simpler to maintain and easy to explain to operators. Automated or high-volume sampling produces large subgroups where the S chart is the natural fit and the by-hand advantage of the range no longer applies.

Worked example: at n = 5 the two agree

Take one subgroup of five bore measurements in millimeters: 10.1, 10.3, 9.9, 10.2, 10.0. The mean is 50.5 / 5 = 10.10 mm.

Worked Example Range R = 10.3 − 9.9 = 0.40 mm
Standard deviation s = √( (0² + 0.2² + 0.2² + 0.1² + 0.1²) / (5 − 1) ) = √(0.10 / 4) = 0.158 mm
Sigma from the range: σ̂ = R / d₂ = 0.40 / 2.326 = 0.172 mm
Sigma from the std dev: σ̂ = s / c₄ = 0.158 / 0.940 = 0.168 mm

The two sigma estimates land within 0.004 mm of each other. At n = 5 the range gives up almost nothing, which is exactly why the R chart is the default for small subgroups. Run the same comparison at n = 15 and the standard-deviation estimate would be the more trustworthy of the two.

Side-by-side

Criterion	X-bar R chart	X-bar S chart
Within-subgroup statistic	Range (max − min)	Standard deviation
Best subgroup size	About 8 or fewer	About 9 or more
Data points used	Two per subgroup	All points per subgroup
X-bar limit constant	A₂ (0.577 at n = 5)	A₃ (1.427 at n = 5)
Spread limit constants	D₃ = 0, D₄ = 2.114 (n = 5)	B₃ = 0, B₄ = 2.089 (n = 5)
By-hand friendly?	Yes — subtraction only	Less so — needs a calculator
Efficiency as n grows	Declines	Holds

Common Mistake Pulling the control-chart constants for the wrong subgroup size. A₂, A₃, D₃, D₄, B₃, and B₄ all change with n. Using the n = 5 row for a subgroup of 8 inflates or deflates your limits and manufactures false signals.

Verdict by role

Quality technician collecting small manual subgroups: use the X-bar R chart. Subgroups of 3 to 5 lose nothing to the range, and the range is fast to compute and easy to chart at the workstation.
Process engineer on an automated, high-volume line: use the X-bar S chart. Large subgroups make the standard deviation the better estimator, and the software computes it anyway, so the by-hand advantage of the range is moot.
Quality engineer standardizing a program across lines: set the cutover once (8, 9, or 10 — pick one and document it), and key the choice to how each line collects data rather than debating the exact threshold.

For the step-by-step build of the small-subgroup case, see constructing an X-bar and R chart step by step. If you are not sure a variables chart is even the right family, the full control-chart selection decision guide covers attribute charts too, and what to do when rational subgroups are not practical at all handles the one-piece-at-a-time case.

To skip the constant lookups entirely, our control chart tool selects the range or standard-deviation pairing from your subgroup size and applies the right constants automatically. The ASQ control chart resource and the NIST/SEMATECH e-Handbook on Shewhart X-bar, R, and S charts document the constants and formulas in full.