Is this the same as an X-bar chart?

No. An individuals chart plots individual measurements; an X-bar chart plots subgroup averages.

Can a stable process fail a requirement?

Yes. Stability concerns its pattern over time, while capability concerns performance relative to requirements.

AI Glossary / Statistics and probability / Intermediate

Individuals chart

Definition

An individuals chart is a statistical process control chart that plots one numerical observation at each monitoring occasion. A center line represents the process level, and control limits help identify observations inconsistent with the established pattern of variation. It is often paired with a moving-range chart, which tracks differences between consecutive observations.

What it means

The word individuals refers to individual measurements, not necessarily individual people. A measurement might be one batch's processing time, one item's weight, or one scheduled reading from a machine.
The chart asks whether the process appears to be behaving consistently with its baseline. It does not directly ask whether performance meets a customer requirement. A stable process can still be too slow, too variable, or otherwise unsuitable for its intended purpose.

How it works

Choose a clearly defined numerical measure and preserve the observations' time order. The usual individuals-chart setting has a subgroup size of one at each monitoring occasion. Combining unlike processes or changing how measurements are taken can make a single chart misleading.
Use a baseline period to understand the process and estimate its level and variation. Investigate identifiable changes, data errors, and unusual operating conditions. Do not remove inconvenient observations solely to make the baseline appear stable.
A common construction estimates short-term variation from moving ranges of length two. For consecutive observations x[i-1] and x[i], the moving range is their absolute difference. There are one fewer such ranges than observations.
Let x-bar be the mean of the individual observations and MR-bar the mean of those ranges. The usual estimate of the process standard deviation is MR-bar divided by 1.128. This constant comes from the expected range of two independent normal observations.
The corresponding three-sigma individuals-chart limits are x-bar minus and plus three times that standard-deviation estimate. The center line is x-bar. These are limits for individual observations; dividing by the square root of the number of baseline observations would instead target a different uncertainty quantity.
Review the moving-range chart as well as the individuals chart. A process can change in variability without a simple shift in average. Be careful to use the correct formulas for each chart; individuals limits and moving-range limits are not interchangeable.
For ongoing monitoring, use established limits until a justified process review supports changing them. Recalculating limits after every unusual observation can absorb the very change the chart is intended to reveal.
A point beyond a limit is a signal to investigate, not proof of a particular cause. Patterns within the limits may also matter under preselected supplementary rules. Adding many rules increases opportunities for false alarms, so define the decision procedure in advance.
The familiar limits rely on a suitable process model. Strong serial dependence, substantial skewness, changing operating conditions, coarse measurements, or poorly estimated baseline parameters can alter their behavior. Consider a different chart or a model-based residual approach when the assumptions do not fit the application.

Three practical examples

A manufacturing team monitors one measured batch characteristic at each production run.
A service team tracks a consistently defined processing-time measurement taken at scheduled intervals.
An AI infrastructure team monitors standardized benchmark latency while keeping workload and measurement conditions comparable.

Think of it this way

Think of a familiar route with a typical travel time and ordinary day-to-day variation. A markedly unusual journey prompts investigation. The historical pattern helps identify a change, but it does not determine the arrival time your customer requires.

Worked example

An original arithmetic example uses five time-ordered measurements: 10, 12, 11, 13, and 12. Their sum is 58, so the individuals mean is 11.6.
The four moving ranges are 2, 1, 2, and 1. Their average is 1.5. Using the rounded conventional constant, the estimated standard deviation is 1.5/1.128, approximately 1.3298.
The individuals limits are therefore 11.6 minus and plus 3*1.3298: approximately 7.6106 and 15.5894. If those illustrative limits are held fixed, a subsequent observation of 16 lies above the upper limit and triggers the stated beyond-limit rule.
The next moving range would be the absolute difference between 16 and the preceding value 12, which is 4. That value belongs on the separate moving-range chart; it is not compared with the individuals limits.
Five baseline observations are deliberately used only to make the arithmetic transparent. They do not provide an adequate basis for declaring a real process stable or claiming a precisely calibrated false-alarm rate.

Illustrative case study

In a fictional case, Northgate Automation monitors the latency of a standard model-serving test. A chart signal appears immediately after a deployment.
Engineers check workload settings, timing instrumentation, and system logs before attributing the signal to the new model. They discover that the benchmark configuration changed, so the old and new observations are not directly comparable. The team restores a consistent measurement procedure and reviews the baseline. The case shows how a signal guides investigation; it does not claim that a chart alone identifies a software defect.

Common misconceptions and mistakes

Do not confuse control limits with specification limits, service-level targets, or confidence intervals for the mean. Do not assume every point inside the limits proves statistical control.
Do not sort measurements by value before computing moving ranges. Do not assume the usual normal-model false-alarm probability applies exactly when limits are estimated from a small sample or observations are dependent.

Business use

An individuals chart can help separate routine fluctuation from signals that merit investigation when grouped measurements are impractical. It supports consistent escalation decisions and discourages reacting to every ordinary change. Pair the chart with operational knowledge, customer requirements, and a documented response procedure so monitoring produces useful investigation rather than repeated alarm handling.

Check your understanding

Why are there four moving ranges in the five-observation example?

Answer: Each range needs a consecutive pair, so the first observation has no preceding measurement in the baseline.

What does the signal at 16 establish?

Answer: It satisfies the chosen chart rule and warrants investigation, not automatic proof of a specific cause.

Related terms

Statistical process control
Control limits
Specification limits
Moving range
Control limit
Specification limit
Common-cause variation
Special-cause variation
X-bar chart

Learn next

Study baseline assessment and moving-range charts next. Then examine serial dependence, supplementary signaling rules, and the distinction between process stability and process capability.

Sources

NIST/SEMATECH e-Handbook, Individuals Control Charts: https://www.itl.nist.gov/div898/handbook/pmc/section3/pmc322.htm NIST/SEMATECH e-Handbook, What are Variables Control Charts?: https://www.itl.nist.gov/div898/handbook/pmc/section3/pmc32.htm

Previous: Individual fairness Next: Inducing point approximation