Problem Solving & Quality · Quality Tools

Control Chart (SPC)

Plot the process over time with limits calculated from its own variation, to tell common-cause noise from special causes that deserve a reaction.

Time1 h
FormatSmall group
StageQuality Tools

Control Chart (SPC): what it is and why it works

A control chart is a time-ordered plot of a process measurement with a center line and upper and lower control limits calculated from the process's own short-term variation, typically at three standard deviations of the plotted statistic. The chart does not ask whether parts are good or bad; it asks whether the process is behaving the way it usually behaves. Points that stay inside the limits without a pattern reflect common-cause variation, which belongs to the system and can only be reduced by changing the system. A point outside the limits, or a non-random pattern such as a long run on one side of the center line or a steady trend, signals a special cause worth investigating now.

The method works because it separates two costly mistakes: reacting to noise, which adds variation (often called tampering), and ignoring a real shift until scrap piles up. It beats inspection against specification because it shows a change before product goes out of tolerance. A control chart is also the entry ticket for other tools: capability indices are meaningful only on a stable process, the measurement system behind the data should pass a gauge study first, and the chart itself usually becomes the monitoring method written into the control plan.

What you need

A measurable characteristic that matters to the customer or to the next process step, with a defined sampling point
A measurement system known to be adequate, ideally confirmed by a gauge R&R study
20 to 25 subgroups (or individual values) collected under normal operating conditions
A rational subgrouping plan: consecutive parts from the same machine, stream or batch within each subgroup
An agreed set of detection rules and a place to log causes of signals

What you get

A center line and control limits based on actual process behavior
A clear verdict on whether the process is currently stable or disturbed by special causes
A log of signals with their assigned causes, which feeds root-cause work and the FMEA
A simple operator rule: adjust only on a signal, leave the process alone otherwise

When to use it

When operators adjust the process after every measurement and make it worse.

How to do it, step by step

Choose the chart for the data: X-bar/R for subgroups of measurements, individuals/moving range for single values, p or u charts for defect counts.
Collect 20–25 subgroups under normal operation and calculate the center line and control limits from that data — not from the specification.
Plot new points as they are produced, at the workplace, by the people who run the process.
React only to signals: a point beyond the limits, runs on one side of the center line, trends — and log the cause found.
Leave the process alone when there is no signal, and recalculate the limits only after a proven, deliberate change.

Worked example: Stopping the over-adjustment of a liquid filler

Illustrative scenario — figures are realistic but not from a real company.

A specialty chemicals plant fills one-gallon jugs of cleaning concentrate on a six-head rotary filler. The label claims 128 fl oz, and operators check-weighed one jug every 15 minutes, nudging the filler setpoint after almost every reading. Giveaway averaged 1.9 fl oz per jug, and short-fill complaints still arrived from a retail customer.

The quality engineer chose an X-bar and R chart with subgroups of five consecutive jugs, one subgroup every 30 minutes, converting weights to volume with the product's measured density.
Operators were asked to stop adjusting for three days while 24 subgroups were collected. The resulting limits were 129.6 to 130.4 fl oz for the averages, far tighter than the swings seen before.
The chart was posted at the filler and operators plotted each subgroup themselves. They adjusted only when a rule fired: a point outside the limits or eight in a row on one side of the center line.
Two signals appeared in the first month; both traced to a worn nozzle seal on head 4, which was added to the preventive maintenance list.

Result. Setpoint changes dropped from about 30 per shift to two or three per week. With the variation from tampering removed, the team lowered the target by 0.6 fl oz while keeping individual jugs, at three sigma, above the label claim, saving roughly $0.06 per jug in product. The lesson: most of the variation had been created by the adjustments themselves.

Open it in the FieldKit Problem Solving app → Get the Problem Solving template pack — join the waitlist

Common pitfalls and how to avoid them

Drawing specification limits on the chart and calling them control limits.Calculate limits from process data only. Keep specifications for capability studies and product disposition, not for deciding when to adjust.
Mixing streams in one subgroup, such as parts from four cavities or two parallel lines.Subgroup rationally: each subgroup should come from one stream over a short time, or use separate charts per stream.
Recalculating limits every week so that drifts are absorbed into wider limits.Freeze limits after the baseline and recalculate only after a documented, deliberate process change.
Plotting points in an office days later, where nobody can react.Put the chart at the workplace, have operators plot in real time, and write the expected reaction next to it.

Frequently asked questions

What is the difference between control limits and specification limits?

Control limits are calculated from the process data and describe what the process actually does when only common causes are present. Specification limits come from the customer or the design and describe what is acceptable. A process can be in control yet produce out-of-spec parts, or be out of control while all parts are still in spec. The first calls for reducing variation or re-centering; the second calls for finding the special cause.

Which control chart should I use for my data?

Use X-bar and R (or X-bar and S for larger subgroups) when you can take several measurements close together. Use an individuals and moving range chart when you get one value at a time, such as one batch analysis per shift. For attribute data, use a p or np chart for the proportion or count of defective units, and a c or u chart for the count of defects per unit or per area.

How many data points do you need to set control limits?

A common rule is 20 to 25 subgroups, or at least 20 to 30 individual values, collected under normal operation. Fewer points give limits that are uncertain and may shift noticeably as more data arrive. If you must start with less, label the limits as trial limits and recalculate once enough data are available, excluding points with known, removed special causes.

Origin

Control chart — Walter A. Shewhart, Bell Telephone Laboratories, 1924; Economic Control of Quality of Manufactured Product, 1931.

Used in these playbooks

Process stability review 2 weeks

Two weeks to find out whether a disputed characteristic is a measurement, stability or capability problem — and to lock the answer into the control plan.

Related methods

More in “Quality Tools”

The statistical and risk tools of industrial quality: Pareto, SPC, capability, measurement, FMEA.