Reading a Control Chart: A Field Checklist for What the Signals Mean
A control chart is the simplest tool in statistical process control and one of the most misused. Its job is to answer a single question: is this process behaving the way it normally does, or has something actually changed? Get that answer right and you stop chasing noise and start fixing causes. Get it wrong — by reacting to every wiggle — and you tamper with a stable process and make it worse.
In early 2021, processes everywhere were running off their historical baselines as demand swung and supply faltered. That is exactly when a control chart earns its keep: it tells you whether a number moved because the world changed or because nothing real happened at all. Use the checklist below to build one properly and read it before you act.
Build the chart correctly
Match the chart to the data. Use variables charts (X-bar and R, or I-MR for individuals) for measured data like cycle time or weight, and attributes charts (p, np, c, u) for counts and proportions of defects. The wrong chart type gives you wrong limits.
Calculate control limits from the process, not the spec. Control limits come from the data's own variation, typically set at three sigma from the centre line. Specification limits are what the customer wants; the two are different things and confusing them is the most common error.
Establish limits on a stable period. Compute limits from a baseline when the process was in control. Limits drawn from chaotic data simply hide the chaos.
Read the signals before you react
A point inside the limits, varying randomly, is common-cause variation — the normal noise of the process. Leave it alone. A signal means a special cause has appeared and is worth investigating. Watch for these:
A single point beyond a three-sigma control limit — the clearest alarm.
A run of seven or more points all on one side of the centre line, suggesting a shift.
A steady trend of several points climbing or falling, suggesting drift such as tool wear.
An unnatural pattern — cycles, or points hugging the limits — suggesting mixed sources or a sampling problem.
Act on what the chart tells you
Only investigate a special cause when the chart actually signals one — chasing common-cause noise is tampering.
When you find and fix a special cause, recalculate limits so the chart reflects the improved process.
Do not narrow control limits to match the specification; if the process cannot meet spec, that is a capability problem to solve, not a chart to redraw.
Keep the chart where the people doing the work can see it, so signals are caught in hours, not at month-end.
Used well, a control chart is a discipline for restraint as much as for action: it gives you permission to leave a stable process alone and a clear trigger for when to step in. That distinction — between noise and a real change — is most of the value in Lean Six Sigma's Control phase.
If your teams react to every data point and never to the right ones, XNM's strategic advisory can help you build measurement that drives the right decisions.