control chart
/ kun-TROHL chart /
Think of plotting your weight on the same chart every morning. One high day is meaningless, but a chart lets you see at a glance whether you are holding steady or slowly creeping up. A control chart does this for a measurement process: you plot the result of a check sample over time and watch whether the method is staying in control or drifting away.
A control chart is a running graph of a quality-control value — often a check standard or reference material measured with each batch — against time or run number. It carries a central line at the expected value and warning and action limits set at two and three standard deviations away. A point inside the limits and a random scatter mean the process is behaving; a point outside, or a non-random pattern, signals trouble.
It matters because it turns a stream of single numbers into an early-warning system: it catches a slow drift, a sudden shift, or creeping imprecision before they spoil real results. The caveat is that the limits only mean something if they were built from genuine in-control data, and the chart cannot tell you why a point went out — only that it did, prompting you to investigate.
A lab plots its daily result for a quality-control serum on a control chart with a central line at 5.0 mmol/L and action limits at 4.7 and 5.3. After the points trend upward for six straight days and cross the upper limit, the analyst stops, recalibrates and re-runs the day's samples.
Plotting a check sample over time to spot drift before it spoils results.
Warning limits (about two standard deviations) flag that something may be off and bear watching; action limits (about three standard deviations) mean the process is out of control and results must not be released until the cause is found and fixed.