A control chart plots a measurement over time against limits calculated from the process’s own behaviour, distinguishing the variation a stable process always shows from a signal that something actually changed.
That distinction is the whole purpose. Without it, teams react to every movement, and reacting to normal variation makes processes worse — adjusting a stable process in response to noise adds variation rather than removing it. Control limits come from the data, not from the specification; confusing the two is the most common error, and a process can sit comfortably within specification while being wildly out of control.
Which chart depends on the data: X-bar and R for measured values in subgroups, individuals and moving range for single readings, p and c charts for counts. Choosing wrongly produces limits that are simply incorrect.
Charts get printed, posted and never looked at, which is the quiet failure — a chart nobody reads is decoration. The other is recalculating limits after every shift, which guarantees the chart can never show a shift.
