A control plan sets out how a process is monitored after it has been improved: which characteristics are measured, by what method, how often, by whom, and — the part that matters — what happens when a measurement falls outside limits.
That last element is the one routinely left out, and without it the rest is data collection rather than control. A measurement with no defined response changes nothing; it just documents the drift.
The reason it exists is that improvements decay. Conditions shift, people move on, workarounds accumulate, and within a year the gain is gone. A control plan is the difference between an improvement that holds and one that is quietly reversed while everyone believes the project succeeded.
In DMAIC terms this is the Control phase made concrete, which is why a project closed without one has not actually finished whatever the closure report says.
Plans fail by covering too much. Twenty monitored characteristics get ignored inside a month. Control the few that actually predict the outcome and let the rest go.
