Method

Control plan

In French: Plan de contrôle (plan de surveillance)

The control plan writes down what is monitored on a process, how, how often, by whom, and what is done when the measurement leaves the expected range.

What the document contains

For each monitored characteristic: the quantity measured, the measuring device, the size and frequency of the sample, the expected limits, the person who takes the reading, the place where it is recorded, and the planned reaction in case of deviation.

The last column is what distinguishes a useful plan from a decorative table. A reading with no written reaction produces a history nobody consults; it is the course of action that turns measurement into steering.

The reaction must be executable by the person who observes the deviation. "Alert quality" is one only if someone answers; otherwise it is a polite way of planning nothing.

Its place at the end of a project

It is the deliverable of the last phase, the one that decides whether the gain will hold. A project that improves a result without writing down what will maintain it sees that result degrade in the following months, generally at the first change of person or of run.

It also materialises the handover: from its signature, the process belongs to its usual owner and no longer to the project team. Without that moment, the team stays on call indefinitely and the owner never takes ownership of the new way of working.

It is built from what the analysis established, not from what is easy to measure. The characteristics to monitor are those whose influence has been demonstrated — that is the only link that justifies spending inspection time on them rather than on others.

Keeping it alive

A control plan goes out of date. Equipment changes, suppliers change, the causes that were treated cease to be risks and new ones appear. Without periodic review, it keeps having quantities recorded that have become pointless while a real risk is monitored by nothing.

The best trigger for revision is an incident: every deviation that got through points to either a characteristic missing from the plan, a frequency too low, or a reaction that did not work. It is free feedback, and it is rarely used.

An overloaded plan degrades differently: readings are done from memory or at the end of the shift, and the recorded values become plausible rather than exact. Better few characteristics really monitored than many of which nobody knows any more which are reliable.

The common mistakes

Leaving the reaction column empty or vague. It is the flaw that empties the document of its function: data is collected and nothing happens when it leaves the limits.

Putting in everything that is measurable. An overloaded plan is not applied in full, and nobody can say which lines are — which is worse than a short plan that is kept.

Writing it without those who will apply it. A frequency decided in an office runs into the real organisation of the station, and the reading is then done when possible rather than when planned.

On the shop floor — Foundry

The plan provided for a reading every two hours, with "notify the quality department" as the reaction. When asked, the operators were indeed recording the values and had never notified anyone: the department did not answer at night or at weekends, i.e. half of production. The column existed, the reaction did not.

The other tools illustrated by this trade: machining, mechanical engineering and finishing.

Frequently asked questions

What is the difference between a control plan and an inspection instruction?

The inspection instruction describes how to carry out a given check — the operating procedure of the gesture. The control plan says what to monitor across the whole process, how often and with what reaction in case of deviation. One is an execution document, the other a steering document; they complement each other and are not substitutes.

How many characteristics should be included?

As few as possible, retaining only those whose influence on the result was established during the analysis. A plan of a few lines actually applied is better than a plan of thirty of which nobody knows which are kept. The question to ask about each line is simple: what would be done with this measurement if it left the limits.

How often should a control plan be revised?

An annual review is the minimum, but most of the useful updates come from elsewhere: any change of equipment, material or supplier makes part of the plan obsolete, and every deviation that got through points to what was missing — a forgotten characteristic, a sample too far apart, or a course of action that did not work.

Is the control plan reserved for industry?

No, it applies to any process whose result one wants to maintain, including administrative ones: lead times, completeness rates or rework are monitored there rather than dimensions. The structure does not change — what, how, how often, by whom, and what to do in case of deviation — and it is the last column that is most often missing there.