Control plans: the standing agreement on what is checked and what happens on a fail

One row per characteristic, and a reaction that names an action

Reaction entries that read investigate or inform quality are decoration. A usable entry says what to do with the machine, what to do with the material already made, who to tell and what evidence to keep, in that order, in language an operator can follow at three in the morning. Writing it forces useful arguments into the open: does the line stop or finish the run, how far back does suspicion extend, is rework permitted without further approval. Settling those questions once at a desk is far cheaper than settling them repeatedly under pressure with a delivery at risk.

A frequency is a staffing commitment, not an aspiration

Whatever rhythm appears in the plan becomes a labour requirement, and plans routinely promise more checking than the plant is resourced to perform. The result is worse than a modest plan honestly executed: records get completed retrospectively, checks are skipped on busy shifts, and the document becomes something to be managed around rather than followed. Before signing, multiply each entry by the running hours and see whether the total is a job somebody actually has. If it is not, either fund it or write down what you will genuinely do, and be prepared to explain that choice to a customer.

The plan and the process have to change on the same day

Control plans go stale in predictable ways. A fixture is modified, an operation is moved to a different machine, a gauge is replaced with a better one, a supplier changes and the incoming characteristic shifts. Each of these invalidates rows in a document nobody thought to open. Tie plan revision to the change control that already governs tooling and process, so the same approval that authorises the change also releases the updated plan. Where the two live in separate systems with separate owners, expect the drift, and expect to discover it during a customer audit rather than beforehand.

Why the operator has to be in the room when it is signed

Plans written entirely in an office describe checks that cannot be performed as described: the feature is unreachable once the part is on the machine, the gauge does not fit past the fixture, the sample cannot be taken without stopping a cycle, or the measurement takes longer than the operation it monitors. Whoever will actually do the work should walk the plan at the machine before approval and be allowed to object. Ten minutes of that saves months of quiet non-compliance, and it turns the document into something the floor considers theirs rather than something imposed on them.

Using the plan as evidence when a customer arrives

In an audit or a complaint investigation the plan is the first document requested, because it states what you claimed you would do. From there the auditor moves to whether you did it: completed records at the stated rhythm, gauges listed and in calibration, reaction entries matched by real events in the nonconformance log. The damaging finding is not a gap in the plan but a mismatch between plan and practice, since that calls the whole system into question. Ambitious documents are consequently a liability. A modest plan that is genuinely followed inspects better than a thorough one that is not.

Frequently asked questions

How detailed should a control plan be?

Detailed enough that somebody unfamiliar with the part could run the checks correctly, and no longer than that. Every row costs labour forever, so rows should exist because a characteristic genuinely needs watching, not because a feature appeared on a drawing. Plants with plans running to many pages usually have a handful of rows doing real work and the rest generating paperwork, which dilutes attention and makes the honest rows harder to see.

Do we need separate plans for prototype and volume production?

Usually yes, because the questions differ. Early production is about learning how the process behaves, so measurement is broad and frequent even where it will never be justified later. Steady production narrows the effort to characteristics that have proved capable of moving. Running the launch plan indefinitely wastes money; jumping straight to the lean version means reducing checking before anyone knows which features are troublesome. Plan the transition deliberately, with evidence deciding what gets dropped.

Who approves changes to a control plan?

At minimum the process owner and whoever holds quality authority for the part, and for many customer programmes the customer as well, since the plan often forms part of what was agreed at approval. The practical rule is that anyone who could be surprised by the change should have signed it. Sending a revised plan to a customer after the fact, when a characteristic they care about has quietly stopped being measured, is a conversation worth avoiding.