Control Plans: How to Make Process Improvements Stick
Seven elements that keep a gain in place after the project team leaves: control plan, SPC, reaction plans, standard work, layered audits, error-proofing and a proper handover.
Key takeaways
- A project is not finished when the pilot works. It is finished when the gain survives every shift, every operator and every maintenance cycle without the project team.
- A control plan is more than a table. It joins seven elements: the plan itself, SPC, reaction plans, standard work, layered audits, error-proofing and handover.
- Control limits come from the process, specification limits come from the customer. Keep them separate, and react to signals, not to noise.
- Name an owner, set follow-up audits at 3, 6 and 12 months, and let finance validate the savings. Without these, gains quietly decay.
01Why gains fade after the project closes
Every plant has a graveyard of improvements that worked for three months. The pattern is almost always the same: the project team measured a good result in the pilot, wrote a report and moved on. Nobody owned the new way of working, new operators were trained informally, a maintenance item was never added, and the first time production pressure rose, the old shortcuts returned. For the statistical background, see statistical process control on Wikipedia.
Control is the last phase of DMAIC and the most neglected. In Part 1 we described it in one paragraph, and Part 2 showed a control plan excerpt. This part builds the whole system.
Equipment and tooling wear
Parameters slowly leave the optimum. Without monitoring, nobody sees it until scrap rises.
New people, informal training
The new method lives in heads. When the people change, it changes with them.
Production over standards
Under time pressure, steps that “cost time” are skipped first. If skipping is invisible, it becomes normal.
021. The control plan itself
A control plan is a living document that lists, for every critical step of a process, what is controlled, how, how often, by whom, and what happens when it goes wrong. It is built from the outputs of the earlier phases: CTQs from Define, the capability baseline from Measure, verified causes from Analyze and risks from the FMEA in Improve.
| Field | What to write | Example |
|---|---|---|
| Process step | The operation where the characteristic is created | Bore finishing |
| Characteristic | Product or process parameter that matters | Bore diameter |
| Specification | Target and tolerance | Ø 25.000 mm, +0.021 / 0 |
| Measurement method | Gauge, method and calibration interval | Air gauge, calibrated weekly |
| Sample size and frequency | How many, how often | 5 parts every hour |
| Control method | Chart, check or error-proofing | X̄-R chart, Cpk ≥ 1.33 |
| Reaction plan | What to do if the control signals | OCAP-12: stop, check tool offset, sort back to last good check |
| Owner | A named role, not “the team” | Shift leader |
032. SPC: monitor the process, not the parts
Statistical process control charts the process over time and separates signals (special causes) from noise (common causes). Their value is early warning: a chart shows a shift hours before parts go out of tolerance.
Choosing the chart
| Data type | Subgroup or situation | Chart |
|---|---|---|
| Variable (measured) | Individual readings, n = 1 | I-MR (individuals and moving range) |
| Variable | Subgroups of 2 to 8 | X̄-R |
| Variable | Subgroups larger than 8 | X̄-s |
| Attribute, defective units | Varying sample size | p chart |
| Attribute, defective units | Constant sample size | np chart |
| Attribute, defects per unit | Constant area of opportunity | c chart |
| Attribute, defects per unit | Varying area of opportunity | u chart |
When to react
| Rule | Signal |
|---|---|
| 1 | One point beyond 3σ from the center line |
| 2 | Nine points in a row on the same side of the center line |
| 3 | Six points in a row steadily increasing or decreasing |
| 4 | Fourteen points in a row alternating up and down |
Nelson rules. Use the first two or three as a start, adding more rules raises the false-alarm rate.
043. Reaction plans (OCAP)
A chart without a reaction plan is decoration. An out-of-control action plan (OCAP) tells the operator exactly what to do when the chart signals, in order of likelihood of cause.
Stop and contain
Stop or slow the process as defined, tag product made since the last good check and keep it separate.
Check likely causes
Work through a short list of causes ordered by probability. Take it from the fishbone and FMEA.
Correct
Apply the defined adjustment. Anything beyond the operator’s authority goes to the shift leader or maintenance.
Verify
Take a confirmation sample. Resume only when the chart is back in control.
Record and escalate
Log the event and cause. Two activations of the same OCAP in a week are escalated to engineering.
054. Standard work and training
The improvement has to become “the way we do it”. That means updating the standard work, training everyone who runs the process, and making deviations visible.
- Update the work instruction with the new method, key points and the reason for each key point. People follow what they understand.
- Train on all shifts and record sign-off on a skills matrix. No sign-off, no solo operation of the changed process.
- Define who may change the standard and how. Without this, every shift adjusts it a little.
- Plan for resistance. Involve operators in writing the standard and use a structured approach such as the ADKAR model and the guidance on resistance to change.
065. Layered process audits and 5S
Layered process audits (LPA) put short, frequent checks at several management levels. Their purpose is not to catch people, it is to make sure that the standard is visible to leadership every week.
| Level | Frequency | Duration | Focus |
|---|---|---|---|
| Operator | Every shift start | 2–5 min | Start-up checklist, gauges, parameters |
| Team leader | Daily | 10 min | Standard followed, charts current, OCAP used |
| Supervisor | Weekly | 20 min | Audit trend, open actions, training status |
| Plant manager | Monthly | 30 min | Gemba walk, sample of audits, resources |
| Leadership team | Quarterly | Review | Trend of audit results and repeated findings |
Combine LPA with a regular 5S audit. A clean, organized workplace makes deviations visible: a tool without its shadow, a gauge past calibration, an unlabeled bin.
076. Error-proofing and visual management
The most reliable control is the one that does not depend on attention. Poka-yoke either prevents the error or detects it before the defect moves on.
| Method | How it works | Example |
|---|---|---|
| Contact | Shape, size or guide allows only the correct orientation | Asymmetric fixture, keyed connector |
| Fixed value | Checks that the correct number of parts or steps was done | Counter on fasteners, kit with exact quantity |
| Motion step | Enforces the correct sequence | Interlock: machine starts only after clamp is closed |
Add visual management to make the state of the process readable at a glance: parameter windows marked in green, the chart at the workstation rather than in an office, andon signals for out-of-control conditions.
087. Handover, closure and follow-up
Close the project only when the exit criteria are met. The checklist below doubles as the agenda of the closing review.
| Exit criterion | Evidence |
|---|---|
| Control plan approved and in use | Signed document at the workstation |
| Standard work updated and trained | Sign-off for all operators on all shifts |
| SPC and reaction plans running | Charts current, OCAP used in at least one event |
| Process owner named and accepted | Name and role in the control plan |
| Monitoring period completed | 8 to 12 weeks within target |
| Savings validated | Finance confirmation of the figures |
| Lessons learned documented | One page shared with other teams |
| Follow-up scheduled | Audits at 3, 6 and 12 months in the calendar |
The follow-up audits decide whether the project really worked. At each one, check the result metric (scrap, Cpk, OEE), the audit trend, the number of OCAP activations and whether the standard is still current.
FAQFrequently asked questions
How often should a control plan be reviewed?
Whenever the process, equipment, material or a customer complaint changes the risks, and at least once a year as a common practice. Many customer standards require review after any such change.
What is the difference between a control plan and an FMEA?
The FMEA identifies what can go wrong and how severe it is. The control plan describes how each risk is prevented, detected and reacted to. The FMEA is the input, the control plan is the output.
Do I need an SPC chart for every characteristic?
No. Use SPC for critical characteristics, high-RPN failure modes and characteristics with limited capability. For others, a check or an error-proofing device is often enough.
Who should own the control plan?
The process owner, usually the production or area manager. The project lead creates the plan and hands it over. A plan owned by “quality” alone is rarely followed on the shop floor.
Part 4 of the Practical Six Sigma series
You now have the full chain: framework, a worked case, the statistics behind capability and the system that keeps gains in place. The series keeps growing, so bookmark the hub for new parts.
Building control plans for your own projects? See whether a Green Belt or Black Belt is the right next step.