Quality in Practice — Lesson 6: Control Plan and SPC: Measure What Matters

Quality in Practice · 4 min read · 13.11.2026

The sixth lesson of the Quality in Practice series builds on the FMEA. Last time, you uncovered risks before they materialized. Today we move them from the spreadsheet to the shop floor: you will build a control plan that defines what, who, with what, and how often to check. And we will add SPC — statistical process control — which will announce that something is wrong with the machine before the first defective part comes off the line.

The Control Plan Is the Bridge Between FMEA and the Shop Floor

An FMEA without a control plan is an exercise in Excel. It sits on a drive somewhere. The control plan is that bridge: it picks up the risk characteristics from the FMEA and assigns specific controls to them. In audits, we also see the other extreme — plans with 240 rows that operators never read all the way through.

A good control plan answers six questions for every characteristic:

• What — the dimension or parameter, for example a shaft diameter of 20 ± 0.05 mm.
• With what — a micrometer, a go/no-go gauge, welding-force test pliers.
• How often — every piece, 5 pieces per hour, the first piece after a setup change.
• Who — operator, foreman, quality lab.
• Record — where the value gets written down and who reads it afterwards.
• Reaction plan — exactly what to do when a value is out of line.

Things usually fall apart at the reaction plan. “It gets checked” is not a reaction plan. A reaction plan says: stop the machine, segregate parts made since the last good check, notify the foreman, log the change. A supplier of body components had 240 inspection points, of which operators could realistically handle 20. After sorting by FMEA and complaint history, 18 critical characteristics remained, each with a clear frequency and reaction. Since then, the records have actually been filled in.

SPC: A Chart That Warns Before a Defect Occurs

Final inspection sorts parts into good and bad. SPC guards the process. The difference is fundamental: anyone can sort; only a controlled process can predict. The basic tool is the control chart, such as X̄/R, where you plot the averages and ranges of measurement subgroups.

Watch out for one thing: control limits are not tolerances. They derive from the process's own variability and will tell you the process is running abnormally even while every part is still within tolerance.

An example from a CNC shaft operation with a takt time of 12 seconds. On the third day, the averages on the chart started to climb — seven points in a row above the center line, all within tolerance. The operator stopped the machine, found a worn insert, and replaced it in 15 minutes. An hour later the process would have crossed the upper limit and roughly 200 parts would have fallen out of tolerance. The tooling cost stayed the same. One timely stop made all the difference.

Alongside the chart sits the Cpk index, which tells you how much headroom the process has to the tolerance limit. In the automotive industry, 1.33 is a common minimum for a capable production run. When Cpk drops, variability is growing. Look for the cause the same way as in the 5 Whys lesson.

Measure Few Things, but the Right Ones

The most common mistake is not too few measurements, but too many. When an operator has 40 points to tick off per shift, after three weeks they are ticking blind. From audits we know the tell-tale evidence: a value of 20.02 mm repeats in the record five days in a row without a single deviation. That is not a stable process. That is fiction.

How to choose the characteristics worth measuring:

1. Take a Pareto analysis of complaints and internal defects over the past 12 months — often 3 of 27 dimensions account for 80% of the problems.
2. Cross-reference it with the critical characteristics from the FMEA where high occurrence meets severe consequence.
3. Add the features the customer marked as critical on the drawing.
4. Cover the rest with first- and last-piece inspection per shift, plus inspection after every setup change.

What you measure, you manage. What you don't measure, chance manages.

Fewer measurements also mean cheaper inspection. Eighteen well-chosen characteristics at 5 pieces per hour protect the process better than 40 ticked boxes without a single number.

This Week's Exercise

Pick one critical characteristic from your most important operation. Have 25 consecutive parts measured into a single record — not from five spots in the warehouse, but straight off the running line. Calculate the average and the range (maximum minus minimum) and compare the range with the tolerance. Does it exceed half the tolerance? The process needs to be brought under control, even though nobody has produced a defective part yet. Repeat the exercise two days later and compare where the average has moved. These two entries will tell you more about the process than any annual report.

A control plan and SPC are an investment you can put a number on. That is exactly what the next lesson will tackle, translating a defect into euros and tallying the cost of poor quality (CoPQ). If you want to implement control plans and SPC systematically, take a look at our range of e-learning courses.

Quiz — test your knowledge