Skip to content
Skills / Quality Inspection / Quality Inspection Specialist / SPC Basics: Control Charts, Cpk, and What Your Shop Looks For
Quality Inspection

SPC Basics: Control Charts, Cpk, and What Your Shop Looks For

10 min read Training Guide

Learn why shops chart their processes, what you do at the machine, how to read a chart in plain words, what to do when it signals, and what a Cpk number means.

Table of contents

Statistical process control (SPC) is how a shop watches a process while it runs, so it can catch trouble before bad parts ship. Instead of finding scrap at the end of a shift, you measure a few parts as you go and plot them on a chart. If you work on a machine, a press, a molding line or an assembly station in a shop that uses SPC, the chart will be part of your day. You do not need to be good at math. You need to measure honestly and notice when the chart is telling you something.

What You Will Be Able to Do

  • Explain why shops chart their processes.
  • Follow the steps an operator takes at the machine.
  • Read a control chart in plain words.
  • Say what to do when a chart signals, and what not to do when it does not.
  • Explain what a Cpk number means.

Why Shops Chart

Inspection finds bad parts after they are made. A chart shows a process heading the wrong way while the parts are still good, so you can fix it for far less.

The method goes back to the 1920s, when Walter Shewhart developed the control chart at Western Electric and Bell Labs. It is still in daily use today, and quality systems such as ISO 9001 and the automotive standard IATF 16949 expect shops to control their processes.

The core idea is simple: every process varies. No two parts are identical, because material, temperature and tools all change a little. That normal scatter is called common cause variation, and you should leave it alone. Sometimes something real happens, such as a chipped tool, a slipped fixture or a bad lot of material. That is special cause variation, and it needs action. A chart helps you tell which one you are looking at.

What You Do at the Machine

Your shop has a control plan, a document from the quality group that says what to measure, how to measure it, and how often. That might be every 10th part, once an hour, or at the start of the shift and then at set times. Your tools are the right gauge for the feature, the charting software on a tablet or terminal (or a paper chart), and the reaction plan, which says what to do if the chart signals.

  1. Check the control plan for the feature, the gauge and the sample size.
  2. Take your sample at the right time. Do not skip one because you are busy, and do not save them up.
  3. Clean the part and measure it the same way, in the same spot, every time.
  4. Enter the numbers right away. Do not rely on memory.
  5. Look at the chart as soon as it updates.
  6. Note anything that changed, such as a new tool, a new lot, a new operator or a restart.

Many charts plot an average. To get it, add your readings and divide by how many you took. For 1.002, 1.001, 1.003, 1.002 and 1.001, the sum is 5.009, and 5.009 / 5 = 1.0018. Software usually does this for you.

Reading a Chart in Plain Words

A control chart is a line of dots over time, with three lines across it.

  • Center line: the process's usual average.
  • Upper and lower control limits: where the process normally stays. They are calculated from the process's own data.

Control limits are not the tolerance on the print. The print tells you what the customer needs. Control limits tell you what your process normally does. A dot can be inside the print tolerance and still be outside the control limits, and that is still a signal.

Software flags most of these, but learn the patterns yourself.

What you see In plain words A likely reason
A dot past a control limit Something just happened Broken tool, slipped fixture, wrong material, a mismeasure
A long run of dots on one side of center The process has shifted New lot, new tool, changed setting
Dots climbing or falling in a steady line The process is drifting Tool wear, machine warming up
Dots spreading out wider The process got less consistent Loose fixture, worn part in the machine
Dots bouncing randomly around center Normal. Leave it alone Common cause variation

Which patterns count as a signal, and how many dots it takes, is set by your control plan.

When the Chart Signals

When a chart signals, follow your reaction plan. It usually goes like this:

  1. Stop and tell your lead, if the plan says to stop. Do not keep making parts and hope.
  2. Check the measurement. Measure the part again, and check the gauge. A bad reading is a common cause of a false alarm.
  3. Hold the parts made since the last good check, so they can be sorted if needed.
  4. Look for the cause: dull tool, worn fixture, new material lot, temperature change, an operator change, or a gauge out of calibration.
  5. Fix the cause, not the number. Do not start up again until the cause is found or your lead clears it.
  6. Write it down. Note what you found and what you did, right on the chart or in the system.

The opposite matters just as much. If the chart shows no signal, leave the machine alone. Making a small adjustment every time a part reads a little high or low is called tampering, and it makes the process vary more, not less.

Cpk in One Paragraph

Cpk is a single number that tells you how well a process fits inside the print tolerance, counting both how wide it runs and how well it is centered. A bigger number means more room between the process and the nearest limit. A Cpk of 1.0 means the process just barely fits, with no room to spare. A Cpk of 1.33 is a common goal in many shops, because it leaves comfortable margin for drift. Below 1.0, some parts will be out of spec. Your customer or quality group sets the target for your job, and some set it higher. You will not usually calculate Cpk at the machine, but when it starts to fall, your process is losing its margin, even if every part still passes.

What Can Hurt You on Day One

Charts do not hurt people. Getting the sample can.

Warning: Never reach into a running machine to grab a sample. Stop the cycle or use the outfeed your procedure shows you. Follow your machine's guarding and lockout rules, every time.

  • Parts can be hot and edges can be sharp, so use gloves that suit the job.
  • Do not rush a measurement to keep up with the machine. If you cannot keep up, tell your lead.

Some mistakes cost the shop instead:

  • Making up numbers. Filling in a chart at the end of a shift, or writing what you expected, hides real problems. It is also the fastest way to lose trust.
  • Ignoring a signal because parts are in spec. The chart's job is to warn you early.

Try It Yourself

Pick something you can measure each day, such as your commute time, for 20 days. Plot each day and draw a line at the average. Can you see a drift, a shift or a sudden jump? Spotting those patterns is the real skill.

Check Your Understanding

1. Your parts are all in tolerance, but the chart flags a dot past the upper control limit. What do you do?

Treat it as a real signal. Follow the reaction plan and tell your lead. Being in tolerance does not cancel a signal.

2. The last part read a little low. The chart shows nothing unusual. Should you bump the offset?

No. That is normal scatter. Adjusting would be tampering, and it adds variation.

3. A coworker says a Cpk of 0.9 is fine because the last few parts passed. Is it?

No. A Cpk below 1.0 means the process is expected to make some parts out of spec, even if the last few were good.

Key Takeaways

  • SPC watches a process as it runs, so problems show up before bad parts ship.
  • Common cause variation is normal. Special cause variation needs action.
  • Control limits come from the process, and tolerances come from the print.
  • When it signals, follow the reaction plan. When it does not, leave the process alone.
  • Cpk sums up how well the process fits the tolerance, and 1.33 is a common goal.