Control charts are one of the most useful tools in quality engineering. They are also one of the most misread. Not because they are complicated, but because people overthink them. This is a practical guide to reading control charts. No theory-heavy textbook stuff. Just the patterns you will actually see and what to do about them. ## What a Control Chart Actually Tells You A control chart answers one question: is this process behaving consistently? It does not tell you if the process is good. It does not tell you if you are meeting spec. It tells you whether the variation you are seeing is random (common cause) or whether something specific changed (special cause). That distinction matters because: - Common cause variation is baked into the process. You can only reduce it by changing the process itself. - Special cause variation comes from something identifiable. A new operator, a bad material lot, a machine that needs maintenance. If you react to common cause variation as if it were special cause, you will make things worse. That is called tampering, and it increases variation instead of reducing it. ## The Anatomy of a Control Chart Every control chart has the same basic structure: - **Center line (CL)** the process average - **Upper control limit (UCL)** CL + 3 sigma - **Lower control limit (LCL)** CL - 3 sigma - **Data points** your measurements, plotted in time order The control limits are calculated from your data. They are not spec limits. This is the single most common source of confusion. Spec limits come from your customer or your engineering drawing. Control limits come from your process. They may have nothing to do with each other. ## The Rules Most organizations use the Western Electric rules (also called the Nelson rules). Here are the ones that matter most: **Rule 1: One point beyond 3 sigma.** This is the obvious one. A single point outside the control limits. Something happened. Go find out what. **Rule 2: Nine consecutive points on one side of the center line.** The process has shifted. It might still be within limits, but it is no longer centered where it was. This often shows up gradually and gets missed. **Rule 3: Six consecutive points steadily increasing or decreasing.** A trend. Something is drifting. Tool wear, temperature creep, material degradation. Catch it before it goes out of limits. **Rule 4: Two out of three consecutive points beyond 2 sigma (same side).** This is subtler. Any individual point in the 2-3 sigma zone is fine on its own. But two out of three in that zone is unlikely by chance alone. ## What to Do When You See a Signal This is where most guides stop. "Your chart shows a signal! Investigate!" Great. But investigate how? Here is a practical approach: 1. **Mark the point.** Note the time and date. Do not erase it or adjust it. 2. **Check the obvious stuff first.** Was there a shift change? A new material lot? Did someone bump a machine? Did the measurement system change? 3. **Talk to the operator.** They usually know. "Oh yeah, the coolant pump was acting up that morning." Most special causes are found within 5 minutes of asking the people doing the work. 4. **Document what you find.** Even if you cannot fix it right now, write down what caused it. Over time, these notes become your improvement roadmap. 5. **Do not adjust the process for common cause variation.** If nothing unusual happened and the point is just near a limit, leave it alone. The worst thing you can do is tweak settings every time a point looks high. ## Common Mistakes **Using spec limits as control limits.** Your chart will look great (everything "in control") but you will not be able to detect actual process changes. Always calculate control limits from your process data. **Recalculating limits too often.** Control limits should be recalculated when you make a deliberate process change, not every time you add new data. If you recalculate constantly, you will never see a shift because the limits shift with it. **Ignoring the rules beyond Rule 1.** A process can be deteriorating for weeks with runs, trends, and zone violations while never actually breaking a control limit. By the time a point goes out, it has been going wrong for a while. **Charting the wrong thing.** If you are running subgroups, you need both an average chart and a range chart (or S chart). Looking at averages alone hides changes in variation. Looking at ranges alone hides shifts in the mean. ## Which Chart Type to Use This trips people up but the decision is straightforward: - **Individual measurements** (one reading at a time): I-MR chart - **Subgroups of 2-10**: X-bar R chart - **Subgroups larger than 10**: X-bar S chart - **Counting defectives** (pass/fail on each unit): P chart or NP chart - **Counting defects** (multiple defects possible per unit): C chart or U chart If you are not sure, start with I-MR. It works for most situations and does not require subgrouping. ## Building the Habit The value of control charts is not in any single chart. It is in the habit of plotting data over time and paying attention to it. A chart on the wall that nobody looks at is decoration. A chart that gets updated every shift and reviewed every morning is a management tool. Start with one process. The one that gives you the most headaches. Chart it for a month. You will learn more about that process in 30 days of charting than in the previous year of reacting to problems after the fact.